Home ESEM conference 1997 | Search
52th ECONOMETRIC
SOCIETY EUROPEAN MEETING
BOOK OF ABSTRACTS
Toulouse, France
August 27 - August 30, 1997
NOTICE
Abstracts are presented in the order of session numbers. Presenters are listed by their entire last names capitalised. The presenters' affiliation and the co-author(s)' names are also provided. The alphabetical list of all presenters is included at the end of the book of abstracts.
Request of Papers
The procedure is as follows. In an area called "Rue Intérieure" on the map, you will find a display of panels with envelopes. Each envelope has the name of the author(s) and the title of one paper. Envelopes follow the order of session numbers. You will find labels with your name and address in the Meeting bag. If interested in receiving a paper, take one of those labels and place it in the corresponding envelope.
Only the author is responsible for mailing you the paper after the Meeting. No papers will be distributed during the Meeting by the Meeting staff.
To the authors
We shall be grateful to you for distributing the requested papers promptly. This task is an important part of the Meeting.
EC1 THE MACROECONOMIC EFFECTS OF FISCAL POLICY
The Sensitivity of Global Model Simulations to Fiscal Closure Rules
Kenneth WALLIS,
The University of Warwick, U.K.
Keith Church, Peter Mitchell, Joanne Sault
Fiscal closure rules that enforce the government's intertemporal budget constraint are an essential requirement in macroeconometric models. This paper analyses the characteristics and performance of the different rules used in three forward-looking global models - MSG2, MULTIMOD and NIGEM - by carrying out simulations under each rule on each model. The rules target incremental interest payments, the debt stock and the deficit ratio, respectively. The results indicate that if the debt stock is an ultimate target, then it should feature as such in the rule. However GDP outcomes on MULTIMOD are remarkably insensitive to the choice of rule.
Keywords: Macroeconometric models, policy rules, fiscal solvency, simulation
EC1 THE MACROECONOMIC EFFECTS OF FISCAL POLICY
Non-Keynesian Effects of Fiscal Contractions: Theory and Applications for Germany
Bernd LUCKE,
Freie Universität Berlin, Germany
At several occasions, the 1980's witnessed non-keynesian effects of fiscal contractions. Most notably, cuts in government spending in Ireland, Denmark and Germany are known to have coincided with increases in private consumption spending. Self-fulfilling expectations about the effects of stabilization policies may explain these and a diversity of opposing experiences. This paper formulates a real business cycle model with equilibrium indeterminacy, which allows for sunspot fluctuations as well as for systematic consumption effects of government policy in either direction. Cointegration tests and Euler equation estimates suggest that the model is approximately in accord with German data. Parameter estimates imply that the non-keynesian experiences are not due to self-fulfilling expectations but to the productivity effects of government-provided infrastructure.
Keywords: Non-Keynesian, sunspots, externalities, equilibrium indeterminacy, real business cycle,cointegration
EC1 THE MACROECONOMIC EFFECTS OF FISCAL POLICY
Sustainability of Government Deficits in European Countries
Isabelle ARMANVILLE,
Universidad Autonoma de Barcelona, Spain
In this paper, we analyze the financial situation of European governments that are currently candidates for the Economic and Monetary Union (E.M.U.).The results cast doubt on the sustainability of the fiscal policies in most countries, unless drastic changes do occur. Most of the countries may have potential problems in managing their debt per G.D.P. ratio. We also arrive at pessimistic conclusions concerning the candidate members' ability to satisfy the fiscal requirement established in the Treaty of Maastricht.
Keywords: Deficit sustainability, Maastricht fiscal criteria
EC2 HYPOTHESIS TESTS
A LIL for m-Estimator and Application to Hypothesis Testing with Nuisance Parameters
Filippo ALTISSIMO,
Bank of Italy, Italy
Valentina Corradi
The purpose of this paper is twofold: on one side we aim to provide easily cormputable almost sure bonds for inference in presence of nuisance pararneters uniderntified under the mull. On the other hand, more generally. we airn to provide a flexible completely consistent procedure for inference in, possibly mispecified, parametric models, in the case of dependent and heterogeneous observations. With the term cornpletely consistent we mean that the asymptotic size is zero and the asyrnptotic power is one. The small sample behavior of our procedure is analyzed via few Monte Carlo simulations; in particular we consider (i) conditional moment tests. (ii) testing for nonlinearities in the SETAR model. Overall the size approaches zero relatively slowly, while the power approaches one very quickly.
Keywords: law of iterated logarithm, completely consistent, nuisance parameters
EC2 HYPOTHESIS TESTS
On the Sampling Performance of an Inequality Pre-Test Estimator of the Regression Error Variance under LINEX Loss
Alan WAN,
City University of Hong Kong, China
Gary Geng
We consider the estimation of the error variance of a linear regression model where prior information is available in the form of an (uncertain) inequality constraint on the coefficients. Previous studies on this and other related problems use the squared error loss in comparing estimators' performance. Here, we adopt the asymmetric LINEX loss function and derive the risks of the inequality constrained estimator and the inequality pre-test estimator which results after a preliminary test for an inequality constraint on the coefficients. The risks based on squared error loss are special cases of our analysis, and we draw appropriate comparisons.
Keywords: asymmetry, error variance, LINEX loss, inequality constraint, pre-test
EC3 COINTEGRATION (A)
Spectral Principal Components Analysis of Cointegrated Time Series
David HARRIS,
Monash University, Australia
In this paper, we consider statistical inference for cointegrated time series based on principal components methods in the frequency domain. The analysis may be carried out on the first differences or the levels of the time series. Existing methods for testing for cointegration using spectral principal components on the differences are extended, and estimation of cointegrating vectors is also considered. The same problems are then addressed using spectral principal components on the levels of the time series. We provide an asymptotically efficient estimator of the cointegrating vectors, tests of hypotheses on the cointegrating vectors, and a test for cointegration. Theoretical advantages of the levels analysis over the differences analysis are given, and simulation evidence is also presented.
Keywords: Cointegration, Principal Components, Spectral Analysis
EC3 COINTEGRATION (A)
A Gibbs Sampling Approach to Cointegration
Pierre GIOT,
Université Catholique de Louvain, Belgium
Luc Bauwens
This paper reviews the application of Gibbs sampling to a cointegrated VAR system. Aggregate imports and import prices for Belgium are modelled using two cointegrating relations. Gibbs sampling techniques are used to estimate from a Bayesian perspective the cointegrating relations and their weights in the VAR system. Extensive use of spectral analysis is made to get insight into convergence issues.
EC3 COINTEGRATION (A)
Testing for a Valid Normalization of Cointegrating Vectors in Vector Autoregressive Processes
Pentti SAIKKONEN,
University of Helsinki, Finland
Ritva Luukkonen, Antti Ripatti
This paper develops test procedures which can be used to check the validity of general normalizing restrictions imposed on cointegrating vectors in vector autoregressive processes. Such test procedures are of importance because a proper normalization is required to make cointegrating vectors identified and because imposing a normalization is not innocuous. The tests obtained in this paper are based on the null hypothesis that the employed normalization is valid and they can also be interpreted as tests for the null hypothesis that certain linear combinations of the considered series are not cointegrated. The latter fact explains why the asymptotic distribution theory of these tests and likelihood ratio tests for cointegration is similar. The practical usefulness of the proposed tests is demonstrated by Monte Carlo simulation and an empirical application to interest rate data.
Keywords: Cointegration, Normalization, Vector autoregressive process
EC4 EMPIRICAL HEALTH ECONOMICS
Health, Wealth and Individual Effects - A Panel Data Analysis
Frank WINDMEIJER,
IFS, U.K.
Christian Dustmann
In this paper, we model visits to general practitioners. Doctoral attendance is viewed not as an input in health production, but as a random deviation from an optimally chosen path of health capital. In our model, the stock of health can be shown to depend on income only via the individual specific initial conditions of the marginal values of wealth and health. In the empirical analysis this is taken into account by estimating panel data models for binary and count data that allow for correlated fixed effects. The results show that only the permanent component of income bears an effect on the stock of health.
Keywords: Health Capital, GP Visits, Binary/Count Panel Data, Correlated Fixed effects
EC4 EMPIRICAL HEALTH ECONOMICS
A Hedonic Insight at the Pricing of French Prescription Drugs: Are Prices Really Rising That Slowly?
S. JACOBZONE,
CREST INSEE, France
We analyze a system of fixed prices in the case of the French drugs market. The regulatory fixed price scheme can be partially offset by the introduction of new products. The intensity of the bias may depends on the type of the product, traditional or innovative, and the ability of the regulator to fix a "perfect" price. Empirically, we estimate hedonic prices on a panel dataset (1980-1993) covering a quarter of the market, and including a chemical description of compounds. We find a bias of standard Paasche indices up to 30 % downwards. We find also strong firms effects.
Keywords: price index, price regulation, Heterogeneous commodities, hedonic methods, Fixed effects
EC4 EMPIRICAL HEALTH ECONOMICS
Resource Use and Cost Heterogenity in Swedish Geriatric Rehabilitation
Gunnar LJUNGGREN,
Karolinska Institute, Sweden
Almas Heshmati
Abstract not received from the author
EC5 MODELS OF EXPECTATION
How Certain are Dutch Households about Future Income? An Empirical Analysis
Bas DONKERS,
Tilburg University, The Netherlands
Marcel Das
In this paper we empirically analyze subjective income uncertainty in the Netherlands. Data come from the Dutch VSB panel. We measure income uncertainty directly by asking questions on expected household income in the next twelve months. The growing literature on precautionary saving clearly indicates the need for measurement of income uncertainty. When we compare our results with studies done in the US and Italy, we find that American households are far more uncertain about their future income than households in the two European countries. Further, we find strong relationships between labor market characteristics and the subjective income uncertainty as reported by the heads of households. Information on other expected changes is also taken into account.
Keywords: subjective information, income expectations, income uncertainty
EC5 MODELS OF EXPECTATION
Comparing Predictions and Outcomes: Theory and Application to Income Changes
Marcel DAS,
Tilburg University, The Netherlands
Arthur van Soest
Household surveys often contain questions on intentions or predictions of future behavior or circumstances. If panel data or repeated cross-sections are available, predictions can be compared with actual (future) behavior. This is not straightforward, however, since the prediction will refer to some location measure of the household's subjective distribution, while the outcome is one draw from the actual distribution. Even if subjective and actual distribution coincide, the two are not directly comparable. Manski has studied this problem for the binary case. In this paper, the analysis is extended to the case of more than two (ordered) outcomes. Relations between predicted and actual variables are derived that should hold if subjective and actual distribution are the same. These relations can be tested using panel data. An application to income change expectations illustrates the use of these tests.
Keywords: predictions, categorical data, income growth
EC5 MODELS OF EXPECTATION
Dynamic Factor Models: Estimation and Test of the Number of Factors with an Application to French Business Surveys
Catherine DOZ,
Université de Cergy-Pontoise, France
Fabrice Lenglart
In this paper, we use a factor model to build a composite index which summarizes the information given by French industrial business surveys. We first show that static factor analysis leads to consistent estimators even in a dynamic framework, when the variables under study are stationary. We then use the estimators obtained this way to build an asymptotic test procedure of the number of factors. Finally we estimate again the model through a Kalman filter technique, which also allows us to deal simultaneously with monthly and quaterly data.
Keywords: Dynamic factor models, Kalman filter, composite index, business surveys
EC6 TERM STRUCTURE
Term Structure, Non-Neutral Inflation and Economic Growth: A Three Factor Model
Andrea BERARDI,
LBS, U.K.
In this paper we develop a general equilibrium three-factor model which investigates the mutual relationships between the term structure, inflation and economic growth. The three state variables are the instantaneous real interest rate, the instantaneous expected inflation rate and a zero-mean variable which induces time-varying central tendencies in the other two variables. The model accounts for non-neutral effects of inflation. Real and monetary variables are strictly interrelated and Jointly influence the term structure, while bond yields convey information about expectations of macroeconomic variables, like real growth and inflation. The model has many empirically testable implications which allow us to shed some light on related theories, such as the Fisher relation, the Mundell-Tobin effect, the money-neutrality issue, the predictive ability of the term structure for future inflation and real economic activity. Estimation is performed using a maximum likelihood method where the Kalman filter algorithm is applied to compute the unobservable state variables. The empirical work is based on quarterly data for the United States over the period 1964-1990.
Keywords: Term structure of interest rates, economic fundamentals, stochastic differential equations, time-varying mean, Kalman filter, UC-ARIMA.
EC6 TERM STRUCTURE
The Term Structure of Interest Rates and the Uncovered Interest Rate Parity: Empirical Results for Germany and the US
H. DANKENBRING,
Humboldt-Universität zu Berlin, Germany
Extensive empirical research concerning either the term structure of interest rates or the uncovered interest rate parity has been carried out. Within the framework of a four dimensional vector autoregressive model this paper combines these two approaches in order to investigate whether national or international effects dominate the determination of German long term interest rates. The paper argues that analysing either of the two approaches individually gives misleading results.
Keywords: VEC model, Term Structure, Uncovered Interest Rate Parity
EC6 TERM STRUCTURE
OLS Bias in Tests of the Expectations Hypothesis of the Term Structure
Richard HARRIS,
University of Exeter, U.K.
George Bulkley, Paul Weller
Evidence that the term structure of interest rates does not satisfy the expectations hypothesis has been reported in a number of papers. However, the nature and degree of this rejection depends on whether regression tests are based on short term changes in the long yield or on long term changes in the short yield. In this paper, we identify a source of small sample bias that is consistent with this empirical finding. We assume that agents have some degree of foresight about future changes in the short rate. This induces a correlation between the repressor and the regression error resulting in a downward bias in the estimated coefficients. Using simulation experiments, we that demonstrates that for the long yield regression, this downward bias is substantial. For the long yield regression, the bias is mach less severe.
EC7 CONSUMPTION SMOOTHING
Asset Holding and Consumption Volatility
James BANKS,
IFS, U.K.
Orazio Attanasio, Sarah Tanner
Abstract not received from the author
EC7 CONSUMPTION SMOOTHING
Consumption over the Lifecycle
Pierre-Olivier GOURINCHAS,
Stanford University, U.S.A.
Jonathan Parker
This paper employs a synthetic cohort technique and Consumer Expenditure Survey data to construct average age-profiles of consumption and income over the working lives of typical households across different education and occupation groups. Even after controlling for family and cohort effects, typical consumption profiles are not flat, and seem to track income at young ages. Using these profiles, we estimate a structural model of optimal life-cycle consumption expenditures in the presence of realistic income uncertainty. The model fits the profiles quite well. In addition to providing tight estimates of the discount rate and risk aversion, we find that consumer behavior changes strikingly over the life-cycle. Young consumers behave as "buffer-stock" agents. Around age 43, the typical household starts accumulating liquid assets for retirement and its behavior mimics more closely that of a certainty equivalent consumer. This change in behavior is mostly driven by the life-cycle profile of expected income. Our methodology provides a natural decomposition of saving into its precautionary and retirement components.
EC7 CONSUMPTION SMOOTHING
Wealth and Consumption of the Unemployed: Do they Smooth or are They Too Shocked?
Hans BLOEMEN,
Tilburg University, The Netherlands
Elena Stancanelli
This paper studies the consumption and savings behaviour of individuals that have become unemployed. The life cycle model of consumption predicts that if the unemployment spell is completely expected, and if there are no liquidity constraints, the unemployed smoothes consumptionand runs down assets. The conditions underlying this prediction are unlikely to be met completely. In studying the impact of the unemployment shock on consumption and wealth, it is important to disentangle it into factors that are due to the income decrease (replacement ratio), the uncertainty about becoming unemployed (layoff rate), and liquidity constraints. Policies based on changing the level of unemployment benefits will only work if the unemployed are liquidity constrained as pointed out by Browning and Crossley (1996).
Keywords: Unemployment, Savings, Consumption
EC8 SWITCHING REGIME MODELS
Measuring the Probability of a Business Cycle Turning Point by Using a Multivariate Qualitative Hudden Markov Model
Fabrice LENGLART,
INSEE, France
Stephane Gregoir
We construct a coincident composite indicator of activity with an approach based on the following descriptive principle : a phase of high (resp. low) activity is supposed to begin with the occurence of shocks or unpredictable events whose result drives the activity in a positive (resp. negative) direction. The procedure is in two steps.First, the weak innovation is filtered out of each preselected series through univariate linear modeling, and we take only the innovation signs into account. After this preliminary "coding step", a hidden markov model is used to detect a temporary coincidence and persistence of these signs. The states that the hidden markovian variable may occupy qualify the current phase (favorable or unfavorable) of the activity. The model is estimated on French business survey data ; it provides satisfying results.
Keywords: business cycle, turning point, business survey, Kalman filter, Hidden Markov Model
EC8 SWITCHING REGIME MODELS
Change in Regime and Markov Models
Anders RYGH
University of Oslo, Norway
Abstract not received from the author
EC9 EDUCATION AND TRAINING
The Impact on Training Programs on Duration of the First Job Search Spell
E. RETTORE,
Universita di Padova, Italy
It is discussed how to evaluate the impact of training programs in the italian institutional context. It is argued that the relevant outcome is the duration of the unemployment spell as measured from an intake event (school completion, dislocation from a previous job) to the beginning of a regular job. Some features of the institutional context are exploited to generate restrictions, at least partly testable. Evidence from an application to a sample of students completing high school suggests that the program considered picks up people occupying very disadvantaged positions in the queue and let them get a better position.
Keywords: training program evaluation, unemployment spell, unrepeatable outcome, bounds on impact
EC9 EDUCATION AND TRAINING
Work Orientated Training and the Determinants of Wage Growth
L. CORRADO,
University of Warwick, U.K.
Traditional approaches of the determinants of wage growth, stress the role played by work-oriented training and general education variables. We want to investigate the extent to which any transferability depends on the qualification received, and the extent to which the unobservable components related to such variables impact on earnings growth In the first section competing specifcations of wage growth equation will be examined and panel data estimates will be carried out. In the second section the issue of endogeneity of training and educational dummies will be discussed to establish whether the results of the preceding section can be confirmed.
EC9 EDUCATION AND TRAINING
The Duration of Higher Education: A Dependent Competing Risks Model
Nicole JONKER,
Faculty of Economics and Econometrics, The Netherlands
Hans Van Ophem
In this paper we develop a theoretical and econometric model to analyze the duration of education. Students have the choice to either complete or quit an education. A dependent competing risks approach is used to estimate the factors influencing this decision. The model is estimated on Dutch data.
Keywords: duration of education, dependent competing risks
EC10 LABOUR DEMAND
Labor Contracts and Flexibility: Evidence from the Spanish Labor Market Reform
Cesar ALONSO-BORREGO, ,
Universida Carlos III de Madrid, Spain
Victor Aguirregabiria
In this paper we analyze the effects on employment, job turnover and proportion of defined duration contracts of the 1984 Spanish labor market reform. This reform, similar to the labor market reforms in other European countries during the 80s, was characterized by the elimination of many of the previous restrictions to hire and fire workers with defined duration or temporary contracts. Our analysis is based on the estimation of a dynamic labor demand model and on counter-factual experiments using the estimated model. The model is estimated using an unbalanced panel of 2356 Spanish manufacturing firms between 1982 and 1993 obtained from the Balance Sheets of the Bank of Spain database. In order to reduce the computational complexity associated to the estimation of the model we use random uniform grids for solving the model (Rust 1994), and the GHK simulator to compute the conditional probabilities of the choice histories. Our estimate of the effect of the reform on total employment is positive and significant but with a small magnitude (1.5% increase). It contrasts with a much larger estimate (6.1%) of the increase in employment associated to a reform that cuts off 50% of firing costs for all workers. We also find that the increase in temporary contracts after the reform has not just been the result of lower firing costs for this type of workers but also of their lower relative wage.
Keywords: Labor demand, Firing costs, Dynamic panel data models, Simulation-based estimation
EC10 LABOUR DEMAND
Uncertainty and Labour Demand Rigidity
Marianne PAUCHET,
THEMA, France
Brigitte Dormont
This paper studies the relationships between labour demand, uncertainty and the use of flexible labour contracts. We set up a labour demand model with heterogeneity as regards labour contracts. Short-term contracts and long-term contracts (with linear adjustment costs) can be signed. Our theoretical model predictions are validated when tested on a panel or 915 French firms. Adjustments on short-term contracts are used to compensate smaller reactions or inaction on long-term employment. A higher incertainty reduces the reactions of the hiring rate on long-term contract. More wokers hired on short-term contract improve total labour flexibility, whereas they protect long-term workers from firing.
Keywords: labour demand, flexibility, incertainty
EC11 GENDER WAGE DIFFERENTIALS
Gender Wage Differences in Malaysia: Parametric and Semiparametric Estimation
Marcia SCHAFGANS,
LSE, U.K.
This paper is an empirical study on the labor force in (Peninsular) Malaysia. Parametric and semiparametric estimated wage equations, which correct for sample selection bias, are used to assess the returns to education and extent of gender 'discrimination'. The Andrews-Schafgans (1996) estimator is used to estimate consistently the wage equation intercept, which is needed for the Oaxaca wage decomposition, in the semiparametric case. The results suggest that 'discrimination' favoring men in Malaysia is still quite prevalent, while for Malays (the 'sons of the soil') the strong level of 'discrimination' favoring Malay men is negated by the semiparametric estimation results.
Keywords: Nonparametric sample selection correction, returns to education, gender wage differences, Oaxaca wage decomposition, 'discrimination'
EC11 GENDER WAGE DIFFERENTIALS
The Unequal Distribution of Unequal Pay- An Empirical Analysis of the Gender Wage Gap in Switzerland
Michael GERFIN,
Universität Bern, Switzerland
Dorothe Bonjour
Abstract not received from the author
EC11 GENDER WAGE DIFFERENTIALS
Gender Wage Differences in Spain. A Quantile Regression Approach
Angel LOPEZ,
Universitat Pompeu Fabra, Spain
Jaume Garcia, Pedro Hernandez
In this paper we use quantile regression techniques to estimate wage equations from Spanish data in order to obtain measures of gender discrimination along the wage distributions. Our results suggest that the traditional decomposition of mean observed wages for individuals with mean characteristics is not representative of What happens along the range of wages. In particular, the proportion of wage differentials attributable to discrimination is larger the further we move up the wage distribution.
Keywords: Quantile regression, gender wage discrimination
EC12 GOVERNMENT POLICY AND CAPITAL CONSTRAINTS
The Role of Government in Capital Constrained Economics
Tove EDSTRAND,
Stockholm School of Economics, Sweden
General conclusions regarding the role of government activity in developing countries should be avoided. This study analyses role of government for growth in 71 LDCs. Both government expenditure and financing are considered and the robustness of growth determinants tested. While the relations between growth and most measures of government activity are fragile, public investment is a positive robust determinant of growth, emphasizing the importance of efficient resource allocation of government expenditure. Although increased public borrowing is a negative robust determinant of growth, surprisingly, neither the level of indebtedness nor the debt service ratio are significantly related to growth in LDCs.
Keywords: Growth, developing countries, government, public investment, debt
EC12 GOVERNMENT POLICY AND CAPITAL CONSTRAINTS
Intertemporal Output and Employment Effects of Public Capital
Panicos DEMETRIADES,
Keele University, U.K.
Theofanis Mamuneas
The paper utilizes an intertemporal optimization framework with a rich dynamic structure to study the effects of public capital on output supply and input demands. A system of equations derived from this framework is estimated on a panel of aggregate manufacturing data for twelve OECD economies. We find that public capital has positive long-run effects on both output and input demands in all countries. These effects tend, however, to be smaller in the G-7 countries than in the other countries in our sample. We also find that in the long-run, public capital is a complementary input to labor and a substitute for private capital in all the countries examined.
Keywords: Public capital, output, employment
EC12 GOVERNMENT POLICY AND CAPITAL CONSTRAINTS
Capital Market Integration in the Pacific Basin Region: An Impulse Response Analysis
Kate PHYLAKTIS,
City University Business School, U.K.
We examine the extent of capital market integration in a group of Pacific Basin countries following the deregulation of their markets, and explore whether the influence of Japan in the region has overtaken that of US. Looking at long-run comovements of real interest rates through the use of cointegration, and using impulse response analysis to examine the speed of adjustment of real interest rates to long-run equilibrium following a shock in one of the markets, which is another indicator of the degree of capital market integration, we find that these countries are closely linked with world financial markets and more so with Japan than with US.
Keywords: Capital Market Integration, Impulse Response Analysis
EC13 FINANCIAL ECONOMETRICS (A)
'Separable Nonstationarity' in Speculative Prices during Prolonged Periods of Uncertainty
Elena ANDREOU,
University of Manchester, U.K.
Nikitas Pittis
The paper proposes an alternative framework for studying the time heterogeneity properties of financial variables. Assuming that these variables follow a generalized Wiener process, a more general framework of nonstationarity arises (namely, the separable type nonstationarity) through a particular parameterization of the covariance matrix, which accommodates the unit root as a special case. The resulting univariate autoregressive model provides a conditional mean with time-invariant parameters and a homoskedastic conditional variance, in the sense that it is free of the conditioning variables, which is, however, a linear function of time. An interesting feature of these models is that the implied nonstationarity can coincide with the autoregressive coefficient being less than one, as opposed to the unit root case. Therefore, an identification factor between the separable nonstationary models and the difference or trend stationary ones, is the time dependence of the conditional second moment. An empirical illustration is provided for the pre-realignment behavior of daily EMS exchange rates that went through prolonged periods of increasing uncertainty.
Keywords: unit root, separable nonstationarity, time-variant conditional variance, exchange rates
EC13 FINANCIAL ECONOMETRICS (A)
Estimation of Jump-Diffusion Processes based on Indirect Inference
George JIANG,
University of Groningen, The Netherlands
Jump-diffusion processes have been widely used in the finance literature to model the dynamics of stock returns and currency exchange rates to reflect discontinuities of the sampling paths. Available literature generally agrees upon the fact that jumps are important components of asset returns and in some cases can count partly for the option pricing errors of the Black-Scholes model. However, difficulties involved in the identification and estimation of general jump-diffusion processes have prevented their implementation in empirical applications and testing. This paper proposes the estimation of generalized parametric continuous-time jump-diffusion processes, from discretely observed data, based on indirect inference through discrete-time instrumental or auxiliary models. A simple Monte Carlo experiment is performed to examine the statistical properties of the indirect estimators and to show that the asymptotic biases caused by discretization can be removed. Applications of the indirect estimation method are also undertaken to various currency exchange rate models to illustrate its implementation on the one hand, and to present interesting results on the other.
Keywords: Jump-Diffusion, Indirect Inference, Exchange Rate Dynamics
EC13 FINANCIAL ECONOMETRICS (A)
Conditional Means of Time Series Processes and Time Series Processes for Conditional Means
Gabriele FIORENTINI,
Universidad de Alicante, Spain
Enrique Sentana
We study the processes for the conditional mean and variance given aspecification of the process for the observed time series. We derive general results for the conditional mean of univariate and vector linear processes, and then apply it to various models of interest. We also consider the joint process for a subvector and its expected value conditional on the whole information set. In this respect, we derive necessary and sufficient conditions for one of the variables in a bivariate VAR(1) to have a white noise univariate representation while its conditional mean follows an AR(1) with a high autocorrelation coefficient.We also compare the persistence of shocks to the conditional mean relative to the observed variable using measures of total and iterim persistence of shocks for stationary processes based on the impulse response function. We apply our results to post-war US monthly real stock market returns and dividend yields. Our findings seem to confirm that stock returns are very close to white noise, while expected returns are well represented by an AR(1) process with a first-order autocorrelation of .9755. We also find that small unexpected variations in expected returns have a large negative immediate impact on observed returns, which is thereafter compensated by a slowly diminishing positive effect on expected returns.
Keywords: Time Series Processes, Conditional Moments, Expected Returns, Persistence
EC14 ECONOMETRIC THEORY
Non-normal Variation and Regression to the Mean
Andrew CHESHER,
University of Bristol, U.K.
Non-normal variation across repeated measurements leads to non-linear and heteroskedastic regression to the mean unlike the simple linear and homoskedastic regression to the mean found in normal models. This paper investigates the nature of the regression to the mean phenomenon in non-normal settings using (a) small variance approximations and (b) exact results obtained using normal mixtures to approximate non-normal distributions.
Keywords: regression to the mean, small variable approximations, measurement error models
EC14 ECONOMETRIC THEORY
Relative Efficiency in Equivalence Classes of Asymptotic Covariances
Carlos MARTINS-FILHO,
Oregon State University, U.S.A.
D. Mandy
Generally, the investigation of asymptotic efficiency in econometrics makes restrictive convergence assumptions on certain
sample moments. To alleviate this problem a generalized asymptotic covariance has been proposed to investigate asymptotic
efficiency. This paper establishes new sufficient conditions for the generalized covariance to be an equivalent class, and then
proposes efficiency concepts that are compatible with the class. Conclusions from the new concepts are weaker than conclusions
from existing efficiency concepts. Thus, all known efficiency conclusions based on the generalized covariance hold for the new
concepts, and additional conclusions can be obtained in some cases.
Keywords: Asymptotic Efficiency
EC14 ECONOMETRIC THEORY
Union-Intersection and Sample Split Methods in Econometrics with Applications to Mand Sure Models
Olivier TORRES,
Université de Lille 3, France
Jean-Marie Dufour
We develop inference procedures for two apparently distinct classes of situations: first, problems of pooling information from several samples whose stochastic relationship is not specified; second, problems where the distributions of standard test statistics are difficult to assess, while it is possible to obtain more tractable distributional results for statistics based on appropriately chosen subsamples. A large number of econometric models lead to such situations. We propose a general approach which uses union-intersection techniques. This approach is easy to apply and transposable to a wide spectrum of models. In addition to being robust to various misspecifications, the approach studied turns out to have good power properties with respect to other available techniques.
Keywords: union-intersection, combination of information, MA processes
EC15 COINTEGRATION AND INFERENCE
Priors, Posterior Odds and Lagrange Multiplier Statistics in Bayesian Analyses of Cointegration
Richard PAAP,
Erasmus University Rotterdam, The Netherlands
Frank Kleibergen
The traditional Bayesian analysis of models containing locally nonidentified parameters leads to ill-behaved marginal posteriors. These models are nested in embedding models whose parameters have well-behaved posteriors such that a controvercy exists between the traditional Bayesian analyses of these models. By defining models as parameter restrictions on more general embedding models, we show that a highly informative prior is imposed on the parameters of the embedding model when we do not explicitly analyze the model containing the locally nonidentified parameters as a parameter restriction of the embedding model. This explains the ill-behaved marginal posteriors. A coherent framework for conducting Bayesian analysis results when models are explicitly analyzed as parameter restrictions of embedding models. We illustrate this using some commonly used statistical models, i.e. linear, ARMA, cointegration and simultaneous equation models.
EC15 COINTEGRATION AND INFERENCE
The Role of Stationary Regressors in the Cointegration Test
Anders RAHBEK,
University of Copenhagen, Denmark
Rocco Mosconi
The issue of including stationary explanatory variables in VAR models is addressed. It is shown that this will lead to nuissance parameters in the asymptotic distribution of the trace statistic for cointegration rank. The nuissance parameters are characterized as canonical correlations between the accumulated common trends and the accumulated level of the process analysed for cointegration. In particular the trace test is not similar and an alternative way to include explanatory variables, which will lead to similarity, is discussed. This is achieved by analyzing the extended VAR model where the (stationary) explanatory varibles enter the system not only in levels, but also cummulated in the cointegration space.
Keywords: Cointegration, Asymptotic distributions, Stationary regressors
EC15 COINTEGRATION AND INFERENCE
Persistent Granger Non-Causality, with an Application to International Links between Long-Term Interest Rates
Catherine BRUNEAU,
Banque de France, France
Eric Jondeau
Abstract not received from the author
EC16 MONETARY ECONOMICS
Understanding a Monetary Conditions Index
Eilev JANSEN,
Norges Bank, Norway
N. Ericsson, R. Nymden
Numerous central banks, governmental organizations, and businesses now calculate a Monetary Conditions Index (or MCI) as an indicator of the stance of monetary policy. Two central banks, those for Canada and New Zealand, use their MCIs as operational targets. This paper describes and defines this concept, summarizes how central banks implement MCIs in practice, reviews some of the operational and conceptual issues involved, and evaluates the sensitivity of MCIs to an inherent source of uncertainty in their calculation. Empirically, this uncertainty typically renders MCIs uninformative for their ostensible purposes, so some possible alternatives are briefly considered.
Keywords: inflation, MCI, monetary conditions index, monetary policy, output, uncertainty
EC16 MONETARY ECONOMICS
Monetary Transmission in a Small, Open Economy: A Structural VARX Analysis for the Netherlands
J. KAKES,
University of Groningen, The Netherlands
J. Jacobs, H. Van Ees
Abstract not received from the author
EC16 MONETARY ECONOMICS
An Empirical Analysis of the Fed's Open Market Operations
Steven WEI,
CORE, Belgium
Abstract not received from the author
EC17 INVENTORY MODELS/INVESTMENT
Balladurette and Juppette: A Discrete Analysis of Scrapping Subsidies
Jérôme ADDA,
INRA, France
Russell Cooper
This paper studies the effects of subsidies on durable goods markets. In particular, we study a recent policy in France in which the governments of Balladur and Juppé subsidized the replacement of old cars with new ones. To study this policy, we construct a dynamic stochastic discrete choice model of car ownership at the household level. The resulting decision rules and equilibrium conditions are used to estimate, using aggregate data, the underlying parameters of the model. These policy functions are used to evaluate the short and long run effects of the French policies. We find that these policies do stimulate the automobile sector in the short run but, through the induced changes in the cross sectional distribution of car ages, create the basis for subsequent low activity. Further, while these policies increase government revenues in the short run, revenues in the long run are lower relative to a baseline without intervention.
Keywords: Discrete choice, aggregation, estimation, heterogeneity
EC17 INVENTORY MODELS/INVESTMENT
Inventories and Asymmetric Business Cycle Fluctuations in the UK: A Structural Approach
Marianne SENSIER,
University of Oxford, U.K.
Abstract not received from the author
EC18 ENVIRONMENTAL ECONOMICS
What Explains the Environmental Kuznets Curve? The Analysis of Environmental Efficency
Osman ZAIM,
Bilkent University, Turkey
Fatma Taskin
The role of the environment is an important issue in the policy making and the accurate assessment of the environmental conditions is vital. In this paper using non-parametric techniques, we propose to develop an environmental quality index for each of the OECD countries that would allow one both to do cross section comparisons on the state of each country's production process in its treatment of undesirable outputs and also to trace each country's modification of their production processes in response to changes in societies' environmental consciousness. Furthermore in this study we try to understand the factors underlying the societies' environmental consciousness that eventually leads to changes in the environmental efficiency. The results provides further empirical evidence for the environmental Kuznets curve hypothesis.
Keywords: Environmental quality index, environmental Kuznets curve, non-parametric efficiency measurement
EC18 ENVIRONMENTAL ECONOMICS
Improving Revelation of Willingness to Pay for Natural Assets Application to Biodiversity
Caroline GAUTHIER,
Université de Toulouse, France
Brigitte Desaigues
The 1992 Earth Summit recommanded to safeguard biodiversity. Contingent valuation allows to reconstruct biodiversity use and non-use benefits. valuing biodiversity is complex because of individual non-familiarity with that good and biases the contingent method generates. In this paper, we present a pilot study on the french Garonne river forests biodiversity. We give principal econometric estimations of this analysis. We focuse on conclusions concerning the embedding effect and the problem of the open versus closed-ended format for the WTP revelation question. We show that the relevance of the method to value biodiversity can be improved.
Keywords: Survey methods, Renewable resources and Conservation Management, Environmental Management, General, Demand and Supply
EC18 ENVIRONMENTAL ECONOMICS
Performance Heterogeneity in a Fishery
Frode STEEN,
Norwegian School of Economics, Norway
Frank Asche, Kjell Salvanes
The performance heterogeneity in a fishery is analyzed using an unbalanced panel data set for sealing. We analyze the importance of factors under the fishermen's control such as effort and stochastic factors such as weather conditions and fishing luck. We address the fishermen's possibility of forming well defined expectations by analyzing the development in performance over time. The time series properties of the industry as a whole is compared with the properties of the individual boats. The results indicate that the industry has a mean reverting technology, but that the individual boats are exposed to considerably more stochastics and are therefore more difficult to regulate.
Keywords: Efficiency frontiers, unit-root tests, Monte-Carlo simulations, efficiency decomposition
EC19 EMPIRICAL MODELS OF SOCIAL INTERACTION
Residential Neighborhood Effects
Yannis IOANNIDES,
Tufts University, U.S.A.
We develop a model of residential neighborhood interactions and test it by means of data from the American Housing Survey for 1985, 1989 and 1993. These data are a panel of dwelling units, and they also provide information on the socioeconomic characteristics of their occupants. We explore in greater detail a unique feature of those data, namely the availability of data for neighborhood clusters in different United States metropolitan areas. We perform a great variety of reduced form regressions in order to explain variations in values of homes, appreciation rates, and home maintenance in terms of three different groups of explanatory variables, namely geographical variables and neighborhood (cluster) characteristics, dwelling unit characteristics, and socioeconomic characteristics of occupants. The statistical properties of these regressions are quite satisfactory. Neighborhood-average variables perform quite well in several regressions and imply nonlinear effects associated, in particular, with such variables as cluster-averages for homeownership, race, unit quality and duration of vacancies. Our estimation results with dynamic versions of our models suggest that the averages of decisions made by one's neighbors are more important than the dependent variable's own lagged value. We interpret this as strong evidence in favor of neighborhood effects. Our estimates imply elements of instability in the intertemporal evolution of property valuations, but not in maintenance behavior.
Keywords: Empirical neighborhood effects, housing, dynamic interactions
EC19 EMPIRICAL MODELS OF SOCIAL INTERACTION
Social Interactions, Local Spillovers and Unemployment
Giorgio TOPA,
New York University, U.S.A.
I analyze a model that explicitly incorporates local interactions and allows agents to exchange information about job openings within their social networks. Thus agents are more likely to be employed if their social contacts are also employed. The model generates a stationary distribution of unemployment that exhibits positive spatial correlations. I estimate the model via an indirect inference procedure, using Census tract data for Chicago. I find a significantly positive level of social interactions across neighboring areas. This finding is robust to several controls for sorting and unobserved characteristics. The local spillovers are stronger for areas with less educated workers and higher fractions of minorities.
Keywords: Local Interactions, Indirect Inference, Informal Hiring
EC20 EMPIRICAL INDUSTRIAL ORGANISATION
Demand and Supply: The Relationship between Price Elasticities and Profitability in the UK
Naercio MENEZES-FILHO,
UCL, U.K.
Abstract not received from the author
EC20 EMPIRICAL INDUSTRIAL ORGANISATION
Cost Efficiency and Regulatory Schemes: Evidence from a Panel of Urban Transport Networks
Philippe GAGNEPAIN,
Université de Toulouse, France
Marc Ivaldi
In this article, we deal with the theory of regulation under asymmetric information from an econometric perspective applied to the urban transport industry in France and emphasize the importance of taking into account regulatory schemes when estimating production or cost frontiers. The objective is to interpret contractual relations between a principal (usually a political authority) and a monopoly which runs a urban transport network. Two main contractual mechanisms are used in practice: Fixed-price and cost-plus schemes. The model assumes that each operator possesses private information about its technology and can also reduce operating costs by exerting a certain level of effort, unobserved by the authority. First, we estimate the operating costs and transport demand functions to recover estimates of the unknown efficiency and effort parameters from a panel data set on transport operators for the 1985-1993 period. To the opposite of the usual assumptions chosen in the related literature, we consider that actual regulatory mechanisms are not necessarily optimal. Second, according to those results, we simulate the optimal regulation contract and evaluate the associated allocation. By this way, we are able to measure the discrepancy between practical arrangements and optimal contracts.
Keywords: Regulation, Asymmetric information, Efficiency, Effort, Costs, Welfare
EC20 EMPIRICAL INDUSTRIAL ORGANISATION
Distributional Consequences of Airport Congestion Pricing
J. DANIEL,
University of Delaware, U.S.A.
Abstract not received from the author
EC21 BEHAVIOUR UNDER RISK
The Attitudes Towards Risk of Racetrack Bettors
Bernard SALANIE,
Laboratoire d'Economie Industrielle, France
Bruno Jullien
We investigate in this paper the attitudes towards risk of bettors in British horse races. The model we use allows us to go beyond the expected utility framework and to explore various alternative proposals by estimating a multinomial model on a 34443-race dataset. We find that rank-dependent utility models do not fit the data better than expected utility models. On the other hand, cumulative prospect theory has higher explanatory power. Our preferred estimates suggest a pattern of local risk-aversion similar to that proposed by Friedman-Savage.
Keywords: choice under risk, non-expected utility, risk-aversion, gambling
EC21 BEHAVIOUR UNDER RISK
Risk Sharing by Households With and Across Regions and Industries
Gregory HESS,
Faculty of Economics and Politics, U.K.
Abstract not received from the author
EC21 BEHAVIOUR UNDER RISK
Efficient Farm/Off-Farm Work Allocation and Labor Market Dynamics in Taiwan (1976-1992)
Marc GURGAND,
Université de Paris I, France
We develop a model to test how efficiently agents adapt their labor supply to changing labor market conditions. Optimal allocation should be insensitive to the amplitude of past changes, as measured by potential wage variation. The focus is on rural household labor allocation between farm and off-farm work in Taiwan. Our model relates farm profit to wage variation so as to identify possible inefficiency. This simple reduced form is implemented based on pseudo-panel data : lagged wage and wage variation simulation is proved possible in this empirical setup. The hypothesis that efficient labor reallocation occurs in response to labor market changes cannot be rejected.
EC22 BOOTSTRAP METHODS/NONLINEAR MODELS
Bootstrap Testing in Nonlinear Models
Russell DAVIDSON,
GREQAM, France
James MacKinnon
With nonlinear models, bootstrap testing can be expensive because one needs to perform at least one nonlinear estimation per bootstrap sample. We show that it may be possible to reduce computational costs by performing only a fixed, small number of artificial regressions, or Newton steps, per bootstrap sample. The number of iterations needed is smaller for likelihood ratio tests than for other classical tests. The suggested procedures are applied to tests of slope coefficients in the tobit model, where asymptotic procedures often work surprisingly poorly. In contrast, bootstrap tests work remarkably well, very few iterations being needed to compute them.
Keywords: Bootstrapping, Hypothesis testing, Artificial regressions, Nonlinear models, Tobit model
EC22 BOOTSTRAP METHODS/NONLINEAR MODELS
The Robustness and Reliability of Heteroskedasticity Tests and the Use of the Bootstrap
Chris ORME,
The University of Manchester, U.K.
Leslie Godfrey
Heteroskedasticity tests are widely used in applied work, involving the estimation of muliple regression models, and form part of the routine output of many estimation programs. However, Monte Carlo evidence that is cited to justify their use is derived from studies using simple regression models and/or a test with only one degree of freedom. Although some of the tests appear reliable and robust when studied in this context, more realistic settings with multiple regression models and standard alternatives produce much less convincing evidence. In this paper, the reliability of a number of heteroskedasticity tests is examined, as is their robustness to nonnormality. Since asymptotic theory often fails to provide a useful guide to finite sample behaviour, in this more general setting, the use of the bootstrap is proposed and its effectiveness is evaluated using Monte Carlo experimentation.
EC22 BOOTSTRAP METHODS/NONLINEAR MODELS
The Wild Bootstrap
Emmanuel FLACHAIRE,
Université de la Méditerranée, France
Russell Davidson
In this paper we are interested in inference based on heteroskedasticity consistent covariance matrix estimators, for which the appropriate bootstrap is a version of the wild bootstrap. We show that the wild bootstrap is not well-behaved in the presence of observations with high leverage. We show that the resampling distribution's choice is important and an arbitration is still necessary. We determine the optimal solutions frontiers, and we propose new implementations.
Keywords: HCCME, wild bootstrap, Monte Carlo methods
EC22 BOOTSTRAP METHODS/NONLINEAR MODELS
Learning on Latent Variables for Estimating Nonlinear Structural Econometric Models
Valentin PATILEA,
Université Catholique de Louvain, Belgium
Eric Renault
Many structural econometric models (nonlinear rational expectations, option pricing, auction models, ...) characterize observable variables as highly nonlinear functions of some latent variables.These functions are one-to-one, but they depend on the distribution of the latent variables through the equilibrium of the game and/or the learning process. Therefore numerical complexity of the equilibrium definition generates substantial obstacles for the direct implementation of maximun likelihood inference. Motivated by the fact that the law of motion of the latent variables is often defined in a fairly simpler way, simulation-based strategies (e.g. indirect inference) have been developed recently. Herein we propose alternative estirmation strategies based on recursive learning on the latent variables in order to perform approximatod MLE directly inside the more tractable latent morlel. This leads to build various generalizations of Robbins-Monro recursive estimators. These new methobologies of inference appear to be well-suited for both empirical implementation of nonlinear rational expectation models and bounded rationality modelling.
Keywords: Structural Econometrics, learning, recursive estimation, option pricing
EC23 INEQUALITY AND POVERTY MEASUREMENT
Welfare Judgements in the Presence of Contaminated Data
Frank COWELL,
LSE, U.K.
Maria-Pia Victoria-Feser
The economic analysis of income distribution and related topics makes extensive use of dominance criteria to draw inferences about welfare comparisons. However it is possible that - just as some inequality statistics can be very sensitive to extreme values - conclusions drawn from empirical implementations of dominance criteria may also be influenced by data contamination. We show the conditions under which this may occur and propose empirical methods to work round the problem.
Keywords: Welfare dominance, Lorenz curve, robustness
EC23 INEQUALITY AND POVERTY MEASUREMENT
Atkinson and Bourguignon's Dominance Criteria: Extended and Applied to the Measurement of Poverty in France
Eric MAURIN,
INSEE, France
C. Chambaz
In this paper, we demonstrate that the criteria put forward by Atkinson and Bourguignon (1987) can be satisfactorily applied to the analysis of poverty even when the demographic composition of society and the definition of poverty change over time. We apply these criteria to a set of data tracking annual income distribution movements in France from 1977 to 1994. Atkinson and Bourguignon criteria posit that recent changes in poverty are a function of macroeconomic fluctuations in activity.
Keywords: Sequential dominance, Welfare, Poverty
EC23 INEQUALITY AND POVERTY MEASUREMENT
Transient Seasonal and Chronic Poverty of the Peasants. Evidence from Rwanda
Christophe MULLER,
CSAE, U.K.
Using panel data from Rwanda, we estimate seasonal transient and chronic poverty indices, for different poverty line, poverty indicators, equivalence scale, and with and without the corrections for price variability and for the sampling scheme. We also estimate sampling standard errors for the poverty indices. The worse poverty crises occur after the dry season at the end of the year. Most of the severity of poverty comes from the seasonal transient component of annual poverty, while the seasonal component of the incidence of poverty is much smaller. Thus, the actual differences of the severity of poverty either between developing and industrial countries or between rural and urban areas in LDCs, may be much worse than what is shown by usual chronic annual poverty measures or by measures of seasonal incidence of poverty. The importance of the transient component suggests a need for income stabilisation policy. However, the contribution of the global transient seasonal poverty is important for household clustered around the poverty line but low for the poorest part of the chronically poor. Thus, policies fighting seasonal transient poverty are likely to concern the moderately poor rather than the very poor, as compared with policies against chronic poverty, which affect the very poor. The probability transition analysis across seasonal living standard distributions shows that mobility across quintiles is always very strong. The poverty crisis in the last season is more the result of many peasants falling into poverty than a decrease in the flow out of poverty. A "safety net" policy aimed at the poor and the non-poor at this period would then be appropriate. We estimate censored quantiles regressions both for household chronic and transient seasonal poverty. The agricultural choices of peasants are found to affect differently the two components of annual poverty that could therefore be addressed by a combination of policies specific to each component.
Keywords: Poverty analysis, Agricultural Households in LDCs
EC23 INEQUALITY AND POVERTY MEASUREMENT
Decomposition of the Change in the Distribution of Primary Family Incomes: A Microsimulation Approach Applied to France, 1979-1989
F. BOURGUIGNON,
DELTA, France
M. Martinez
This paper proposes a decomposition analysis of the evolution of inequality of primary family incomes based on the distinction between estimated changes of behavior in the labor market and changes in the distribution of socio-demographic characteristics in the population. This methodology is based on the simultaneous modeling of earnings determinants, labor force participation behavior and unemployment incidence among the members of a family. Applied to the evolution of the equivalized primary household income distribution in France between 1979 and 1994, this decomposition permits to clearly identify some of the various forces whose complex interaction resulted into an unambiguous increase in the inequality of primary incomes throughout the 1980s and the early 1990s.
EC24 EDUCATION AND THE LABOUR MARKET
Using Rank Order as an Instrumental Variable: An Application to the Return to Schooling
Marno VERBEEK,
Katholieke Universiteit Leuven, Belgium
Francis Vella
In many situations there is an unobserved variable whose presence is thought to bias the measured effects of the observed covariates. In this paper we assume that the data can be divided into multiple sections in which the distribution of this variable is identical. Next, we approximate the unobserved heterogeneity by the rank order of an individual in his reduced form error distribution. Through this approach we are able to produce instrumental variables estimators that are consistent for all model parameters in the absence of exclusion restrictions. The approach is illustrated with an empirical study on the returns to schooling.
Keywords: endogenous regressors, instrumental variables, rank order, returns to schooling
EC24 EDUCATION AND THE LABOUR MARKET
Estimating the Returns to Education from a Non-Stationary Dynamic Programming Model
Christian BELZIL,
Concordia University, U.K.
Jorgen Hanson
We estimate an optimal stopping model of the decision between continuing schooling or entering the labor market using a cross-section of Swedish labor force participants. The model has finite horizon and is solved in closed-form using backward recursion methods. The model incorporates unobserved ability at school and in the labor market which can be estimated non-parametrically. The construction of the likelihhod based on education choices and wage data enables us to estimate the wage returns to education as well as the effect of education on non-wage benefits. We also investigate the effects of parental background (education) on schooling
Keywords: Dynamic Programming, Returns to Education, Returns to Experience, Human Capital, Schooling Decisions
EC24 EDUCATION AND THE LABOUR MARKET
Educational Debt, Loan-Forgiveness and Adverse Selection in the Market for Lawyers
Robert SAUER,
Tel-Aviv University, Israel
This paper analyzes the effect of educational debt and loan-forgiveness programs on both law school entry decisions and the subsequent career choices of law school graduates using a dynamic optimization model and data on several cohorts of graduates from the University of Michigan Law School. The study shows that both the returns and costs to a highly-selective legal education are strongly related to parental occupational status. Law school graduates who do not have a professional or manager father are less likely to be assessed highly talented attorneys and thus have lower expected lifetime earnings as well as higher educational debt burdens. The implication is that loan-forgiveness programs will differentially attract less talented attorneys to enter law school and participate more often in the nonprofit sector, suggesting that loan-forgiveness programs may be subject to an adverse selection problem.
Keywords: Adverse Selection, Lawyers
EC24 EDUCATION AND THE LABOUR MARKET
Quality Adjusted Measures of Services in Public Schools
Almas HESHMATI,
Göteborg University, Sweden
This study considers specification and estimation of cost and production functions in public schools. A number of production characteristics are included in the specification to control for and explain the quality differentials as well as producer effectiveness across municipalities in the provision of their school services. Both parametric and non-parametric approaches are used to take into account the quality differences in school services. These approaches are then compared to the alternative output measures without any adjustment for quality. The sensitivity of different elasticities and returns to scale measures with respect to alternative model specifications and quality adjustments are also analyzed. In the empirical section we examine performances of 286 Swedish municipalities in the production of primary and secondary school education during the 1992/31994/5 school years.
Keywords: Economics of education, primary and secondary schools, production and cost functions, output quality adjustment
EC25 FIRMS' FINANCIAL STRUCTURE
Economic Recession and Firms Financing Decisions
Hossein ASGHARIAN,
University of Lund, Sweden
The paper studies determinants of the firm leverage during the economic downturn in Sweden 19891992. I use a panel data framework to study variations in leverage ratios during the recession. The model allows for firm and time effects as well as time varying coefficients. The consistency and efficiency of the estimators regarding problems such as self selection bias and heteroscedasticity are taken into account. Among other things, I find that dividend payments and stock returns affect firms' leverage negatively.
Keywords: Leverage, Recession, Panel data, Sample selection
EC25 FIRMS' FINANCIAL STRUCTURE
Causes and Effects of Financing Constraints at the Firm Level
Peter WINKER,
University of Konstanz, Germany
Asymmetric information is one reason for financing constraints at the firm level. A standard model of credit rationing is used to derive hypotheses on possible causes. Firms' age and size are among the factors which should influence the probability of financing constraints. Furthermore, changing business conditions might strengthen the degree of informational asymmetry. Firm panel data for Germany are used for an econometric analysis of the causes of limited access to the loan market and its real effects. Both, the influence of asymmetric information on financing constraints and real effects of financing constraints are supported by the data.
Keywords: Financing constraints, Asymmetric information, Firm panel, Real effects
EC25 FIRMS' FINANCIAL STRUCTURE
Credit Constraints, Investments and Financial Deregulation: An Euler Equation Approach to Panel Data for Swedish Firms
Sten HANSEN
Uppsala University, Sweden
Using Swedish firm level data, we study the role of capital market imperfections for the investment behavior of different categories of firms. We also investigate whether financial constraints were relaxed by financial deregulation in the mid 80's. The dynamic investment model we estimate is based on the Euler equation governing the optimal path of investments in the presence of convex adjustment costs and agency costs of debt.
Keywords: Capital Markets, Agency Cost, Investment, Financial deregulation
EC25 FIRMS' FINANCIAL STRUCTURE
Skewness of Earnings and the Believability Hypothesis: How Does the Financial Market Discount Accounting Earnings Disclosure?
Murugappa KRISHNAN,
University of Minnesota, U.S.A.
S. Santaraguruswany, H. Shin
When firms attempt to manage their earnings disclosures by presenting evidence selectively, sophisticated inference on the part of financial market participants entails a positive association between the market to book ratio of a firm and the skewness of the distribution of its announced earnings. In this paper, we put this hypothesis to the test, and confirm its main predictions. Our regressions suggest that in setting security prices the financial market's scepticism about how much private information managers have plays a significant role. In addressing a question similar to the empirical discretionary accruals literature in accounting, by using our model of equilibrium, we are able to sidestep the problem of measurement error in measuring discretionary accruals.
Keywords: Earnings skewness, discretionary disclosure, market-to-book
EC26 MACROECONOMIC FORECASTING
A Framework for Simulated and Analytical Properties of Economic Forecasts
Neil ERICSSON,
Federal Reserve Board, U.S.A.
Jaime Marquez
This paper proposes a tripartite framework of design, evaluation, and post-evaluation analysis for generating and interpreting forecasts. The value of this framework is illustrated by re-examining the properties of mean square forecast errors from dynamic models, and of forecasts from empirical models of U.S. external trade. Properties of interest include the possible nonmonotonicity and nonexistence of the mean square forecast error, and the nonlinearity bias of deterministic forecasts. Each property has been previously studied in isolation from other aspects of the forecasting process. The framework developed helps reveal how each property results from integrating all the activities that generate the respective forecasts.
Keywords: approximations, forecasts, mean square forecast error, Monte Carlo, nonlinearity bias, trade balance
EC26 MACROECONOMIC FORECASTING
A Monte Carlo Study of the Forecasting Performance of Empirical SETAR Models
M.P. CLEMENTS,
University of Warwick, U.K.
J.Smith
In this paper we investigate the multi-period forecast performance of a number of empirical self-exciting threshold autoregressive (SETAR) models that have been proposed in the literature for modelling exchange rates and GNP, amongst other variables. An indicator of when such models are likely to forecast well is suggested based on the serial dependence of regimes, as a means of distinguishing between types of nonlinearities that can be exploited for improved fit versus those that contribute to a better (relative to linear models) out-of-sample forecast performance. In our study the indicator provides a reasonable guide to those models which embody nonlinearities that may yield improved conditional mean forecasts.
EC26 MACROECONOMIC FORECASTING
The Econometrics of Macroeconomic Forecasting
David HENDRY,
Nuffield College, U.K.
When an econometric mode l coincides with the mechanism gene rating the data in an unchanging world, the theory of economic forecasting is reasonably well developed. However, theory has established less about forecasting when model and mechanism differ in a non-stationary and changing world. The paper addresses the basic concepts of (un)predictability and forecastability underlying forecasting and considers their implications. Next the role of causal information is investigated, and it is shown that noncausal variables may outperform in forecasting when the model and mechanism differ in a world subject to structural breaks. The possible role of parsimony in h-step ahead forecasting is discussed, but only a small effect is found for constant-parameter cases. Further, collinearity is shown to have little effect on forecast accuracy in constant worlds, but a larger effect if the collinearity alters, so parsimony may have a justification in non-constant processes. A reformulated taxonomy of forecast errors is presented which helps explain why differencing and intercept corrections robustify forecasts against biases due to shifts in deterministic factors.
EC26 MACROECONOMIC FORECASTING
Forecasting Short-Term Eurocurrency Interest Rates: The Spread of Information
Marcelo FERNANDEZ,
Université Libre de Bruxelles, Belgium
This paper investigates the relationship between short-term interest rates and spreads in the Euromarket. More specifically, six Eurocurrencies are analysed: the Belgian and French franc, the German mark, the Danish crown, the British pound and the Dutch guilder. A multivariate test for unit roots is performed and strongly rejects the null hypothesis of integration in the 1-month and 12-months rates of these Eurocurrencies, indicating that the spread cannot be seen as a cointegration vector. Then, it is shown using a codependence analysis that the spread can still be interpreted as a long-run relationship between the short and long-term Eurorates. A flexible non-linear error correction model is proposed for the short rate changes to take into account both the short- and long-run adjustments. The models fit the data quite well, but they are unable to significantly improve the forecasting accuracy of the random walk benchmark.
Keywords: multivariate unit root test, codependence, non-linear error correction model, conditional heteroskedasticity
EC27 SEARCH MODELS
Equilibrium Search with on the Job Search and a Continuous Distribution of Opportunity Costs of Employment
C. BONTEMPS,
CREST, France
J.M. Robin, G.J. Van den Berg
In this paper we present and estimate a version of Burdett and Mortensen's equilibrium search model, allowing for a continuous heterogeneity of workers' opportunity costs of employment and a continuous heterogeneity of firms' productivities. We also assume that job offer arrival rates are independent of employment state. We derive the theoretical implications of these assumptions, provide simulations and a semi-parametric estimation procedure that we apply to a set of professional histories drawn from the 1991-1993 French Labor Survey.
EC27 SEARCH MODELS
Job Search and Asset Accumulation Under Borrowing Constraints
Silvio RENDON,
New York University, U.S.A.
In this paper I show how borrowing constraints and job search interact. Assets influence job search outcomes by allowing wealthier people to be more selective and obtain higher wages, so that initial wealth positively affects success in the labor market. On the other hand, employment transitions influence asset accumulation: unemployed individuals maintain their consumption by running down their assets, and employed agents save to buffer against future unemployment spells and future lower wages. I fit the model to data from the National Longitudinal Survey (1979-cohort) and I quantify the two main effects mentioned above. Twenty quarters after leaving high school, an individual who started off with $5,000 of initial assets can have quarterly wages $50 higher than an individual who started off with no assets. A permanent decrease of $100 in unemployment net transfers leads to a decrease of between $37 and $96 in assets holdings.
Keywords: Job Search, Unemployment, Savings, Asset Accumulation, Borrowing Constraints
EC27 SEARCH MODELS
Does the Public Employment Service Affect the Search Intensity and Job-Findings Methods? Evidence from an Econometric Search Model
Muriel ROGER,
MAD, Université Paris I, France
Jacqueline Pradel, Denis Fougere
In this paper, we are interested in the influence of the Public Employment Service on the intensity of search and consequently on the exit rate of unemployment. For that purpose, we specify and estimate a structural job search model in which unemployed workers adapt optimally their personal (active) search intensity to the exogenous rate of jobs offered by the Public Employment Service (PES). Because the theoretical effect of an increase of this exogenous job offer arrival rate on the structural exit rate from unemployment is ambiguous, we turn to the estimation of the structural model using individual data. Results show that in most cases, the elasticity of the exit rate with respect to the arrival rate of PES offers is positive.
Keywords: Job Search, Search Intensity, Unemployment Duration, Structural Econometric Model, Individual Data
EC27 SEARCH MODELS
The Hiring Function Reconsidered: On Closing the Circle
Peter SMITH,
University of York, U.K.
Karen Mumford
Abstract not received from the author
EC28 GROWTH AND CONVERGENCE
Human Capital, Demographics and Growth across the US States 1920-1990
Joakim PERSSON,
Stockholm University, Sweden
Bo Malmberg
This paper finds robust evidence that age structure matters for subsequent growth in per capita income across the US States 1920-1990. The age groups 25-65 are positively related to subsequent per capita income growth. Another conclusion is that average years of schooling affects per capita income growth positively when age structure is controlled for. Moreover, the estimated speed of convergence increases substantially when schooling and age structure are held constant in the income growth regressions. Keywords: Demographics, Human Capital, Regions, Growth, Convergence
EC28 GROWTH AND CONVERGENCE
Patterns of Economic Development and the Formation of Clubs
Alain DESDOIGTS,
Université d'Evry Val d'Essonne and EPEE, France
This paper offers a contribution to the convergence debate by investigating the economic issues of development patterns and club formation. By focusing on the heterogeneity across nations of their fundamentals, our approach allows us to investigate a key economic issue that neither traditional cross-country growth regressions nor models of distribution dynamics can adequately address: whether clubs for which the assumption of a similar steady state is realistic, naturally emerge on economic grounds? Conditioning on variables reflecting both the resource allocation and the resource accumulation, the empirical evidence strongly supports the convergence club hypothesis. Moreover, unspecified and unanticipated cultural (Protestant versus Catholic), geographical (continents), and institutional (OECD versus non OECD) clubs form endogenously and emerge naturally as homogeneous classes on the basis of their economic structure.
EC28 GROWTH AND CONVERGENCE
Country Size and Comparative Advantage: An Empirical Study
J. TORSTENSSON,
Sweden
We formulate a simple model that captures two recent hypotheses: i) that countries with an abundant absolute endowment of skilled labour will be net exporters in R&D-intensive industries and ii) that countries with a large domestic market will be net exporters in scale-intensive industries. I he hypotheses are empirically tested by studying the trade of each and every OECD-country individually. Thereafter, we pool the data into one regression by introducing simultaneously country and industry characteristics. The results offer relatively strong support for hypothesis i) and some support for hypothesis ii).
Keywords: Country Size, Comparative Advantage, Economic Geography
EC28 GROWTH AND CONVERGENCE
The Role of Demand Constraints in Growth Empirics
Hélène POIRSON,
DELTA, France
This paper argues for and implements a model that allows to account not only for the influence of inputs per worker dynamics but also for that of excess demand dynamics on the dynamics of output per worker in a panel of countries. The model integrates a recent production function-based cointegration approach with a disequilibrium or "regime-switching" approach. The estimation results suggest that demand effects generated by excess demand dynamics are correlated positively with accumulation rates of physical capital and negatively so with accumulation rates of human capital. Failing to account for them thus leads to biases in growth regressions.
Keywords: economic growth, demand constraints, physical and human capital accumulation, international per-worker ouput, analysis of panel data
EC29 PORTFOLIO PERFORMANCE
Testing Portfolio Performance
Alistair HALL,
University of Birmingham, U.K.
Peter Knez
In this paper we develop a set of econometric techniques for testing the performance of a portfolio of financial assets. These techniques are based on the nonparametric measures proposed by Chen and Knez (1996) and cover the following cases: (1) the performance measure is based on a set of benchmark assets; (2) the performance measure is based on a set of prespecified economic factors; (3) the performance measure is based on a statistical factor model. For each case both unconditional and conditional performance measures are considered. The techniques are applied to mutual fund data from the U.S.
Keywords: stochastic discount factors, Law of One Price, non-arbitrage condition, factor models, asymptotic principal components
EC29 PORTFOLIO PERFORMANCE
Extreme Returns, Tail Estimation and Value-at-Risk
J. DANIELSSON,
LSE, U.K.
Accurate prediction of extreme events are of primary importance in many financial applications. The properties of historical simulation and RiskMetrics techniques for computing Value-at-Risk (VaR) are compared with a method which involves modelling the tails of financial returns explicitly with a tail estimator. The methods are compared using a sample of U. S. stock returns. For predictions of low probability worst outcomes, RiskMetrics type analysis underpredicts while historical simulation overpredicts. However, the estimates obtained from applying the tail estimator are more accurate in the VaR prediction.
EC29 PORTFOLIO PERFORMANCE
Estimating Inflation Risk with Indexed Bonds
Mike WICKENS,
University of York, U.K.
Frank Gong, Eli Remolona
Abstract not received from the author
EC29 PORTFOLIO PERFORMANCE
Home Bias and Conditioning Information: Methodological Issues and Empirical Results
Alessandro SBUELZ,
LBS, U.K.
The investors' portfolio is biased towards domestic assets (home bias puzzle). This paper proposes a general model for measuring the time-series pattern of such phenomenon. Supposing that the investors' actual portfolio choices are dynamically rational, a conditional CAPM stemming from the classic Euler rationality restriction provides their implicit ex-ante returns. Those implicit ex-ante returns embed their rational choice and are therefore compared with the pure conditional expectations of returns, which do not involve any apriori structure. This comparison produces an intertemporal home bias measure: lack of optimal diversification should typically be offset by improved implicit expected returns on the domestic assets. Exchange risk premium, inflation risk premium and market frictions are integrated. The model helps understanding how accrued information affects the parochial attitude; it also provides the temporal behaviour of the implicit cross-border transaction costs which justify the actual bias observed period after period. The econometric framework designed in the paper
EC30 COINTEGRATED SYSTEMS
Oil Shocks and Long Run Price and Import Demand Behavior
Jan URBAIN,
University of Limburg, The Netherlands
F. Kleibergen, H. van Dijk
The effect which the oil price time series has on the long run properties of Vector AutoRegressive (VAR) models for price levels and import demand is investigated. As the oil price variable is assumed to be weakly exogenous for the long run parameters, a cointegration testing procedure allowing for weakly exogenous variables is developed using a LU decomposition of the long run multiplier matrix. The likelihood based cointegration test statistics, Wald, Likelihood Ratio and Lagrange Multiplier, are constructed and their limiting distributions derived. Using these tests, we find that incorporating the oil price in a model for the domestic or import price level of seven industrialized countries decreases the long run memory of the inflation rate. Second, we find that the results for import demand can be classified with respect to the oil importing or exporting status of the specific country. The result for Japan is typical as its import price is not influenced by GNP in the long run, which is the case for all other countries.
Keywords: Oil price shocks, cointegration, exogeneity, import demand
EC30 COINTEGRATED SYSTEMS
Estimating the Kronecker Indices of Cointegrated Echelon Form VARMA Models
Holger BARTEL
Humboldt Universität zu Berlin, Germany
Cointegrated VARMA models can be parameterized by using the echelon form, which is characterized by the Kronecker indices. Three different methods for estimating the Kronecker indices of cointegrated echelon form VARMA models are discussed and compared. They have the common feature of estimating the individual equations of the system separately and using order selection criteria. The small sample performance of the methods is compared in a simulation study. It is found that the performance is better if all echelon form restrictions implied by the Kronecker indices found in preceeding steps are incorporated immediately. Keywords: VARMA models, cointegration, echelon form, Kronecker indices
EC30 COINTEGRATED SYSTEMS
A Review of Systems Cointegration Tests
Kristin HUBRICH,
Humboldt Universität zu Berlin, Germany
Helmut Lutkepohl, Pentti Saikkonen
The relevant literature on systems cointegration tests is reviewed, and the various sets of assumptions for the asymptotic validity of the tests are compared within a general unifying framework. The comparison includes likelihood ratio tests, a modification of these tests for processes without deterministic trends, lag augmentation tests, LM type tests, tests based on canonical correlations, the Stock-Watson tests and Bierens' nonparametric tests. Asymptotic results regarding the power of these tests and previous small sample simulation studies are discussed. We present our own simulations to compare all tests under uniform conditions. A special emphasis is given to the sensitivity of the test performance towards the trending properties of the DGP.
Keywords: systems cointegration tests, LR tests, nonparametric tests, asymptotic power, small sample simulations
EC30 COINTEGRATED SYSTEMS
The Influence of VAR Dimensions on Estimator Biases
Kaddour HADRI,
Brunel University, U.K.
Karim Abadir, Elias Tzavalis
In this paper, we consider the implications of varying the dimensions of VARs on the biases of Maximum Likelihood and Least Squares Estimators (MLE and LSE, respectively). In the purely nonstationary case, estimator biases are exactly equal to the dimension of the system times the univariate bias, even when the variates are generated independently of each other. Even though we do not derive an exact formula for the variance, we show (qualitatively) that it too increases with the dimension of the system, sometimes dramatically, hence also raising the Mean Squared Error (MSE) of the estimator. When some stable linear combinations exist, the biases are generally smaller and are asymptotically proportional to the sum of the characteristic roots of the VAR. One source of such combinations is meaningful economic relations that are represented by the cointegration of some of the components of the VAR. Adding economically-irrelevant variables to a VAR is thus shown to have more serious negative consequences in integrated time series than in classical ergodic or cross section analyses. The findings strengthen the case for parsimonious modelling and for the reduction step of the general-to-specific marginalization method. They also support the use of seasonally unadjusted data whenever possible.
Keywords: Vector AutoRegression (VAR), Integrated and Cointegrated series, Spurious correlation, Error Correction Mechanism (ECM), Jordan decomposition
EC31 PROGRAMME EVALUATION
A Decision-Theoretic Approach to Programme Evaluation
R.H. DEHEJIA,
U.S.A.
This paper develops a decision-theoretic approach to program evaluation, and applies it to data from California's GAIN experiment (a randomized trial of a welfare-to-work alternative to AFDC). I use a mixture-of-normals specification to model individual earnings in a flexible way, and then make use of individual-specific predictive distributions of earnings to examine a range of decision problems relevant in the context of program evaluation. I show that thinking of program evaluation as a decision problem leads both to much sharper and richer rankings of the treatment and control policies than is available through the alternative methodology.
Keywords: Program Evaluation, Bayesian Decision Theory
EC31 PROGRAMME EVALUATION
Labour Supply with In-Work and In-Kind Transfers
Paul BINGLEY,
Arhus University & Arhus Business School, Denmark
Ian Walker
There have been a number of changes to the nature of in-work transfers over the last 20 years. Further expansions have been proposed and in-work transfer schemes have recently been promoted as a device for encouraging labour force participation and reducing the severity of the disincentives associated with out-of-work income support schemes. In-kind transfers have also been promoted as a way of proving poverty relief without inducing severe adverse incentive effects. Here we estimate a discrete choice labour supply model which allows for endogenous participation in multiple cash and in-kind transfer programmes, and for involuntary unemployment. This is applied to the Family Credit, Housing Benefit, Free School Lunch and Milk Token programmes for a sample of 4527 lone mothers drawn from UK Family Expenditure Surveys. We find that Family Credit and In-Kind transfers have larger labour supply incentive effects as they expand up the hours distribution; there are economies to multiple programme participation; and Family Credit and in-kind programme participation are non-separable from labour supply. Also there is quite strong evidence of a utility loss associated with programme participation per se, which is consistent with there being some stigma driving low participation rates.
Keywords: labour supply, programme participation, in-kind transfers
EC31 PROGRAMME EVALUATION
Training of the Unemployed. Evaluation by Means of a Natural Experiment
Isabelle BARDOULAT,
IRES, Université Catholique de Louvain, Belgium
Bart Cockx
In this paper we estimate the impact of vocational training for un-employed workers on the propensity to leave unemployment in the Wal-loon region in Belgium for the May 1989 - March 1993 period. As we base our analysis on non-experimental data, we face the well-known selectivity problem. We argue that the financing rule of the vocational training induces exogenous variation in the participation rate across sub-regions. This allows to identify the true effect of training without imposing restrictive distributional assumptions. The method is largely based on Cockx and Ridder (1996). We extend this work by allowing for the presence of other training programs and by distinguishing the participation effect from the post-program effect. Training is found to increase unemployment duration during program participation, but this is more than compensated by the post-program decrease. The selection bias leads to a slight underestimation of the effects.
Keywords: training, evaluation, duration data, grouping
EC31 PROGRAMME EVALUATION
A Regression-Discontinuity Evaluation of The Effect of Financial Aid Offers on College Enrolment
Wilbert VAN DER KLAAUW
New York University, U.S.A.
This paper discusses and applies a powerful, but largely overlooked quasi-experimental design, the `Regression-Discontinuity' design. It is shown how the design can be exploited to obtain credible program or treatment effect estimates without having to rely, as in usual selection bias correction methods, on arbitrary exclusion restrictions, and/or functional form and distributional assumptions on errors. This is illustrated in the context of an evaluation of the effect of financial aid offers on college enrollment decisions.
Keywords: Regression-Discontinuity design, selection bias, instrumental variables
EC32 INDIRECT ESTIMATION
Indirect Estimation of Just-Identified Models with Control Variates
Girgio CALZOLARI,
Universita di Firenze, Italy
Francesca Di Lorio, Gabriele Fiorentini
Simulation estimators, such as indirect inference or simulated maximum likelihood, are successfully employed for estimating models where the likelihood function does not have a simple analytical expression. They adjust for the bias (inconsistency) produced by the estimation of an auxiliary model that can be manageable, but is essentially misspecified. The price to be paid is an increased variance of the estimated parameters. There is, in fact, an additional component of the variance, which depends on the stochastic simulation involved in the estimation procedure. To reduce this undesirable effect, one should properly increase the number of simulations (or the length of each simulation) and thus the computation cost. Alternatively, this paper shows how variance reduction can be achieved, at virtually no additional computation cost, by use of control variates. This technique can be applied in a particular just-identified context, that is when a one-to-one correspondence exists beteween parameters of the econometric model (the model of interest) and of the auxiliary model. This is a case which often occurs in practical applications. Several models are explicitly considered and experimented with: moving average, ARMA, stochastic differential equations, stochastic volatility models. Monte Carlo experiments show that, for some parameters of interest, a global efficiency gain up to almost 50\%. over the simplest indirect estimator is obtained at about the same computation cost.
Keywords: Efficient Monte Carlo, Variance reduction techniques, Control variates, Indirect inference, ARMA models, Stochastic volatility, Stochastic differential equations
EC32 INDIRECT ESTIMATION
A Toolbook for Using Indirect Estimation Methods
E. BURGAYRAN,
CREST, France
Serge Darolles
In this paper, we study the different blocks involved in the usual estimation methods using simulated data. We first present in details the choices available in the building of the method. Then, we perform Monte-Carlo studies to measure the time required by each method, and their small sample properties. We conclude that decreasing the size of auxiliary optimization problem leads to an important gain of computation time, without loss of efficiency. The simulation method is important, and does not give equivalent final estimators, and Monte-Carlo studies allows us to propose a solution to these difficulties. An application to the Constant Elasticity of Variance diffusion model with a linear mean-reverting drift gives us a generic example to illustrate all the solutions proposed at each step.
Keywords: Simulated Methods, Monte-Carlo, Diffusion
EC32 INDIRECT ESTIMATION
Calibration of Structural Models by Indirect Inference
Ramdan DRIDI,
Université de Toulouse, France
Laurence Broze, Eric Renault
We propose in this paper a generalization of the indirect inference (I.I.) methodology, termed G.I.I., in order to account for calibrationnist approach where a structural model is studied through its ability to reproduce some well chosen moments. We stress that little attention has been payed in the I.I. literature on the consequences of misspecification of the structural model while they are even more detrimental than for direct inference, since the misspecified model is used for building simulated path. The only way to protect one against likely misspecification of the structural model is to examine it through a convenient instrumental model which does not capture what goes wrong in the simulated paths. We argue that a well-driven G.I.I. strategy might be seen as rigorous calibrationnist approach, that is that captures both the advantages of this approach (account for structural "a - statistical" ideas) and of the inferential approach (precise appraisal of loss functions and of conditions of validity). We provide a collection of examples where such a strategy allows one to maintain consistency of the estimators of the structural parameters of interest, in spite of misspecification. These examples typically involve unobserved stochastic processes that enter the structural model non-linearly (stochastic volatility, equilibrium asset-pricing).
Keywords: Indirect Inference, GMM, Structural Models, Instruments, Calibration
EC32 INDIRECT ESTIMATION
Indirect Estimation of ARFIMA and VARFIMA Models
Vance MARTIN,
The University of Melbourne, Australia
Nigel Wilkins
Indirect estimation methods are proposed for handling both univariate as well as more complex multivariate fractionally integrated ARMA models. Special attention is given to comparing the finite sampling properties of the indirect estimator with Sowell's (1992a) exact time domain maximum likelihood estimator and the Geweke and Porter-Hudak (1983) estimator. The indirect estimator is shown to be computationally much faster than the maximum likelihood estimator while generating similar small sample properties. The relative computational gains of the indirect estimator over maximum likelihood increase as the complexity of the data generating process increases.
EC33 LONG MEMORY PROCESSES
Long-Term Dependence in Trading
Joanna JASIAK,
York University, Canada
E. Ghysels
In this paper we examine long-term dependence in times betweentrades and quotes on financial markets. The autocorrelation functions ofseveral intra-trade duration series show a slow, hyperbolic rate of decay typical for long memory processes. For example, a shock to a wating time to trade the IBM stock on the New York Stock Exchange (NYSE) may persist in the transactions time for a long period of 1000 or 2000 ticks. With an average duration of 35 seconds between transactions this may amount to ten or twenty hours in calendar time. Given the complexity of trading dynamics it is very important to capture the long-range persistence to predict trading activity.In this paper we characterize and estimate long memory in data from floor- and screen-trading stock markets as well as FX Reuter screen exchange rate quotations. We propose a fractionally integratedautoregressive conditional duration (FIACD) model as well as a periodic model to accomodate intra-day variation in the degree of persistance.To relate the duration persistance to the stock price dynamics we study durations between consecutive returns to states of null, positive or negative returns. These series are characterized by distinct autocorrelation patterns and provide information about the expected time to increases (decreases) in stock prices. The persistance in durations is also directly related to times spent in a given state of returns. The average visiting time is state dependent, features intra-day variation and may be considered as an alternative measure of persistence.
Keywords: durations, long memory, periodicity, persistence measures
EC33 LONG MEMORY PROCESSES
Double Long-Long Memory Financial Time Series
Gilles TEYSSIERE,
University of London, U.K.
We show that some financial time series display long-memory in both their conditional mean and their conditional variance. We model these double long-memory series by combining a fractionally integrated regression function and a fractionally integrated skedastic function. We examine the case of a double long-memory (DLM) model, the "ARFIMA-FIGARCH", and we propose an estimation procedure by QML which requires pre-sample values. We show, by using Monte Carlo simulations and kernel density estim.ation, that; the QML Estimates for the parameters of this model have nice properties: they are root-n consistent, asymptotically normal except for one parameter, and the level of bias is negligible. We analyze the consequences of d disregarding t; he long-memory component in the regression specification, even for rnoderate values of the degree of long-memory which are not detected by appropriate statistical tests. We use two DLM models, the ARFIMA-FIGARCH and the ARFIMA-FIEGARCH, for modelling absolute returns on T-Bill futures.
Keywords: integrated processes, double long-memory, heteroskedasticity, second-generation models, Monte Carlo methods
EC33 LONG MEMORY PROCESSES
Estimation of Distributioned Lag Models with Long-range Dependence
Javier HIDALGO,
LSE, U.K.
This paper provides central limit theorems for the estimation of the parameters and consistent estimation of the true order of a distributed lag polynomial in the presence of long-range dependence in both regressors and errors. We show that the estimator proposed by Hannan (1963,1967) with slight modification is root-T consistent and asymptotically normal. The method of estimation becomes, when the data is long-range dependent more attractive than Least Squares, since the latter can be neither root-T consistent nor asymptotically normal as is the case with weakly dependent data. Also, due to the fact that the errors are autocorrelated, to provide consistent estimation of the true order of the distributed lag, we need to modify the determination order criteria already given in earlier works. Moreover, based on the estimates of the parameters of the model, a nonparametric test is derived for the null hypothesis of non-causality between economic variables, which does not rely on a complete knowledge of the underlying structure of the data.
EC33 LONG MEMORY PROCESSES
Analyzing Continuous-Time Long-Memory Models with Wavelets
E. HOEG,
Aarhus School of Business, Denmark
It is natural to analyze long-memory models with wavelet methods as these models possess the self-similarity property and wavelet methods enable analysis according to scale. Fractional Brownian Motion is used to model continuous-time long memory processes or processes with the self-similarity property. By means of the theory of wavelets we obtain the covariance matrix for the wavelet transformed stochastic process. This in turn facilitates the likelihood analysis as it appears that the covariance matrix in the transformed model is sparse. We use a discrete wavelet transform algorithm in conjunction with a numerical optimization routine. The relationship to (the discrete) ARFIMA class of models is analyzed. The method is applied to a US interest rate series, where the conjecture is that the series is non-stationary but mean reverting.
Keywords: wavelets, mean reversion, interest rates, Fractional Brownian Motion
EC34 COST FUNCTIONS/PRODUCTION FRONTIERS
The Expected Minimum Cost Function: a Non Parametric Approach
Catherine CAZALS,
Institut D'Economie Industrielle, France
Jean-Pierre Florens
The purpose of this paper is to define the concept of expected minimum cost function and to present the methodology for a non parametric estimation of it. Generally, in cost econometric analysis, inference is made on the cost function, namely the conditional expectation of cost given some level of output. Here, we consider the conditional distribution on the cost, given that the output level is greater than a fixed level. Our methodology is applied to a data set concerning cost and outputs of the delivery activity for french post offices.
Keywords: expected minimum cost function, cost function, conditional survivor function, frontier, non parametric estimation
EC34 COST FUNCTIONS/PRODUCTION FRONTIERS
Specification and Estimation of Multiple - Output Stochastic Ray Frontier Production Models
Michael LÖTHGREN,
Stockholm School of Economics, Sweden
This paper proposes an approach to specify and estimate multiple input, multiple output production frontiers and technical efficiency using a stochastic ray frontier production model. A possible model extension is to incorporate a technical efficiency effects model to allow estimation of the effects of various explanatory variables on technical efficiency. An empirical application using Swedish health care data reveals a significant positive effect on technical efficiency of an "internal market" reform while the effect on the production frontier is negative. Technical change is found to be positive while technical efficiency has decreased over time.
Keywords: Composed error model, Distance function, Panel data, Stochastic ray frontier model, Technical change, Technical efficiency
EC34 COST FUNCTIONS/PRODUCTION FRONTIERS
Stochastic Estimation of Firm Inefficiency Using Distance Functions
Scott ATKINSON,
University of Georgia, U.S.A.
Daniel Primont, Rolf Fare
Abstract not received from the author
EC35 COINTEGRATION (B)
Testing for the Cointegrating Rank of a VAR Process with a Time Trend
Helmut LÜTKEPOHL,
Institut fuer Statistik und Oekonometrie, Germany
Pentti Saikkonen
Standard tests for the cointegrating rank of a vector autoregressive (VAR) process have nonstandard limiting distributions which depend on the characteristics of intercept terms and time trends in the system. In practice these characteristics are often unknown. Therefore modified tests are proposed with limiting distributions which do not depend on the characteristics of deterministic terms under the null hypothesis. One type of tests makes use of lag augmentation, that is, a VAR process of order p+1 is fitted when the true order is p while the tests are based on the coefficient matrices of the first p lags only. It is shown that Chi^2 limiting distributions are obtained in this way. The price for this simplicity will be reduced power, however. Therefore, we also consider LM (Lagrange multiplier) type tests. These tests are shown to have nonstandard limiting distributions which do not depend on deterministic terms.
Keywords: Cointegration, Cointegrating rank, Wald test, Lagrange multiplier test, vector autoregressive process
EC35 COINTEGRATION (B)
Testing for the Rank of the Co-Integration Space and the Form of the Intercept Using the BIC Criterion
Antonio AZNAR,
University of Zaragoza, Spain
M. Salvador
Abstract not received from the author
EC35 COINTEGRATION (B)
On The Distribution of Tests For Cointegration Rank
B. NIELSEN,
Nuffield College, U.K.
The likelihood test for the hypothesis of reduced cointegration rank in a Gaussian vector autoregressive model is analysed. In finite samples the rejection probability for the hypothesis may be quite different from the promised asymptotic size. The likelihood analysis is a combination of canonical correlation analysis and autoregressive analysis. In both cases tests are often non-similar. This is also the case for cointegration tests which explains the small sample problems. Some key examples are analysed. Some alternative graphical diagnostics are discussed.
Keywords: Cointegration, Vector autoregressive model, Likelihood Ratio Test, Similarity, Small sample distribution
EC36 ASSET PRICING
Addressing Nonlinear Dynamics in Asset Prices with a Behavioural Capital Asset Pricing Model: with an Empirical Application to Noise Trade Behaviour in Financial Markets
Craig HEIMSTRA,
University of Strathclyde, U.K.
The principal motivation of this paper is to develop a behavioural capital asset pricing model by which certain issues relating to the existence and nature of noise traders in financial markets can be examined empirically through a parsimoniously parameterized structural econometric model. A behavioural asset-pricing model describing the actions of agents with heterogeneous expectations is derived. This model admits a mean belief across the agents relating to the expected return on the market portfolio. This mean belief is modeled using ideas taken from statistical mechanics and discrete-choice theory. The behavioural parameters are then linked econometrically to trading-rule information. This link allows the general aims and objectives of the paper to be addressed empirically.
Keywords: Interacting Particles Systems Theory, Mean Field Theory, Noise Traders, Capital Asset Pricing
EC36 ASSET PRICING
Human Capital Risk and Asset Pricing
Ignacio PALACIOS-HUERTA,
Stanford University, U.S.A.
This study derives labor-income based measures of human capital returns directly from the Euler equations, and uses them (i) to evaluate the pricing of human capital assets and compare their pricing implications across individuals with different demographic characteristics and (ii) to compare the asset pricing implications of human capital returns and financial returns. The period of study is 1964-1996 and the methodology consists of comparing the Hansen-Jagannathan bounds on the stochastic discount factor (intertemporal MRS) implied by human capital and financial returns, with and without using conditioning information. Finally, it calculates the increase in expected returns that the different individuals can achieve by diversification into other human capital assets or into financial assets during the period of study.
Keywords: Human capital risk and returns, pricing of human capital and financial assets, asset pricing and portfolio
EC36 ASSET PRICING
Volume Volatility Dynamics in an Intertemporal Asset Pricing Model
Chiente HSU,
Duke University, U.S.A.
Abstract not received from the author
EC37 WAGE INEQUALITY
Wages, Prices, and International Trade: Trends across Industries for an "Export Champion"
Bernd FITZENBERGER,
Universität Konstanz, Germany
This paper provides an empirical analysis of the structure of earnings in West Germany across skill groups and industries. Our analysis is based on data from the German Socioeconomic Panel for the period 1984 to 1994. We estimate quantile regressions in order to obtain a finer picture of the earnings structure compared to conventional least squares methods. For robust standard error estimation, this study uses a block bootstrap procedure. We also suggest a simple one-step procedure to obtain a consistent estimate of inter-industry earnings variability. Our main findings are: first, pooled estimation comprising a uniform time trend is not rejected by the data, and second, the effects of human capital variables and industry dummies on earnings differ considerably across quantiles.
Keywords: Earnings Structure, Quantile Regression, Block Bootstrap
EC37 WAGE INEQUALITY
Changing Distribution of Wages in the UK and West Germany 1984-1992
Amanda GOSLING,
IFS, U.K.
The aim of this paper is to compare the differing response of the UK and West German labour markets to the global demand changes of the 1980s. Using comparable datasets we show that the oft-cited increases in the inequality of wages between and within skill groups in the UK have not occurred in West Germany. Moreover, over the period considered, there has been no change in the relative employment rates of unskilled workers in contrast to the UK. It is argued that this is because changes in the relative demand for skills are better absorbed into changes in relative supply in West Germany.
EC37 WAGE INEQUALITY
International Comparisons of Wage Inequality: How Robust are the Findings?
Hans VAN OPHEM
FEE, The Netherlands
Edwin Leuven, Hessel Oosterbeck
According to Blau and Kahn (1996) international differences in male wage inequality cannot be explained by a simple model of supply and demand for skill. We argue that this conclusion may be due to employing an inappropriate measure of skill. Their measure operates on the strong assumption that years of schooling and years of experience are comparable across countries. A different skill measure is obtained from an international comparative literacy test. Using this alternative skill measure, we find that international differences in male wage inequality by skill between the U. S. on the one hand and Germany, the Netherlands and Sweden on the other hand, are consistent with international differences in demand and supply of skill.
Keywords: wage inequality, international comparison
EC38 CONDITIONAL HETEROSKEDASTICITY
Second Order Pseudo-Maximum Likelihood Estimation and Conditional Variance Misspecification
Bernard LEJEUNE,
Université de Liège, Belgium
In this paper, we study the behavior of second order pseudo-maximum likelihood estimators under conditional variance misspecification. We first show that, even if mean and variance parameters vary independently, standard PML2 estimators are not robust to conditional variance misspecification. Further, we outline sufficient and essentially necessary conditions for a second order pseudo-maximum likelihood estimator to be consistent for both mean and variance parameters when both the conditional mean and the conditional variance are correctly specified, and to remain consistent for the mean parameters when the conditional mean is correctly specified but the conditional variance is misspecified. Finally, we provide limiting distribution results for this restricted class of second order pseudo-maximum likelihood estimators under different assumptions regarding the degree of misspecification present in the model and compare its relative merits with those of standard QGPML1.
Keywords: Pseudo-maximum likelihood methods, misspecified models
EC38 CONDITIONAL HETEROSKEDASTICITY
Estimating Weak GARCH Representations
J.M. ZAKOIAN,
CREST, France
C. Francq
The classical definitions of GARCH-type models rely on strong assumptions on the first two conditional moments of the process. Using a different approach based on linear conditional expectations, we define a new class of weak ARMA-GARCH representations. The generality of the formulation is illustrated through a collection of examples. We propose a two-stage estimation procedure based on the minimization of sums of squared deviations about linear innovations. We provide proofs of strong consistency and asymptotic normality of the estimator. Finally, a small simulation experiment illustrates the method.
Keywords: Weak GARCH, Conditional Heteroskedasticity, Conditional Least Squares, Mixing, Asymptotic normality
EC38 CONDITIONAL HETEROSKEDASTICITY
On Estimation and Testing Goodness-of-Fit for m-Dependent Stable Sequences Conditionally Heteroskedastic Stable Processes
Rohit DEO,
Stern School of Business, New York University, U.S.A.
A class of estimators of the characteristic index of m-dependent stable sequences is proposed. The estimators are sown to be consistent and asymptotically normal. In addition, a class of goodness-of-fit tests for stability is also obtained. The performance of the estimators and the goodness-of-fit tests is evaluated through a simulation study. One of the estimators is shown to have a reasonably high relative efficiency which is uniformly superior to that of the regression estimator, which is currently most widely used. Our results have significance for modelling financial data like stock returns and exchange rates which have thick tailed distributions and exhibit non-linear behaviour.
Keywords: thick tailed distributions, characteristic index, U-statistics, conditional heteroscedasticity
EC39 VAR ANALYSIS AND MACROMODELLING
Fiscal Policy in the Transition to Monetary Union: a Structural VAR Model
Olivier DE BANDT
European Monetary Institute, Germany
Catherine Bruno
In order to assess the effect of fiscal rules in EMU, we use a structural V.A.R. model for Germany and France, including the general government financial balance. This allows to distinguish between fiscal and monetary shocks. Fiscal shocks do not contribute to a significant part of the dynamics of output in both countries, especially in France, so that relinquishing fiscal autonomy should not matter too much. In addition, fiscal shocks in the two countries are rather uncorrelated, although it is difficult to conclude that it reflects the influence of the political cycle, rather than a different policy-mix.
Keywords: Budget deficit, Structural VAR, EMU
EC39 VAR ANALYSIS AND MACROMODELLING
Rational Expectations in a VAR with Markov Switching
Marten BLIX,
Institute for International Economics Studies, Sweden
This paper shows how a well known class of rational expectations hypotheses using linear vector autoregressions (VAR:s) can be extended to allow for unobservable Markov switching. The regime shift model here differs to the centered Hamilton model in that it has the intuitive appeal of symmetric state dependence, as well as being easier to estimate. The contribution of the paper is to derive testable restrictions implied by rational expectations, which are linear when the forecast horizon is infinite. As an illustration, we test the expectations hypothesis on 3 and 6 month US bonds for quarterly data.
Keywords: VAR, Markov chain, regime switching, rational expectations, expectations hypothesis
EC39 VAR ANALYSIS AND MACROMODELLING
Modelling the Demand for Imports and Domestic Output
Bjorn NAUG,
Statistics Norway, Norway
The paper models the ratio of domestic output to imports in Norwegian expenditure on manufactures. Using data spanning 1967(1)-1994(4), multivariate cointegration analysis establishes a long-run relationship which implies a trend decline in the output-imports ratio of 3% per annum and a unit long-run substitution elasticity. Increasing international specialisation is presumed to be the most important aspect captured by the trend. Deviations from the cointegrating relation are highly significant in a single equation model of the output-imports ratio, which also contains marked negative effects of capacity utilisation in Norwegian manufacturing.
Keywords: Imports, cointegration, error correction model
EC40 EMPIRICAL MACROECONOMICS
Real and Monetary Determinants of the Real Exchange Rate in South Africa
Janine ARON,
University of Oxford, U.K.
Ibraham Esbadawi, Brian Kahn
The real exchange rate is a key policy variable in South Africa's open economy. A cointegration framework is used with single equation equilibrium correction models to investigate the short-run and long-run equilibrium determinants of the quarterly real exchange rate, 1970:1-1995:1.The cointegrated equilibrium is obtained from a theoretical model characterising equilibrium as the achievement of internal and external balance for sustainable capital flows and trade and tax regimes, and given terms of trade and technology. For a spectrum of exchange rate regimes, we discuss the evolution of a broad range of theoretical economic fundamentals and short-run policy factors.
Keywords: Real Exchange Rate, South Africa
EC40 EMPIRICAL MACROECONOMICS
Long Run Money Demand in the EU: Evidence for Area-Wide Aggregates
J. HENRY,
European Monetary Institute, Germany
G. Fagan
Abstract not received from the author
EC40 EMPIRICAL MACROECONOMICS
Measuring Inflation: An Attempt to Operationalize Carl Menger's Concept of the Inner Value of Money
C.K. FOLKERTSMA,
De Nederlandsche Bank NV, The Netherlands
M. Fase
This paper attempts to operationalize Carl Menger's concept of the 'innerer Tauschwert des Geldes', i.e. the inner value of money. Since the change in the inner value of money is the component of price movements which is due to monetary influences, the operationalization provides an alternative measure of inflation. We consider several approaches for gauging the change in the inner value of money. Of these, we use Quah and Vahey's structural VAR model to identify the price movements in the Netherlands and the EU due to monetary shocks.
Keywords: core inflation, monetary policy, European Monetary Union
EC41 EFFECTS OF TRAINING AND SENIORITY
Nonparametric Bounds on Employment and Income Effects of Continuous Vocational Training in East Germany
Michael LECHNER,
Universität Mannheim, Germany
This paper explores the potential of an approach suggested by Manski of obtaining nonparametric bounds for treatment effects in evaluation studies without knowledge of the participation process. The practical concern is the effects of continuous vocational training in East Germany. The empirical application is based on a large cross-section that covers about 0.6% of the total population in 1993. The results are rather mixed. The large width of the intervals obtained emphasise the fundamental problem of all evaluation studies without good knowledge of the relationship between potential outcomes and the participation process. However, in some cases suitable exclusion restrictions are indeed capable of bounding the treatment effects strictly away from zero.
Keywords: Nonparametric estimation of treatment effects, training evaluation, East German labour markets, Mikrozensus
EC41 EFFECTS OF TRAINING AND SENIORITY
Firm Heterogeneity and Worker Self-Selection Bias Estimated Returns to Seniority
David MARGOLIS,
Université de Paris 1, France
I develop a model under which workers with different marginal productivities self-select into firms based on the firm's seniority reward policy. I show how this may bias upwards the estimates of returns to seniority in cross-sectional and even some longitudinal studies, when differences in workforce composition are ignored. I develop a new estimator of "true" returns to seniority and empirically test the implications of the model. I show how several previous estimation strategies over-estimate returns to seniority, particularly in firms that offer zero or negative returns to job seniority, using a large sample of French firms and workers.
Keywords: matched panels, returns to seniority, self-selection, worker heterogeneity, firm heterogeneity, composition bias
EC41 EFFECTS OF TRAINING AND SENIORITY
Training and Exclusion: Is the German Apprenticeship System in Decline?
Klaus ZIMMERMAN,
University of Munich, Germany
Christoph Schmidt
German vocational training has become a reference model for the reform debate in many countries, especially in the US. However, this contrasts with the internal German debate that is concerned with a decline in the system of vocational training in favor of university training. The paper provides econometric evidence to disentangle the arguments and the empirical facts. It is true that manual training is in decline, however, the training system remains effective to integrate excluded groups such as unemployed and migrants in the labor market. Hence, the perceived decline in the German training system is actually a restructuring of training. Keywords: Vocational Training, Apprenticeship System
EC42 PANEL DATA MODELS
GMM Bootstrapping and Testing in Dynamic Panels
Matz DAHLBERG,
Uppsala University, Sweden
Pal Bergström, Eva Johansson
Two different bootstrap approaches for GMM estimation have recently been suggested for use in dynamic panel data models, Brown & Newey (1992) and Hall & Horowitz (1996). In this paper we compare the small sample properties of these estimators as well as their performance in a sequence of tests. This comparison is carried out by means of Monte Carlo experiments. Our findings are that i) the Brown and Newey method has superior size properties, but cannot be used in a sequence of tests without modifications. ii) the Hall and Horowitz method works better than the usual asymptotic tests in a sequence of tests.
Keywords: GMM, Panel Data, Bootstrapping, Test sequences
EC42 PANEL DATA MODELS
Robust GMM Estimation of an Euler Equation Investment Model with German Firm Level Panel Data
Norbert JANZ,
Centre for European Economic Research (ZEW), Germany
In this paper the outlier robust GMM panel data estimator recently proposed by Lucas, van Dijk, and Kloek is applied to an Euler equation model of firm investment behaviour with imperfectly competitive product markets for a small panel of German nonfinancial stock companies. Truncated QQ-Plots for checking distributional implications and Denby-Mallows-Plots for the selection of tuning constants are provided. Whereas the estimation results from the usual GMM estimator would contradict the theory, the empirical results using the Robust GMM estimator largely support it.
EC42 PANEL DATA MODELS
Simultaneous Equations with Incomplete Panels
Badi BALTAGI,
Texas A&M University, U.S.A.
Young-Jae Chang
This paper compares the performance of several single and system estimators of a two equation simultaneous model with unbalanced panel data. The Monte Carlo design varies the degree of unbalancedness in the data and the variance components ratio due to the individual effects. One useful result for applied researchers is that the feasible error components 2SLS and 3SLS procedures based on simple ANOVA type estimators of the variance components perform well with incomplete panels and are recommended in practice.
Keywords: Simultaneous Equations, Incomplete Panels, Unbalanced ANOVA
EC43 SAMPLE SELECTION MODELS/GMM
Estimation of Dynamic Panel Data Sample Selection Models
Ekaterini KYRIAZIDOU,
University of Chicago, U.S.A.
This paper considers the problem of estimation in dynamic panel data sample selection models with a binary sample selection rule (Type 2 Tobit models) when the latent equations contain strictly exogenous regressors, lags of the dependent variables and additive unobserved individual-specific effects that are possibly correlated with the observed covariates. We propose two-step kernel-weighted GMM-type estimators, where the kernel weights depend of initial first-step estimates of the selection equation. The estimation exploits moment conditions similar to the ones used in the estimation of linear dynamic panel data models. The proposed estimators are shown to be consistent and asymptotically normal atlhough their rate of convergence is slower than the square root of the sample size.
Keywords: Panel Data, Sample Selection
EC43 SAMPLE SELECTION MODELS/GMM
A New Estimator For Panel Data Sample Selection Models
Maria ROCINA
UCL, U.K.
We are concern with the problem of estimation in a panel data sample selection model, where both the selection rule and the regression equation contain individual fixed effects. In particular, the estimation procedure is an extension of Heckman's sample selection technique to the case where one correlated selection rule in two different time periods generates the sample. Some non-parametric components are present in our approach. Monte Carlo simulation experiments and extensions towards more robust estimators against missespecification of the error distribution in the main equation are presented.
Keywords: panel data, sample selection, fixed effects, non-parametric components
EC43 SAMPLE SELECTION MODELS/GMM
Moment Estimation with Attrition
Francis KRAMARZ,
CREST, France
John Abowd, Bruno Crepon
To avoid the difficulties encountered when modelling firm exit and entry, we formulate simple assumptions on the stochastic process by which firms or individuals exit from samples (Rubin (1976) and Robins, Rotnitzky, and Zhao (1994), (1995)). Then, based on these hypotheses, we show that a simple adaptation of the GMM framework allows consistent estimation of any equation using moment conditions. In particular, we propose a class of estimators based on Weighted Generalized Method of Moments, in which weights are computed using the exit process, that lead to consistent estimators of the parameters of interest. We apply our method on French firm-level data and show that the method makes a real difference when compared to traditional strategies that address these attrition problems.
Keywords: GMM, exit, attrition
EC44 BUSINESS CYCLE MODELS
Business Cycle Nonlinearities in UK Consumption and Production
Denise OSBORNE,
University of Manchester, U.K.
Nadir Ocal
This paper finds linear and two-regime models to be inadequate for capturing the business cycle characteristics of UK real consumers' expenditure and industrial production, but two-threshold exponential smooth transition autoregressive (ESTAR) models are satisfactory. These are interpreted as three-regime models, with the regimes defined as recession, normal growth and high growth. The transitions describing recovery from recession for consumption and production are found to be very similar. Of the regimes, only the normal growth one is stationary in the sense that the dynamics of the models imply that each series will move out of the other regimes back to normal growth. Dynamic simulations are used to illustrate the complex nature of the dynamic responses of these models, where the responses depend on the specific period in which the shock occurs, as well as on its magnitude and sign.
Keywords: nonlinearity, business cycles, smooth transition autoregressive models, consumption, production
EC44 BUSINESS CYCLE MODELS
International Capital Movements and the Locomotive effect
Catherine BRUNO,
OFCE, France
Abstract not received from the author
EC44 BUSINESS CYCLE MODELS
Trend and Cycle Dichotomy and the Welfare Gains of Cycle Smoothing
Joao ISSLER,
Graduate School of Economics EPGE, Brazil
Afonso de Mello Franco
If consumption is log-Normal with a dichotomous trend and cycle, a surprising result in business-cycle research is that the welfare gains of eliminating uncertainty are relatively small - $8.50 per-capita yearly. A possible problem with such calculations is that controlling volatility may change the rate of growth of consumption, i.e., the Lucas Critique. This is indeed our findings using a panel of 85 countries covering 1962 to 1989. Incorporating the effect of volatility on the drift of consumption results on a dramatic increase of the welfare gains of cycle smoothing, which can reach up to $1,792.14 for reasonable parameter values.
Keywords: trend-cycle decomposition, welfare analysis, unit-roots, panel-data estimation
EC45 INVESTMENT AND R&D
Investment, R&D and Financial Constraints in Britain and Germany
Stephen BOND,
IFS, U.K.
Dietmar Harhoff, John Van Reenen
This paper analyses empirical models of investment in R&D and physical capital for the UK and Germany. Using firm level panels of R&D performing firms we estimate identically specified models for both countries. In our preferered error-correction models there is no evidence that British companies are more financially constrained than their German counterparts. Indeed cash flow sensitivity appears to be a feature only of non-R&D performing firms. We demonstrate this by examining a larger sample of 762 UK firms (R&D and non-R&D firms). Firms likely to encounter serious financial constraints do not seem to engage in R&D.
Keywords: investment, R&D, financial constraints, panel data
EC45 INVESTMENT AND R&D
Research Investment, Innovation and Productivity: An Econometric Analysis at the Firm Level
Emmanuel DUGUET,
University of Paris 1, France
B. Crepon, J. Mairesse
This paper presents an econometric study of productivity and innovation at the firm level. We introduce three new features: (i) A four equations structural model that explains research participation and research investment by market structure and sectoral opportunities, innovation output by endogenous research investments and finally total factor productivity by endogenous innovation output; (ii) We use new French data: patent counts and innovation sales and (iii) We account for selection biases on research and development, for the count data features of patents and for the simultaneity biases. We find that innovation explains strong productivity differences among manufacturing firms.
Keywords: Systems of limited dependent and count data variables, M-estimators, asymptotic least squares, research and development, innovation, productivity
EC45 INVESTMENT AND R&D
Do R&D Tax Credits Work? Evidence from an International Panel of Countries 1979-94
Rachel GRIFFITHS,
IFS, U.K.
Nicholas Bloom, John Van Reenen
This paper examines the impact of fiscal incentives on the level of R&D investment. An econometric model of R&D investment is estimated on a country level panel including Japan, the US, the UK, France, Italy, Germany, Canada and Australia. Evidence is found that tax incentives are effective in increasing R&D, even allowing for country fixed effects and world macro shocks. The tax price elasticity of R&D is lower than suggested by recent studies which focus solely on the U.S. Additionally there is some evidence that changes in R&D tax credits effects decisions over the international allocation of R&D as suggested by models of tax competition.
Keywords: Tax credits, R&D, panel data
EC46 FORECASTING
Evaluating Density Forecasts
Francis DIEBOLD,
University of Pennsylvania, U.S.A.
Todd Gunther, Anthony Tay
We propose methods for evaluating and improving density forecasts. We focus primarily on methods that are applicable regardless of the particular user's loss function. Throughout, we take explicit account of the relationships between density forecasts, action choices, and the corresponding expected loss. We illustrate the methods with a detailed series of examples, and we discuss extensions to improving suboptimal density forecasts, multi-step-ahead density forecast evaluation, multivariate density forecast evaluation, monitoring for structural change and its relationship to density forecasting, and density forecast evaluation with known loss function.
Keywords: prediction, uncertainty, decision theory, risk management, volatility dynamics
EC46 FORECASTING
A Comparison of the Forecast Performance of Markov-Switching and Threshold Autoregressive Models of US GNP
Hans-Martin KROLZIG,
University of Oxford, U.K.
Michael Clements
While there has been a great deal of interest in the modelling of non-linearities in economic time series, there is no clear consensus as to whether allowing for non-linearities has led to an improved forecast performance. In this paper we evaluate the performance of two leading non-linear models in forecasting post-war US GNP, the self-exciting threshold autoregressive model and the Markov-switching autoregressive model. Two methods of analysis are presented: an empirical forecast comparison accuracy over a historical period, and a simulation-based comparison. We find that, at best, the non-linear models yield some improvement at very short horizons.
Keywords: forecasting, non-linear time series models, Markov-switching, SETAR models, business cycle analysis
EC46 FORECASTING
Measuring Predictability: Theory and Macroeconomic Applications
Lutz KILIAN,
University of Michigan, U.S.A.
Francis Diebold
We propose a measure of predictability based on the ratio of the expected loss of a short-run forecast to the expected loss of a long-run forecast. This predictability measure can be tailored to the forecast horizons of interest, and it allows for general loss functions, univariate or multivariate information sets, and stationary or nonstationary data. We propose a simple estimator, and we suggest resampling methods for inference. We then provide several macroeconomic applications. First, based on fitted parametric models, we assess the predictability of a variety of macroeconomic series. Second, we analyze the internal propagation mechanism of a standard dynamic macroeconomic model by comparing the predictability of model inputs and outputs. Third, we use predictability as a metric for assessing the similarity of data simulated from the model and actual data. Finally, we sketch several promising directions for future research.
Keywords: time series econometrics, predictability of macroeconomic time series, quantitative macroeconomics, model evaluation, propagation mechanisms
EC47 INCOME MOBILITY
On the Non-Stationarity of German Income Mobility
Christian SCHLUTER,
LSE, U.K.
The intra-distributional mobility of German income dynamics is analysed using GSOEP. Transition probabilities are found to be time-varying. The tested models comprise various mixed Markov chains in discrete time and a non-stationary over-stayer model is proposed. In order to explain the observed mobility profiles, we concentrate on one important income class -the poor- instead of the entire transition matrix. Various poverty duration models accomodating unobserved population heterogeneity and duration dependence are examined.
EC47 INCOME MOBILITY
The Persistence of Poverty: Evidence from Spanish Panel Data
Olga CANTO
European University Institute, Italy
This paper analyses, using longitudinal data on Spanish households for the period 1985 and 1992, income dynamics of the household in time and transitions out of poverty when the probability of transiting is studied both as a state duration independent and a state duration dependent variable. Results indicate that there is a group of households who suffer poverty during short periods which one could identify as transitorily poor. This group is seldom captured by cross-section poverty analysis. Regarding the determination of the probability a household has to leave poverty, certain characteristics of the household such as the unemployment situation of the head of household, head over 55 years of age, a large poverty gap or a female head impose a much lower probability of exiting poverty. Some evidence is also found in the data on the duration dependence of the probability of leaving poverty: the longer a household remains in poverty, the more difficult for it to step out of it.
EC48 ECONOMETRICS OF AUCTION MODELS
Affiliated Private Values in OCS Wildcat Auctions
Isabelle PERRIGNE,
University of Southern California, U.S.A.
T. Li, Q. Vuong
This paper analyses the 1954-1969 wildcat auctions on the Outer Continental Shelf of the Mexico Gulf. We consider the affiliated private value model, which allows us to account for companies' efficiencies and opportunity costs, while permitting affiliation among firms' private values. We establish the identification of the model and propose a computationally convenient two-step nonparametric procedure for estimating the joint private value distribution from observed bids. Our empirical analysis shows that the model is supported by the data and that affiliation is significant though weak. We also find that the auction mechanism was unsuccessful in capturing firms' willingness-to-pay for tracts.
Keywords: Affiliated Private Value, OCS Wildcat Auctions, Structural Econometrics, Nonparametric Estimation
EC48 ECONOMETRICS OF AUCTION MODELS
A Structural Econometric Ascending Auction Model, with an Application to the PCS Spectrum Auctions
Han HONG,
Stanford University, U.S.A.
Matthew Shum
We derive a econometric structural model of a rnulti-round ascending price auction in wich bidders update their valuatiorr of an object's value by observing, the dropout behavior of the other bidders. We estimate this model using data from the PCS spectrum auctions run by the FCC. Our two major findings are: evidence of PCS-cellular complementarities and, more significantly, no evidence that the multi-round auction format led to information revelation, and therefore higher sailing prices, relative to an single-round format.
Keywords: auction, learning, rnulti-round, Personal Commucations (PCS), structural econometric model
EC48 ECONOMETRICS OF AUCTION MODELS
Optimal Nonparametric Estimation of First-Price Auctions
Quang VUONG,
University of Southern California, U.S.A.
Emmanuel Guerre, Isabelle Perrigne
This paper considers the structural estimation of first-price auction models within the private value paradigm. First, we solve the nonparametric identification of the bidders' private value distribution from observed bids. Then we propose a computationally convenient nonparametric estimation procedure. In a first step kernel estimation of the density of observed bids is performed to construct a sample of pseudo private values. In a second step this pseudo sample is used to estimate nonparametrically the density of private values. We establish the uniform consistency of our procedure and its optimality in the local minimax sense.
Keywords: First-Price Auctions, Nonparametric Identification, Nonparametric Estimation, Minimax Theory
EC49 CROSS-SECTION DYNAMICS
The Macroeconometrics of Cross-Sectional Heterogeneity
Claudio MICHELACCI,
LSE, U.K.
Detrended aggregate time series display a very high degree of persistence. In frequency domain this shows up in a typical spectral shape. The paper uses this well known empirical regularity as testing bench for a simple transmission mechanism: cross-sectional heterogeneity. To highlight its relevance the paper analyses a vintage model where all aggregate uncertainty takes the form of a reallocative shock that can be driven by either nominal or real factors. The paper argues that this transmission mechanism is an important one as it is powerful, realistic, robust and general. It is powerful because a sufficient amount of cross sectional heterogeneity is able to generate typical spectral shapes without necessarily relying on real shocks. It is realistic because it is shown how two of the most striking empirical regularities in economics, the typical spectral shape and the Gibrat's law, might be just two faces of the same coin. It is robust because it is able to generate almost any kind of spectral shapes provided that reallocations cleanse the economy reallocating between very low and very high technological states. It is general because the paper gives micro-foundation for fractional cointegration of aggregate economic variables high-lightening why a typical spectral shape might be such a generalized phenomenon.
Keywords: Cross-Sectional Heterogenity, Typical Spectral Shape, Macroeconometrics, Gibrat's Law
EC49 CROSS-SECTION DYNAMICS
Europe of Regions or Europe of Nations? A Dynamic Factor Analysis
Catherine FUSS,
Université Libre de Bruxelles, Belgium
The paper groups 144 EU regions' per capita GDP growth rates over the period 1980-1992, according to an OCA criterion, using the dynamic factor analysis of Forni and Reichlin (1996a, 1996b). The results show that (1) an EU core is not defined as a set of nations but as a set of regions: Belgium, most French regions, Northern Italian regions, some German and Dutch regions; (2) there is evidence of a two-speed Europe, Greece, Luxembourg and the United Kingdom respond to the common shock in opposite way to the core; (3) regions of existing monetary unions must not satisfy the OCA criterion.
Keywords: EU regions, European Monetary Union, Optimal Currency Area, factor analysis, comovements
EC49 CROSS-SECTION DYNAMICS
A Partially Linear Additive Regression Model of Cross Country Growth Regressions
Thanasis STENGOS,
University of Guelph, Canada
Zhenuan Liu
Abstract not received from the author
EC50 ISSUES IN MACROECONOMETRICS (A)
Some Temporal Aggregation Issues in Empirical Analysis
Massimiliano MARCELLINO,
European University Institute, Italy
We derive the generating mechanism of a temporally aggregated process when the original one belongs to the ARMA class. We then study the effects of temporal aggregation on a set of characteristics of usual interest such as exogeneity, causality, cointegration and common features. An empirical example illustrates the main issues.
Keywords: Temporal aggregation, ARMA process, invariance
EC50 ISSUES IN MACROECONOMETRICS (A)
Data Transformations and De-Trending in Econometrics
D.S.G. POLLOCK,
Queen Mary College, U.K.
Abstract not received from the author
EC50 ISSUES IN MACROECONOMETRICS (A)
Constructing Quadratic Objective Functions under an A Priori Monotonicity Restriction
Andranik TANGIAN,
Fern Universität Hagen, Germany
Josef Gruber
A model for constructing quadratic objective functions from interviewing experts is considered. The person interviewed is presented a set of incomplete alternatives (vectors of target variables, with one coordinate not being fixed) and is asked to complete these alternatives (to adjust this one coordinate) to make these vectors equivalent in preference to some given reference vector. The objective function is determined by the indifference hypersurface fitted to the equivalent vectors constructed. The optimal fit is constrained by the requirement of monotonicity of the objective function which is proved to be reducable to a finite set of linear inequalities in the function coefficients. The computational implementation of the optimal fit is a modification of Fletcher's `active set method'. Besides formal properties of the model, a WINDOWS computer program will be introduced together with experiments on constructing an objective function of German economic policy in four target variables (inflation, unemployment, GNP growth, and public debt).
Keywords: Preference, objective function, data from interviewing an expert, ordinal regression, constrained least squares, econometric decision mode
EC51 STOCHASTIC VOLATILITY
Specification Tests for Asymmetric GARCH
Gustaf HAGERUD, ,
Stockholm School of Economics, Sweden
Two new Lagrange multiplier statistics for testing the null of GARCH(1,1), against asymmetric GARCH are presented. In a Monte Carlo experiment the empirical size is shown to be equal to the theoretical for reasonable sample sizes. Furthermore the power of both tests is superior to that of the asymmetry tests proposed by Engle and Ng [1993]. This is true if the true data generating process follows any of the models, EGARCH, GJR, TGARCH, A-PARCH, LSTGARCH, GQARCH, and VS-ARCH. Thus, the two tests are tests for general GARCH asymmetry.
Keywords: Asymmetric GARCH, Specification Tests, LSTGARCH, GQARCH
EC51 STOCHASTIC VOLATILITY
Stochastic Volatility Models: Conditional Normality versus Heavy-Tailed Distributions
Robert JUNG,
Universität Tubingen, Germany
Most of the empirical applications of the stochatic volatility (SV) model are based on the assumption that the conditional distribution of returns given the latent variance is normal. In this paper the SV model based on a conditional normal distribution is compared with SV specifications using conditional heavy-tailed distributions, especially the Student t-distribution and the generalized error distribution. To estimate the SV specifications a simulated maximum likelihood approach is applied. The results based on German stock market data reveal that the SV model with a conditional normal distribution does not adequately account for the empirical facts that the distributions of the returns are leptokurtic and that the autocorrelation functions of the squared returns are low but very slowly decreasing. Furthermore it is shown that these emprical facts are more adequately captured by a SV model with a conditional heavy-tailed distribution. Finally, it turns out that the choice of the conditional distribution has systematic effects on the parameter estimates.
Keywords: Volatility persistence of returns, Leptokurtic return distribution, Simulated maximum likelihood, Monte Carlo integration
EC51 STOCHASTIC VOLATILITY
Fourth Moment Structure of the GARCH Model
Changli HE,
Stockholm School of Economics, Sweden
Timo Teräsvirta
In this paper, a necessary and sufficient condition for the existence of the unconditional fourth moment of the GARCH(p,q) process is given as well as an expression for the moment itself. Furthermore, the autocorrelation function of the centred and squared observations of this process is derived. The statistical theory is further illustrated by a few special cases such as the GARCH(2,2) process and the ARCH(q) process.
Keywords: Autoregressive conditional heteroskedasticity, conditional variance, fat-tailed error distribution, time series, volatility
EC52 ECONOMIC GEOGRAPHY/TRADE
Estimating Equilibrium Models of Local Jurisdictions
Holger SIEG,
Duke University, U.S.A.
D. Epple
Abstract not received from the author
EC52 ECONOMIC GEOGRAPHY/TRADE
Trade Patterns, Technology Flows, and Productivity Growth
Wolfgang KELLER,
University of Wisconsin, U.S.A.
Earlier findings of international R&D spillovers suggest that economies benefit from R&D conducted abroad. In this paper, we analyze a particular mechanism through which this might occur: we estimate the extent to which a country benefits from intermediate goods imports which embody new technology--the result of foreign R&D investments. We distinguish this mechanism from others wich are unrelated to international trade. The empirical analysis employs industry level data for eight OECD countries between 1970-91 to estimate the underlying model of trade and growth. Two results concern international R&D spillovers in general: First, the productivity effects of foreign R&D vary substantially by which country is conducting the R&D, suggesting that the quality of newly created technology varies. Second, while estimating substantial international R&D spillovers effects, we also find that the importance of own R&D exceeds that of the average foreign country in determining productivity levels. Our analysis further demonstrates why it is difficult to empirically separate the international spillover effect through importing embodied technology--intermediate goods--from a more general international spillover effect: often, both are simultaneously present. We, thirdly, estimate that in our sample of industrialized countries, international trade contributes about 20% to the total productivity effect derived from foreign R&D. We conjecture that this effect could be higher for less industrialized countries importing from OECD countries, but stress at the same time that alternative mechanisms should be included in studies examining the role of international trade in the process of international technology diffusion.
Keywords: International R&D Spillovers, Intermediate Goods Trade, Embodied Technical Change, Monte Carlo Estimation
EC52 ECONOMIC GEOGRAPHY/TRADE
Specialisation Patterns in Europe
Mary AMITI,
Universitat Pompeu Fabra, Spain
The purpose of this paper is to analyse whether specialisation has increased in European Union countries, and to determine whether specialisation patterns are consistent with trade theories. I present evidence of increasing specialisation in European Union countries between 1968 and 1990. I identify which industries have increased in geographical concentration and show that the characteristics of these industries are consistent with what is predicted by trade theories. The industries with increasing geographical concentration are characterised by high scale economies and high proportions of intermediate goods in production, providing support for the new trade theories and the economic geography theories.
Keywords: geographical concentration, inter-industry trade, intra-industry trade, specialisation
EC53 SAVINGS AND WEALTH ACCUMMULATION
The Wealth Accumulation of Different Cohorts: A Panel Data Analysis
Anna LUSSARDI,
Dartmouth College, U.S.A.
Arie Kapetyn, Rob Alessie
In this paper, we examine the accumulation of wealth over the life cycle for different cohorts. It is well known that individuals born in different periods of time (cohorts) exhibit different accumulation paths. While previous studies have used cohort dummies to proxy for this fact, research in this area suffers from a serious identification problem, i.e., how to disentangle age, time, and cohort effects from a simple cross-section or from time series of cross-sections (pseudo-panels). The innovation we propose in this paper is to go beyond the simple use of cohort dummies to capture the differences in wealth accumulation across individuals born in different time periods. We propose an indicator of economic conditions that represents a good measure of productivity growth over time: the aggregate level of national product when the head of the household entered the labor market. We use a similar methodology to construct another measure of economic conditions: the level and generosity of the Social Security system at the start or during the working life. The empirical work is carried out on a panel data set that provides yearly data on wealth from 1987 to 1991. Panel data allow us to control for many of the factors that can explain wealth and, additionally, we can work with saving, which are derived by first differencing wealth. Our empirical findings show that, while explaining income, productivity growth alone cannot explain all the movements in wealth holdings across different cohorts. We find that Social Security plays also a role and that these two variables alone can explain most, if not all, of the differences in wealth holdings of different cohorts.
Keywords: Life-Cycle Models, Wealth Accumulation, Saving, Panel Data
EC53 SAVINGS AND WEALTH ACCUMMULATION
Measuring Financial Liberalization in the UK Mortgage Market
John MUELLBAUER,
Nuffield College, U.K.
Increased consumer credit availability has important repercussions for consumer behaviour, and hence for macroeconomic fluctuations. The aim of this paper is to derive an index of mortgage credit conditions that captures the financial deregulation of the 1980s together with more recent changes in mortgage markets. Previously proposed measures, including credit/GDP ratios, interest rate differentials between borrowing and lending rates and loan-to-value and loan-to-income ratios, are defective: they depend also on interest rates, income, wealth, expectations of income, of inflation and of asset prices and on consumer uncertainty. This paper analyses a panel of UK regional data on loan-to-value ratios for first-time buyers for 1971-1995. Time effects uniform across regions for the 1980s and 1990s are used to estimate a credit conditions index, afer controlling for economic variables.
Keywords: financial liberalization; mortgage markets, loan-to-value ratios
EC54 VARS
Bias Correction in Vector Autoregressions with Possibly Integrated Processes
Taku YAMAMOTO,
Japan
Eiji Kurozumi
The present paper concerns without the ordinary least squares (OLS) estimator for the lag augmented vector autoregressions with possibly integrated processes. It is known that the OLS estimator is biased in small samples and consequently the empirical size of the Wald test of parameter restrictions is severely distorted. The present paper proposes a version of jackknife method in order to reduce the OLS bias. Firstly, it derives the asymptotic theory for the proposed method. Secondly, several Mote Carlo experiments show that the proposed method effectively reduces the bias and corrects the size distortion of the Wald test, in small samples
Keywords: bias correction, integrated process, jackknife method
EC54 VARS
On The Asymptotics of Maximum Likelihood Estimation In Gaussian Cointegrated Vector Autoregressive Models Under Nonlinear Restrictions
H. BOSWIJK,
University of Amsterdam, The Netherlands
This paper obtains consistency and asymptotic distribution results for maximum likelihood estimators in Gaussian cointegrated vector autoregressive models under non-linear restrictions. The approach is to derive boundedness in probability of a normalized estimator and unform weak convergence of thelog-likelihood ratio on shrinking compact sets. Specific attention is given to partially identified parametrizations. Some examples illustrate the results.
Keywords: Maximum Likelihood, Cointegration, Asymptotic Theory, Non-Ergodic
EC54 VARS
Cointegrated VAR-X Models
Rocco MOSCONI,
Politecnico di Milano, Italy
Anders Rahbek
A Gaussian VAR with exogenous variables (VAR-X) is discussed. The exogenous variables are generated by a Gaussian VAR, and may be stationary or I(1) (possibly cointegrated). We show that, when the exogenous variables are allowed to be non stationary, inference on the VAR-X requires some preliminary analysis of the marginal subsystem. In particular, the analysis of the VAR-X is asymptotically equivalent to the analysis of the full system, provided that the contegration rank and the cointegrating vectors among the exogenous variables have been consistently estimated in the marginal subsystem. The issue of identification of the long run coefficients in the proposed model is also addressed.
Keywords: Cointegration, VAR-X model, Identification
EC54 VARS
Autocovariance Functions and Maximum Likelihood in a VAR Model Under Markov Switching
Anders WARNE,
Stockholm University, Sweden
Abstract: The mean and the autocovarinaces conditional on the regime are derived for a Markov switching VAR model and related to the unconditional mean and autocovariances. The former parameters can be particularly useful when attempting to interpret the regimes, while properties of the latter make it possible to evaluate the necessity of a sufficient condition for weak stationarity. Based on the ML estimator, it is shown that the conditional moments can be approximated by weighting the data with the estimated smooth probabilities, thus simplifying the computation considerably. For monthly US data (1959-95) on income and money, the differences between the estimated moments are negligable.
Keywords: Autocovariances, Markov process, maximum likelihood, regime switching, stationarity, vector autoregression
EC55 MACROECONOMIC MODELLING
Evaluating Gains to Cooperation in the G-3
Brian HENRY,
LBS, U.K.
Guglielmo Maria Caporale, Michael Chui, Stephen Hall
A new model of the G-3 is introduced which has a transparent structure including a simple but analytically tractable supply side. Extenstive policy experiments are conducted on this multicountry model including full Nash solutions and a fully cooperative solution. In the case of a fiscal shock originating in the US we find evidence of significant spillovers to other countries, which can only be ameliorated in the fully cooperative regime.
Keywords: Supply side, Prices and Wages, Policy Rules, Solvency and Inflation Targets, Econometric Modelling
EC55 MACROECONOMIC MODELLING
Asymmetries in Unemployment Rates: An International Comparison
Claudio LUPI,
ISPE, Italy
Patrizia Ordine
Abstract not received from the author
EC55 MACROECONOMIC MODELLING
Procyclucal Labor Productivity, Labour Effort and Technology Shocks A Study on Injury Rates in Swedish Manufacturing 1972-1992
Jyry HOKKANEN,
Uppsala University, Sweden
In empirical studies the finding of procyclical average labor productivity is common. Three possible explanations to these results have been suggested in the literature: (1) Increasing returns to scale/Imperfect competition, (2) Labor hoarding/labor utilization, and (3) the RBC productivity/technology shock explanation. This study tries to discriminate between the two latter hypotheses by estimating the labor input coefficient in a production function regression with instrumental variables that are assumed to be unrelated to technology shocks, and, by using industrial injury rates as a proxy for unobserved labor effort. The results from panel data estimations for the Swedish industry are in favor of the labor hoarding/utilization hypothesis.
Keywords: injury rates, labor effort, panel data
EC55 MACROECONOMIC MODELLING
Cyclical Effects of the Composition of Government Purchases
Jahangir AZIZ,
IMF, U.S.A.
Luc Leruth
Abstract not received from the author
EC56 SIMULATION ESTIMATION AND INFERENCE
Equality Restricted Random Variables: Densities and Sampling Algorithms
Frank KLEIBERGEN,
Erasmus University Rotterdam, The Netherlands
Many common statistical models can be specified as linear models with restrictions imposed on the parameters. A large amount of these models impose restrictions which do not allow for the analytical construction of the probability density function (pdf) of the parameters given the restrictions. This is often implicitly assumed which leads to an inconsistency as the pdf of the parameters of the linear specification under the imposed restrictions is then not nested within the assumed pdf of the unrestricted linear specification. The paper shows how these restrictions need to be incorporated by constructing the pdfs incorporating the restrictions and algorithms to sample from these pdfs. We show how these methods are applied to some common statistical models, i.e. ARMA, cointegration and simultaneous equation models.
EC56 SIMULATION ESTIMATION AND INFERENCE
Bayesian Estimation of Switching ARMA Models
Monica BILLIO,
Universita Da Foscari, Italy
A. Monfort, C.P. Robert
Switching ARMA process have recently appeared as an efficient modelling to nonlinear time series models, because they can represent multiple or heterogeneous dynamics through simple components. The levels of dependence between the observations are double: at a first level, the parameters of the model (AR, MA or ARMA) are selected by a Markovian procedure. At a second level, the next observation is generated according to a standard time series model. When the model involves a moving average structure, the complexity of the resulting likelihood function is such that simulation techniques, like those proposed by Shephard (1994) and Billio and Monfort (1995), are necessary to derive an inference on the parameters of the model. We propose in this paper a Bayesian approach with a noninformative prior distribution developed in Mengersen and Robert (1996) and Robert and Titterington (1996) in the setup of mixtures of distributions and hidden Markov models, respectively. The computation of the Bayes estimates relies on MCMC techniques which iteratively simulate missing states, innovations and parameters until convergence. The performances of the method are illustrated on both simulated examples and real data. We show in addition how maximum likelihood estimators can be derived from these Bayes estimators through increasing replications of the original sample, following Robert (1993) and Robert and Titterington (1996). This work also extends the papers by Chib and Greenberg (1994) and Chib (1996) which deal with ARMA and hidden Markov models, respectively.
Keywords: Markov model, Moving Average model, Hidden Markov chain, MCMC algorithm, ARMA model, Convergence control, Prior feedback
EC56 SIMULATION ESTIMATION AND INFERENCE
Simulation Based Finite and large Sample Inference Methods in Simultaneous Equations
Jean-Marie DUFOUR,
Université de Montreal, Canada
Linda Khalaf
Abstract not received from the author
EC56 SIMULATION ESTIMATION AND INFERENCE
Simulated Closed-Form Estimators for the Multinomial Probit Model
Tom WANSBEEK,
University of Groningen, The Netherlands
Michel Wedel
The multinomial probit model is an often used, general and flexible model to describe the choice from many alternatives. Its estimation is not straightforward since the choice probabilities implied by the model involve M-1-dimensional integrals, where M is the number of alternatives. Over the recent years, the numerical problems have been greatly alleviated by simulation methods. By exploiting the structure of the model to a larger extent than seems to have been done before, we arrive at simple expressions for the score vector and the information matrix. These expressions allow us in particular to propose new simulationbased estimators for the model parameters which are of an extremely simple form.
Keywords: multinomial probit, maximum likelihood, simulation estimation
EC57 SPECIFICATION TESTS/TIME SERIES
The Power of Some Tests for Difference Stationarity under Local Heteroskedastic Integration
Richard SMITH,
University of Bristol, U.K.
Brendan McCabe
This paper considers the power properties of a test (McCabe and Tremayne (MT) 1995) for the difference stationarity of a time-series. The limiting distribution of the MT test is derived under a sequence of locally heteroscedastic and locally explosive autoregressive (AR) alternatives. The limiting distribution of Dickey-Fuller (DF) (1979) statistics is also considered under a sequence of locally heteroscedastic alternatives. Whereas the MT test possesses asymptotic power against both forms of nonstationary local alternative, the DF tests set up to test against explosive AR alternaives display little or no ability to reject difference stationarity under local heteroscedastic integration.
Keywords: Autoregression, Brownian Motion, Dickey-Fuller Test, Nonstationarity, Unit Root
EC57 SPECIFICATION TESTS/TIME SERIES
On the Properties of the Dickey-Pantula Test Against Fractional Alternatives
Juan DOLADO,
Universidad Carlos III, Spain
Francesco Marmol
The limit properties of the testing sequence underlying the Dickey-Pantula test for a double unit root in a time series are derived when the true data generating process is asumed to be nonstationary fractionally integrated.Our basic conclusion is that that the mechanical application of the DP procedure can lead to misleading results,extending previous results by Sowell (Econometrica,1990) on the properties of the DF unit root test under fractionally integrated alternatives.
EC57 SPECIFICATION TESTS/TIME SERIES
Nonparametric Lag Selection for Time Series
Rolf TSCHERNIG,
Humboldt University, Germany
Lijian Yang
A nonparametric version of the Final Prediction Error (FPE) was recently proposed for lag selection in nonlinear autoregressive time series. We derive its consistency where we allow for both, local constant or linear fits. Further asymptotic analysis indicates a greater probability of overfitting than underfitting. To penalize overfitting in small samples we propose a correction factor. Our Monte-Carlo study suggests that the correction factor generally improves the probability of correct lag selection in the presence of linear and nonlinear DGPs. The proposed methods are successfully applied to the Canadian lynx data and daily returns of DM/US-$ exchange rates.
Keywords: consistency, Final Prediction Error, foreign exchange rates, lag selection, nonlinear autoregression, nonparametric methods
EC57 SPECIFICATION TESTS/TIME SERIES
On Adaptive Tests
Param SILLVAPULLE,
La Trobe University, Australia
Mervyn Silvapulle, Ishwar Basawa
Abstract not received from the author
EC58 UNEMPLOYMENT
A Conjecture on the Explanation for High Unemployment in the Industrialized Nations: Part I
Andrew OSWALD,
University of Warwick, U.K.
This paper conjectures that the high unemployment of the Western economies has been produced by the decline of the private house-rental market and the rise of home-ownership. Evidence is provided for the developed nations, the states of the USA, and the regions of the UK, Italy, France and Sweden. Although its calculations should be viewed as tentative, the paper's results imply that a 10 percentage point rise in the owner-occupation rate is associated with an increase of approximately 2 percentage points in the unemployment rate. This would be sufficient to explain a significant part of the observed rise in joblessness.
Keywords: Unemployment, housing
EC58 UNEMPLOYMENT
European Regional Unemployment. The Interaction of Education and Labor Institutions
Alicia ADSERA,
Universitat Autonoma de Barcelona, Spain
Carles Boix
In the last fifteen years two equilibria have arisen in the advanced world. On the one hand, wage dispersion has widened in those countries where unemployment has remained low (with cyclical variations). On the other hand, wherever income inequality has remained unchanged, unemployment has shot upwards. To account for these distinct patterns, a combination of current theories - focusing separately on either technological and trade shocks or institutional arrangements - is required. In a simple model, we show that, controlling for the skills of the population, the effects of exogenous shocks are contingent on the institutional rules in place. Economies with generous unemployment allowances adjust through subsidized unemployment. Lower social protection leads to less unemployment but wider inequality. The model is successfully tested for the sample of European regions and US states.
Keywords: Unemployment, income distribution, inequality, education, institutions, corporatism
EC58 UNEMPLOYMENT
The Puzzle of Jobless Growth in Indian Manufacturing
Sonia BHALOTRA,
University of Bristol, U.K.
In the 1980s, India experimented with deregulation in industry and trade. Manufacturing output accelerated but employment declined, raising doubts about the desirability of the policy reforms. This paper proposes an explanation of employment behaviour in terms of increases in total factor productivity, in actual hours worked, and in the product wage. Using robust methods, it is shown that neglect of hours worked results in a substantial upward bias in estimates of the wage elasticity. Growth in productivity and hours appears to be associated with the reform process, with the increase in hours worked reflecting recovery of lost time. To the extent that hours must hit a ceiling, the drop in employment on this count is expected to be temporary. Other things being equal, employment prospects appear to depend considerably on the nature and the sustainability of productivity growth.
Keywords: employment, hours, deregulation, methodology, panel data, India
EC58 UNEMPLOYMENT
Unemployment Dynamics and Age
A.G.C. VAN LOMMEL
Erasmus University Rotterdam, The Netherlands
G.T. Van den Berg, J.C. Van Ours
Unemployment in general and youth unemployment in particular, is among the top policy issues in Western European countries. In this paper we analyze unemployment dynamics of three agegroups, i.e. youths, adults and elderly. We use quarterly French unemployment data, stratified by gender-type and agegroup on the period 1982-1994. We find the business cycle to effect the outflow probabilities of youths and adults in a similar way, but the seasonal effect to be different. Moreover, we find different unobserved heterogeneity distributions and duration dependence patterns for youths and adults.
Keywords: unemployment, unemployment duration, business cycle, duration dependence, seasonality, age
EC59 DISCRETE CHOICE MODELS
A Simultaneous Probit Model
Marcel DAGENAIS,
CRDE Université de Montreal, Canada
Simultaneous probit models in which the binary endogenous variables appear as explanatory variables may raise problems of coherence. For example, in the two equations model, under the usual assumptions based on the bivariate normal distribution, coherence precludes simultaneous feedback effects between the two dichotomic endogenous variables. However, this constraint seems to be too restrictive in certain applications. The purpose of this paper is to suggest appropriate modifications to the stochastic hypotheses to eliminate the above constraint. Numerical examples illustrate the proposed approach.
Keywords: probit, simultaneous equations, coherence
EC59 DISCRETE CHOICE MODELS
Testing Models of Labour Supply with Discrete Choices
Melvyn WEEKS,
The University of York, U.K.
Alan Duncan
It is often argued that institutional factors constrain the labour supply choices open to the individual. In this paper we present an alternative to the standard neoclassical model of labour supply under the presumption that the range of hours choices are finite. We examine alternative specifications of discrete choice model, using both nested and non-nested statistical tests. Finally, we evaluate the precision of simulated transitions following a policy reform.
Keywords: discrete choice models, non-nested tests, labour supply
EC59 DISCRETE CHOICE MODELS
Utility Maximization and Mode of Payment
Ruud KONING,
Vrije Universiteit, The Netherlands
The hypothesis of utility maximization imposes well-known restrictions on demand equations. In this paper, we discuss the analogue of these conditions in discrete choice models. Recently, Börsch-Supan (1990) and Koning and Ridder (1991,1994) have developed conditions on choice probabilities that ensure compatibility with stochastic utility maximization. These conditions can be applied to both parametric and semi- and nonparametric models. We apply these so-called `compatibility conditions' to a simple model of mode of payment.
Keywords: discrete choice, utility maximization, applied microeconomics
EC59 DISCRETE CHOICE MODELS
Bayesian Identification of Semi-Parametiric Binary Response Models
Michel MOUCHART,
Université Catholique de Louvain, Belgium
Jean-Marie Robin, Eliana Scheihing
In this paper, minimal conditions under which a semi-parametric binary response model is Bayesian identified are presented and compared to the inferential usual conditions.
Keywords: Binary response models, Non parametric Bayesian Statistics, Dirichlet processes, Identification
EC60 FRACTIONALLY INTEGRATED PROCESSES
The Functional Central Limit Theorem and Weak Convergence to Stochastic Integrals: Results for Weakly Dependent and Fractionally Integrated Processes
James DAVIDSON,
Cardiff Business School, U.K.
Robert de Jong
This paper gives new conditions for the functional central limit theorem and weak convergence of stochastic integrals for near-epoch dependent functions of mixing processes. These results have applications in the theory of unit root testing and cointegrating regressions. The conditions improve on existing results in the literature in terms of the amount of dependence and heterogeneity permitted. In particular, these appear to be the first such theorems in which virtually the same assumptions are sufficient for both modes of convergence. Results are also given for fractionally integrated (or I(d)) processes.
Keywords: functional central limit theorem, stochastic integrals, near-epoch dependence, fractional integration
EC60 FRACTIONALLY INTEGRATED PROCESSES
Time-Domain Methods for the Estimation of Fractionally-Integrated Time Series Models
Victoria ZINDE-WALSH,
McGill University, Canada
John Galbraith
Long memory models such as the autoregressive moving average fraction-ally integrated model, ARFIMA(p,d,q), have come to play an increasing role in economics and finance as longer time series have become available. The estimation method we consider in the paper falls in the class of time-domain methods where all parameters are estimated simultaneously. The method is based on autoregressive approximation of an ARFIMA process which provides asymptotically normal estimators. The simultaneous estimation of the parameters offers the potential for lower bias, and therefore mean squared error, in mixed models. The estimator is straightforward to compute; its finite sample performance relative to that of the most often used estimators was examined in a number of Monte Carlo experiments for an ARFIMA(1,1,0) model. The results demonstrate comparatively good performance over a large part of the parameter space, with relatively small loses in efficiency where they occur.
Keywords: fractionally integrated model, autoregressive approximation
EC60 FRACTIONALLY INTEGRATED PROCESSES
Fractional Integration versus Trend Stationarity in Time Series Analysis
Francesco MARMOL,
Université de Toulouse, France
Recently, Lee and Schmidt (1996) show that the Kwiatkozski, Phillips, Schmidt and Shin (KPSS) test is consistent against stationary fractionally integrated alternatives. In this paper we prove that this test is also consistent against nonstationary fractionally integrated alternatives. On the other hand, Sowell (1990) and Haldrup and Marmol (1997) show that the standard Dickey-Fuller (DF) test is consistent against mean reverting alternatives. Herein we prove that the DF test is also consistent against stationary fractionally integrated alternatives. The finite sample behavior of both tests as well as their usefulness in real applications are also discussed.
Keywords: fractionally integrated processes, DF test, KPSS test, inflation series
EC60 FRACTIONALLY INTEGRATED PROCESSES
Fractional Cointegration: Bayesian Inferences Using a Jeffreys Prior
Gael MARTIN,
Monash University, Australia
The concept of fractional cointegration, whereby deviations from an equilibrium relationship follow a fractionally integrated process, has attracted some attention in the literature of late. The extended concept allows cointegration to be associated with mean reversion in the error, rather than requiring the more stringent condition of stationarity. This paper presents a Bayesian method for conducting inference in a fractional cointegration framework. The results are based on an approximation of the exact likelihood function, with a Jeffreys prior being used to offset an identification problem in the model. Inferences are based on marginal posterior density functions, estimated via a combination of Markov chain Monte Carlo sampling algorithms. The procedure is applied to several purchasing power parity relations, with substantial evidence found in favour of parity reversion.
Keywords: Fractional cointegration, Bayesian inference, Jeffreys prior, Markov chain Monte Carlo
EC61 SEASONALITY (A)
Approximating the Census X-11 Filter
Carlos LENZ,
Universität Bern, Switzerland
This paper shows how to construct an approximation to the seasonal filter proposed by Sims (1974) and compares this filter to the linear version of the Census X-11 filter. The comparison makes clear that a very close approximation to the X-11 filter can be obtained with a filter that consists of less than one half of the terms that enter the X-11 filter. In addition it is shown that the X-11 filter is well-suited to deseasonalise time series with certain properties but is not flexible enough to be generally applied. Keywords: seasonal adjustment, band-pass filtering, Census X-11
EC61 SEASONALITY (A)
Decision Bounds for Data-Admissible Seasonal Models
Robert KUNST,
Institute for Advanced Studies, Austria
Johannes Kepler
This paper aims at identifying adequate models for the seasonal behavior in univariate and multivariate time series. The decision between unit-root seasonality and deterministic cycles and, in multivariate models, the number of independent stochastic seasonal factors form discrete parameters of interest. Attention is restricted to data-admissible models. To this aim, an attempt is made to provide an adequate definition of data admissibility. Among data-admissible models, decision rules are constructed from weighting priors and decision bounds analysis. The procedure is applied to exemplary economics series. The evidence supports models with seasonal unit roots and restricted influence of seasonal constants.
Keywords: seasonality, time series, decision bounds analysis, seasonal cointegration
EC61 SEASONALITY (A)
Determining the Order of Differencing in Seasonal Time Series Processes
Robert TAYLOR,
The University of York, U.K.
Philip Hans Frances
In this paper we propose a sequential testing procedure to determine the order of differencing in seasonally observed time series processes, which builds existing approaches developed for nonseasonal series. Multiple unit roots at the seasonal frequencies can be useful in circumstances where one does not wish to take logarithms of the data-set at hand and also appear in commonly applied seasonal adjustment filters such as Census X-11. The testing procedure developed in this paper can therefore be used to obtain some it a priori insight into the likely properties of seasonally adjusted time series. In particular it may be used to investigate whether or not the seasonally adjusted series can be expected to be strictly noninvertible at the seasonal frequencies of the data. The proposed testing procedure is shown be asymptotically consistent and, unlike existing approaches, is designed such that the size of the procedure may be controlled by the practitioner. Empirical applications are provided to illustrate the practical usefulness of our approach.
Keywords: multiple seasonal unit roots, seasonal adjustment
EC61 SEASONALITY (A)
A Two-Step Procedure for the Estimation and Testing of Seasonal Cointegration in Monthly Data
Andreu SANSO,
Universitat de Barcelona, Spain
Jordi Surinadi, Manuel Artis
In this paper we extend the Engle-Granger-Hylleberg-Lee two-step procedure to estimate and test seasonal cointegration in monthly data. We consider the seasonal error correction representation, we calculate the asymptotic distributions and tabulate critical values. Besides, the paper deal with both contemporaneous and polynomial cointegration and formalize spurious regressions at seasonal frequencies.
Keywords: Seasonality, Cointegration, Unit roots
EC62 NONPARAMETRIC ANALYSIS
A Test of Subadditivity
Jean-Pierre FLORENS,
Université de Toulouse, France
This paper proposes an equivalent definition of the subadditivity of a conditional expectation function which may be tested without any estimation of this function. The test procedure requires an estimation of the density of the explanatory variables which can be done parametrically or non-parametrically. In this last case we present a -convergent fully non-parametric test procedure.
Keywords: subadditivity, semiparametric estimation, Kernel smoothing, weighted bootstrap
EC62 NONPARAMETRIC ANALYSIS
Indirect Nonparametric Estimation of a Cumulative Distribution Function with Econometric Applications
Costin PROTOPOPESCU,
Université de Toulouse, France
Jean-Pierre Florens
A large class of game theoretic econometric models may be formalized by the following specification: are i.i.d. signals generated by a distribution F and the actions satisfies where is the strategy derived, e.g., from a Nash equilibrium rule. The structural model specifies , the 's are observable but the are unobservable for the econometrician. The distribution F (or its parameters) are unknown. This paper is concentrated on non parametric estimation of F. The identification is first considered and an indirect estimator based on a fixed point functional theorem is constructed. The speed of convergence depends on the "order" of the game related to the property of the function .
Keywords: Game theoretic models, non parametric identifiction, structural kernel estimation
EC62 NONPARAMETRIC ANALYSIS
Speculative Bubbles and Excess Returns in European Exchange Rates. Evidence from a Nonparametric Approach
Giuseppe DE ARCANGELIS
University of Rome "La Sapienze", Italy
Andrea Bubula
This paper evaluates whether excess returns on holding Deutschmarks against Francs, Liras and British Pounds have been recently characterized by (temporary) speculative bubbles. We propose a two-stop, distributionfree procedure. First, nonzero-median subperiods are significantly withdrawn from the original sample by an elaborate sign test that avoids the objection of data-mining. Second, we apply the Wilcoxon rank test on all selected subsamples. Nonzero-median excess returns characterize all exchange rates, even when adjusting for risk. Heuristically, a speculative bubble seems to be a plausible explanation (especially for Lira/DM) rather than a peso problem or learning.
Keywords: Sign Test, Wilcoxon Rank Test, Efficiency Market Hypothesis, Monte Carlo Experiment, Risk Premium, Peso Problem
EC62 NONPARAMETRIC ANALYSIS
Testing Serial Independence via the Empirical Characteristic Function: A Generalized Spectral Distribution Function Approach
Y. HONG,
Cornell University, U.S.A.
By generalizing the concept of the spectral distribution function via the characteristic function, we propose a test for serial independence by comparing a sample generalized spectral distribution function with a restricted null. The test can detect linear and nonlinear dependencies. They avoid any smoothing parameter and have power against root-n local alternatives. It applies to discrete and continuous distributions, with possibly infinite moments. Tests based on the conventional spectral distribution function can be derived by differentiating the generalized spectral distribution function with respect to some parameters at the origin. We examine the finite sample performance of the test in comparison with a variety of existing tests for serial independence.
Keywords: Empirical characteristic function, frequency domain analysis, serial independence, spectral distribution function, weak convergence
EC63 THE ECONOMICS OF THE FAMILY
Marriage Settlements
Anne LAFERRERE,
INSEE CREST, France
The paper offers models of choice of a marriage settlements (community property versus separate property agreements) and test them on a representative sample of 6135 French married couples. Three models are offered: a cooperative model in view of providing for widowhood, a non cooperative model in view of divorce, and a bargaining model, each with clear testable predictions. A separate property agreement goes along with high level of wealth, both human and non human, higher involvement of the wife in market production, less children and higher differences between spouses' endowments.
Keyword: marriage, family bargaining, contracts
EC63 THE ECONOMICS OF THE FAMILY
Dowry Inflation: A Comment
Lena EDLUND,
University of Michigan, U.S.A.
Dowries have risen in India. Rao (1993) proposed scarcity of men (marriage squeeze) as an explanation, ignoring the fact that women are increasingly scarce in India. Re-examining the same data as Rao used, I show that his results are not robust. The marriage squeeze variable fails to be significant in reasonable alternative specifications. Instead, the empirical results suggest that "dowry inflation" might be a case of omitted variable bias, the omitted variable being wealth, and not of a rising price of husbands as is commonly believed. This interpretation challenges the notion of dowries as determinants of gender bias.
EC63 THE ECONOMICS OF THE FAMILY
Testing Between Cooperative and Noncooperative Models of Intra-household Allocation
Valérie LECHENE,
Hedm-INRA, France
Martin Browning, Wasif Rasheed
Several models of intrahousehold decision making have been suggested in the literature. The principal dichotomy is between non-cooperative and cooperative models (including specific models of bargaining). We present tests between these models using Canadian household data. We also present tests for the presence of altruism within the household.
Keywords: intrahousehold decision making, cooperative models, non cooperative models
EC63 THE ECONOMICS OF THE FAMILY
Family Labour Supply
Pierre CHIAPPORI,
DELTA, France
Richard Blundell, Thierry Magnac, Costas Meghir
This paper applies nonparametric demand theory to the nonparametric statistical analysis of consumer demand. It exploits the idea that price taking individuals in the same market face the same relative prices, in oder to smooth across the demand of individuals for each common price regime. This is shown to provide a stochastic structure within which to examine the consistency of individual data and revealed prel`erence theory. Using a long time series of repeated cross-sections from the 1974-1993 UK Family Expenditure Surveys, we examine whether revealed preference theory is rejectod for particular types of individuals or in particular subperiods of the data.
Keywords: demand analysis, consumer behaviour, nonparametric regression, revealed preference
EC64 AR MODELS
Adjusted Profile Likelihood Applied to a Unit-Root AR(1) Model with Constant
Pekka PERE,
University of Helsinki, Finland
The adjusted profile likelihood (APL) of Cox and Reid (1993) is applied to an AR(1) model with constant. A Wald statistic based on APL is proposed. The Cox--Reid adjusted estimate (AE) for the autoregressive coefficient of the unit-root AR(1) model with zero constant is even asymptotically more accurate, in terms of mean square error (MSE), than the maximum likelihood estimate (MLE). The tests are more powerful than the corresponding Dickey--Fuller tests if the starting value of the process deviates sufficiently from the unconditional mean. We derive additionally an estimate and a Wald statistic which are asymptotically distributed compactly and symmetrically around zero under a unit root. The estimate is asymptotically the most accurate under a unit root but is not consistent in general, though, which can lead to a deterioration of power.
Keywords: Unit root, adjusted profile likelihood
EC64 AR MODELS
Exact Geometry of Explosive Autoregressive Models
Kees Jan VAN GARDEREN
Université Catholique de Louvain, Belgium
This paper derives exact expressions for statistical curvature and related geometric quantities in the first order autoregressive models with stable and unit roots, as well as explosive roots larger than unity. We develop a method for deriving exact moments of arbitrary order in general autoregressive models. The covariance of the minimal sufficient statistic is an applications of this method. Of particular interest is the Efron curvature which is continuous and bounded in finite samples, but increases rapidly when the autoregressive parameter changes from stable to explosive values. The initial value effect is also quantified exactly.
Keywords: curved exponential models, differential geometry, exact distribution theory, statistical curvature, time series, unit roots
EC64 AR MODELS
The Bias and Mean Squared Error of the Least Squares Estimator in a Dynamic Regression Model With a Unit Root
Garry D.A. Phillips
University of Exeter, U.K.
Jan F. Kiviet
This paper uses large-T asymptotics in a dynamic regression model to find approximations to the bias and mean squared error. The results are specialised to the case of the AR(1) model with a constant where it is seen that the ratio of the standard deviation of the disturbance to the constant term plays a crucial role.This case is also examined in a set of Monte Carlo experiments which are used to check the accuracy of the approximations.
Keywords: dynamic regression, unit root, moment approximations
EC64 AR MODELS
Is the Unemployment Rate Non-Stationary or Non-Linear? A Test for Unit Roots Against Threshold Autoregressions
Mehmet CANER,
Koc University, Turkey
Bruce Hansen
In econometrics many testing problems involve unidentified nuisance parameters under the null of unit roots. The limit theory for such tests has not been established before. This article tries to tackle this problem. The existing unit root tests suggest that macroeconomic variables such as unemployment rate are I(1). Even though this may be true locally , globally this result might occur due to the nonlinearity present in the data. So specifically the null of unit root is tested against the alternative of a threshold autoregressive model. We find that the limit laws are functionals of stochastic integrals with Brownian motions.
EC65 FERTILITY/STRUCTURAL MODELS
Uncertain Child Mortality, Learning and Life-Cycle Fertility
Pedro MIRA,
CEMFI, Spain
I examine the links between infant mortality and fertility in an environment whith unobserved heterogeneity in infant mortality rates across mothers. In such an environment, fertility responses to experienced child deaths (replacement behavior) might be influenced by mothers' learning about a family-specific component of infant mortality rates. I explicitely introduce learning by mothers in a dynamic stochastic model of life cycle marital fertility behavior. Maximum likelihood estimates of the model's structural parameters are obtained from Malaysian panel data. I show how increases in infant survival rates induced a small decline in fertility, larger increases in family size, and population growth.
Keywords: Dynamic discrete choice models, structural estimation, Bayesian learning, fertility, demographic transition, replacement behavior
EC65 FERTILITY/STRUCTURAL MODELS
Structural Estimation of Key Labour Market Characteristics in Five OECD Countries, Using Aggregate Data and an Equilibrium Search Model Framework
Gerard VAN DEN BERG
Free University Amsterdam, The Netherlands
Geert Ridder, Niels de Visser
In this paper we develop an estimation method for labor market characteristics in the context of an equilibrium search and matching framework. The method uses aggregate data on marginal distributions of unemployment and job durations and wages. We estimate the degree of labor market frictions, the magnitude of structural and frictional unemployment, and the monopsony power of firms, and we decompose wage variation. Estimation of some of the characteristics is invariant to the way in which wage determination is modeled. We perform separate empirical analyses for the USA, the UK, France, Germany and The Netherlands.
Keywords: wages, search frictions, cross-country comparisons, unemployment, monopsony
EC65 FERTILITY/STRUCTURAL MODELS
Cohort Effects on Career Patterns. Evidence from Italy, Spain and the UK
Patrizia CANZIANI,
LSE, U.K.
This paper addresses the question of whether individual career patterns are affected by the economic conditions when the workers first entered the labour market.Individual data for Italy, Spain and the UK are analysed in an effort to characterise individual career patterns.I find that in Italy current wages and occupational levels are negatively correlated to the unemployment rate at the beginning of the first job only for people with low or very high qualifications. Second, in Spain there is no correlation between wages and unemployment. Apparently, the nature of the employment contract plays an important role at explaining the level of individual wages. Finally, in the UK only people with high qualifications are adversely affected by a slack labour market at their entry into the labour market.
Keywords: Cohort effects on Wages, European Unemployment
EC65 FERTILITY/STRUCTURAL MODELS
A Structural Estimation of a Model With Optimal Participation and Search
Thierry KAMIONKA,
Université de Toulouse, France
In this paper we study a structural model of transition between employment, unemployment and nonparticipation. We model the participation decision. People can enter temporarily into the out-of-labor-force state when they participate to the labor market. We show that the optimal strategy in unemployment depends on the rates of exit from the other states. We consider the estimation problems. We propose to use the pseudo maximum likelihood estimator in order to obtain the estimates of the parameters of the structural model as well as the asymptotic distribution of the estimator.
Keywords: Labor market, Job search, Participation, Pseudo Maximum likelihood estimator
EC66 ISSUES IN MACROECONOMETRICS (B)
Statistical Inference for Computable General Equilibrium Models, with Application to a Model of the Moroccan Economy
Jean-Marie DUFOUR,
Université de Montreal, Canada
Touhami Abdelkhalek
Computable general equilibrium (CGE) models have become over recent years important tools of economic analysis. They are used to evaluate the effects of commercial and taxation policies, regional economic integration as well as other policies, especially in developing countries where data are scarce. Such models typically contain free parameters, which must be assessed by statistical or theoretical methods, and calibration parameters which are evaluated by ''calibrating'' the model on the data of a reference year. Of course the uncertain character of the parameters affects the reliability of the simulation results. We study the problem of measuring the uncertainty of CGE model simulalulations in relation to parameter uncertainty. The general setup considered is similar to the one of Pagan and Shannon (1985). We show first how calibration as usually performed in CGE models can easily be covered by this setup. Then, we describe two systematic approaches for assessing simulation uncertainty, which are based on building confidence sets for the endogenous variables of the model, given minimal information on parameter uncertainty. The first one is a direct extension of the approach proposed by Pagan and Shannon (1985). It is based on a standard Wald-type statistic and assumes that consistent asymptotically normal estimators are available for the free parameters of the model. The second approach assumes that a confidence set (sampling or Bayesian) is available for the free parameters. Given this confidence set, we can then obtain valid confidence sets for the variables of interest by a projection technique. This approach has two important advantages: first, the validity of the confidence sets is not affected by the nonlinear character of the model; second, it allows one to easily build simultaneous confidence intervals for an unlimited number of variables of interest (or transformations of these). Further. we study general conditions under which these confidence sets are connected (not a union of disjoint sets) and/or take the form of intervals. Numerical procedures required to apply these procedures are also discussed. In particular, we show that valid projection-based confidence sets can be obtained easily by using standard methods for solving CGE models [e.g., routines available in GAMS]. We present an application of these procedures to a CGE model of the Moroccan economy and study the economic effects of policy induced increases of transfers from Moroccans working abroad.
Keywords: Confidence interval, Confidence set, Computable general equilibrium models, Sensitivity analysis, Calibration; Projection, Morocco, Fiscal policy
EC66 ISSUES IN MACROECONOMETRICS (B)
Bayesian Simultaneous Equation Analysis using Reduced Rank Structures
Herman VAN DIJK
Erasmus University, The Netherlands
Frank Kleibergen
The use of flat priors in the analysis of SEMs is shown to lead to pathological posterior behavior. The origin of this pathological behavior lies in the local nonidentification of certain parameters in SEMs. When this, a priori known, feature is not captured in the prior, an a posteriori favor for certain specific parameter values will result which is not the consequence of strong data information but of model inadequacies. We, therefore, suggest priors, which are based on conditionalization rules, resulting from the reduced rank structure of the implied reduced form. This reduced rank structure also shows that the rank and order conditions can differ between a limited and full information analysis. The priors are constructed for both limited and full information analyses and are used to construct Importance and Metropolis-Hastings Samplings procedures.
Keywords: Reduced Rank Structure, Markov Chain Monte Carlo, Simultaneous Equation Models, Bayesian Inference
EC66 ISSUES IN MACROECONOMETRICS (B)
Multiple Regime Smooth Transition Autoregressive Models with an Application to US real GNP
Dick VAN DIJK
Erasmus University, The Netherlands
Philip Hans Franses
A limitation of Smooth Transition AutoReg;ressive [STAR] models is that they can deal with only two regimes. The purpose of this paper is to explore generalizations of the STAR model such that more than two regimes can be accommodated. We introduce the class of Multiple Regime STAR [MRSTAR] models, which can be obtained from the basic model in a fairly straightforward way. We discuss the main properties of the model and several issues which might be relevant for empirical specification. In particular, we derive a Lagrange Multiplier-type test which can be used to determine the appropriate number of regimes. The new model class is applied to quarterly growth rates of US real GNP.
Keywords: Smooth transition autoregression, multiple regimes, Lagrange, Multiplier test
EC67 STRUCTURAL BREAKS/FILTERING ECONOMIC TIME SERIES
Further Curiosa in 'Spurious' Cointegration
R. O'BRIEN,
Southampton University, U.K.
A previous paper examined the effect, upon Johansen's cointegration tests, of intercept shifts in I(1) processes. It was shown that unrelated random walks can appear cointegrated with probability 1 as the sample size tends to infinity. This paper generalises these results, then investigates the use of prior unit root tests, and of testing for zeros in the cointegrating vectors, to avoid the problem. Monte Carlo experiments suggest the effect remains noticeable in samples of moderate size. An asymptotic analysis of the Engle-Granger procedure for testing for cointegration explains published results which indicate that this procedure has rather different difficulties here.
Keywords: Johansen tests, spurious cointegration
EC67 STRUCTURAL BREAKS/FILTERING ECONOMIC TIME SERIES
Tests for Cointegration in Models with Structural Breaks
K. KRAUZE,
University of Gdansk, Poland
The paper examines tests which allow for the possibility of break in the cointegrated process when the marginal process of one of the variables in the cointegrating relationship has its own break. Cases where the intercept and/or trend coefficient in both processes have a single break are considered. Cointegration tests based on Engle-Granger's procedure and on error correction model are used. Critical values are calculated for the tests by simulation methods. As an illustration we test for structural breaks in the Polish zloty-US dollar exchange rate and the aggregate currency basket index equations using monthly, weekly and daily data.
Keywords: Cointegration, Error correction, Exchange rates, Monte Carlo, Structural breaks, Tests
EC67 STRUCTURAL BREAKS/FILTERING ECONOMIC TIME SERIES
Different Filtering Methods for Finite Nonstationary Time Series
Victor GOMEZ,
Direccion Gral de Analisis y Programacion Presupuestaria, Spain
To estimate the components in an unobserved ARIMA components model, three different approaches can be used: Kalman filtering plus smoothing, Wiener-Kolmogorov filtering and optimal smoothing. It is shown in the paper that the three approaches are equivalent and that many filters, widely used in the electrical engineering literature, can be obtained as the solution of a simple signal extraction problem. Details to design and construct low-pass and band-pass filters are given and it is shown how the three approaches can be used to estimate smooth trends or business cycles within a model-based approach to signal extraction for ARIMA models.
Keywords: Kalman filter, Signal extraction, ARIMA components model, Smoothing, Butterworth filters, Wiener-Kolmogorov filters
EC68 MODEL SELECTION
An Extension Of Risanen's Bound on the Best Empirical Data-Generating Process
Werner PLOBERGER,
University of St Andres, U.K.
Peter Phillips
In a seminal paper, Rissanen (1986) proved an important theorem about minimal information loss in time series regression. Suppose some data is given and the data-generating process (DGP) is known only to belong to a p-dimensional parametric family and satisfies certain regularity conditions. Then, Rissanen showed that the minimum information distance (based on the relative likelihood) between any candidate probability measure and the true measure is, on average, bounded from below by the product of p and the logarithm of the sample size for almost all parameters, i.e., all besides a Lebesgue null set. The bound provides a yardstick for how "close'' to the true probability measure we can get within a parametric family, assuming that the parameters all have to be estimated with the given data. Here we demonstrate a new way of proving results of this type by utilizing general asymptotic expressions for Bayesian densities and working with joint measures over the sample space and parameter space. We are able to modify the Rissanen theorem so that it applies directly to probabilities about the relative likelihood (rather than averages). We also extend the theory to cases that are relevant to nonstationary time series where there may be unit roots and cointegration of unknown orders. The corresponding bound for the minimal information loss is not a constant anymore, but is shown to be proportional to the logarithm of the determinant of the (possibility stochastic) Fisher-information matrix.
EC68 MODEL SELECTION
Modified Stationary Tests with Data Dependent Model Selection Rules
Stephen LEYBOURNE,
University of Nottingham, U.K.
B. McCabe
We deseribe some simlpe methods for improving the performance of stationarity tests (i.e. tests which have a stationnari null and a unit root alternative). Specifically, we inerease the rate of convergence of the test under the unit root alternative from Op(T) to Op(T2), then suggest a method of selecting the order of the AR component in the fitted ARIMA model on which the test is based. Simulation evidence suggests that theses modifications work well. We apply the modified procedure to U.S. monthly macroeconomic data and uncover new evidence of a unit root in unemployement.
Keywords: stationcarity test, ARIMA model, model selection, asymptotic distribution
EC68 MODEL SELECTION
The Distribution of Estimators after Model Selection: Large and Small Sample Results
Benedikt POTSCHER,
Operations Research und Computerverfahren, Austria
Andreas Novak
Frequently, parameter estimation is preceded by a data-driven model selection procedure. In such a case traditional distribution theory based on the assumption of an a priori given model is no longer valid. In Poetscher (1991) the asymptotic distribution of estimators that are preceded by model selection has been studied. In this paper we study the small sample distribution and, in particular, evaluate the accuracy of the approximation provided by the asymptotic distribution in small samples. We also show, how one of the assumptions in Poetscher (1991) can be weakened substantially.
Keywords: Model selection, post-model-selection estimator, AIC
EC69 LABOUR SUPPLY
Labour Supply Behaviour and Demand Side Constraints in the USA and the FRG - A Comparative Analysis
using a Double Hurdle Model for Panel Data
Reinhard HUJER,
Fachbereich fur Statistik und Okonometrie, Germany
Joachim Grammig
We estimate two structural labour supply models based on US and FRG panel data and study differences in the labour supply behavior in the two countries. The paper emphasises the analysis of demand side restrictions of the participation decision in the two economies. The datasets used for estimation are extracted from the GSOEP and the PSID which have been combined with information on sectoral and regional labour market conditions. The econometric contribution of the paper is the specification and estimation of a panel version of the Double-Hurdle model. Specification tests are developed and applied at each estimation step.
Keywords: US a FRG comparison, labour supply, limited dependent variable, panel data
EC69 LABOUR SUPPLY
Testing the Predictive Value of Subjective Labour Supply Data
Bertrand MELENBERG,
Tilburg University, The Netherlands
Rob Euwals, Arthur van Soest
Empirical implementation of labour supply theories is usually based on realized labour market behaviour. This requires strong assumptions about the impact of labour demand. A possibility to avoid these assumptions is to make use of subjective data on desired labour supply. In this paper we investigate whether respondents' answers to survey questions on the desired number of working hours contain additional information on the preferences of the individuals. Using panel data for the Netherlands, we analyze whether deviations between desired hours and actual hours of work help to predict future job changes or changes in hours worked. We use parametric and recently developed nonparametric tests. The results suggest that subjective information on desired working hours are helpful in explaining female labour supply. For males the evidence is mixed.
Keywords: Labour supply, subjective data, (non)parametric testing
EC69 LABOUR SUPPLY
The Hausman-Macurdy Controversy
Matias EKLÖF,
Uppsala University, Sweden
Hans Sacklen
The two perhaps most influential empirical labor supply studies carried out in the U.S. in recent years, Hausman (1981) and MaCurdy, Green & Paarsch (1990), report sharply contradicting labor supply estimates. In this paper we seek to uncover the driving forces behind the seemingly irreconcilable results. Our findings suggest that differences with respect to the estimated income and wage effects can be attributed to the use of differing nonlabor income and wage measures, respectively, in the two studies. Monte Carlo experiments suggest that the wage measure adopted by MaCurdy et al might cause a severely downward biased wage effect such that data falsely refute the basic notion of utility maximization.
Keywords: Labor supply estimation, Slutsky effects
EC70 DEMAND ANALYSIS
The Demand For Food Products: An Analysis of Interpurchase Times and Purchasing Qualities
Michael VISSER,
CREST INRA, France
Christine Boizot, Jean-Marc Robin
Abstract not received from the author
EC70 DEMAND ANALYSIS
Nonparametric Engel Curves and Revealed Preference
Richard BLUNDELL,
UCL, U.K.
Martin Browning, Ian Crawford
Abstract not received from the author
EC70 DEMAND ANALYSIS
The Long Run Demand for Lottery Tickets
Gauthier LANOT,
Keele University, U.K.
Roger Hartley
This paper proposes a theoretical structure to address the question of the individual inter-temporal allocation of lottery play across weeks. We consider infinitely lived risk averse individuals, who extract a thrill from the play. A typical individual seeks to allocate her life time wealth between consumption and lottery expenditure. We consider the case where the discount rate and the interest rate approach unity, which corresponds to a long run average objective function. Using data from the 1994-1995 Family Expenditure Survey (FES) we estimate a simple model of lottery expenditure which allows for unobserved heterogeneity
Keywords: lottery, dynamic programming, penalised maximum likelihood
EC71 EMPIRICAL MACROECONOMICS/AGGREGATION
The Italian Recession of 1993: Aggregate Implications of Microeconomic Evidence
Rafaele MINIACI,
Universita di Padova, Italy
Guglielmo Weber
This paper uses household-level data covering a ten-year period (1984-93) to investigate the likely determinants of the Italian 1993 recession - the first episode after World War II where annual private consumption fell in real terms in Italy. We find that consumption fell most for working-age households and for the self-employed. The evidence we present is consistent with the response to permanent negative shocks due to the major pension reform of 1992 and the introduction of stricter tax compliance measures (the `minimum tax') for the self-employed. This is still true when we control for the role played by the widespread loss of jobs and the collapse of the traditional retail sector that characterised the early 1990s.
Keywords: consumption, pension reform, micro-data
EC71 EMPIRICAL MACROECONOMICS/AGGREGATION
Markups, Business Cycles and Factor Markets: An Empirical Analysis
Oivind NILSEN,
Norwegian Research Centre in Organization and Management, Norway
Jan Erik Askildsen
We investigate the cyclical variation of markups. According to theory, the cyclical variation of markups may go in either direction. If capital market imperfections or labour adjustment costs are present, markups are more likely to vary countercyclically. Contrary, with no such frictions the estimated markups are expectedly procyclical. These hypotheses are tested on plant level data for Norwegian manufacturing industry for the period 1978-1991. The analysis is based on a dynamic model of the firm. Labour may be costly adjustable, and credit rationing may prevail. Product markets are imperfectly competitive. The results indicate a frequent presence of procyclical markups, insignificant adjustment costs and no capital market imperfections.
Keywords: Market power, business cycles, panel data, labour demand, financing
EC72 FACTOR DEMAND/WAGES
Pensions and Productivity: Evidence from Panel Data
Christopher CORNWELL,
University of Georgia, U.S.A.
Stuart Dorsey
This paper presents new empirical evidence on the relationship between pensions and productivity. The framework for our investigation is a production function which includes the firm's pension status as an additional argument. This allows us to test directly whether the presence of a defined benefit pension affects productivity levels. Estimation is carried out on a sample 361 manufacturing firms drawn from the annual Compustat Industrial File covering the period 1981-92. The longitudinal nature of the sample allows us to use to control for unobserved firm characteristics that may be correlated with the decision to offer a pension plan.
Keywords: productivity, pensions, compensation, panel data
EC72 FACTOR DEMAND/WAGES
Worker and Plant Wages: Estimates from a Multi-Level Model
Paul BINGLEY,
CLS Science Park Aarhus, Denmark
Niels Westergard Nielson
Abstract not received from the author
EC73 METHOD OF MOMENT ESTIMATION
Generalization of GMM to a Continuum of Moment Conditions
Marine CARRASCO,
Ohio State University, U.S.A.
Jean-Pierre Florens
This paper proposes a version of the Generalized Method of Moment that handles both the case where the number of moment conditions is finite and the case where there is a continuum of moment conditions. Typically the moment conditions are indexed by a parameter that takes its values in an interval. The program to minimize is then the norm of the moment conditions in a Hilbert space. The estimator is shown to be consistent and asymptotically normal. The optimal estimator is obtained when minimizing the norm of the moment conditions in the reproducing kernel Hilbert space associated to the covariance. We show how to calculate easily this estimator.
Keywords: continuous time, GMM, panel data, reproducing kernel Hilbert space
EC73 METHOD OF MOMENT ESTIMATION
Integrated Conditional Moment Estimation and Testing of Median Regression Models, with an Application to a Mincer-Type Wage Equation
Herman BIERENS,
Pennsylvania State University, U.S.A.
Donna Ginther
In this paper we propose a consistent test of the linearity of the median regression model, similar to the Integrated Conditional Moment (ICM) test of Bierens and Ploberger. This test requires reestimation of the median regression model by minimizing the test statistic of the ICM test to the parameters, yielding a consistent ICM estimator of the parameter vector involved. Also, we may use the ICM test for testing linear parameter restrictions. The ICM estimator and test do not require continuity and moment conditions on the distribution of the dependent variable, or moment conditions on the independent variables. We apply the ICM test to test the functional form of a Mincer-type wage equation.
Keywords: Consistent test, median regression, integrated conditional moment test, wage equation
EC73 METHOD OF MOMENT ESTIMATION
Efficient Conditional Moment Estimation of Nonlinear Sur Models: Theory and Application to Count Data
Joachim INKMANN,
University of Konstanz, Germany
Winifred Pohlmeier
This paper focuses on the estimation of nonlinear equation systems by efficient conditional moment estimators and compares these estimators to pseudo maximum likelihood estimators belonging to the linear exponential family. Using the fact that PML estimators are specific conditional moments estimators with suboptimally chosen instruments we show that efficient conditional moment estimation is asymptotically as efficient as maximum likelihood if the model correctly belongs to the linear exponential family. For systems of count variables we provide Monte Carlo evidence that efficient conditional moment estimation performs well even for moderate sample sizes. However, the relative performance of the efficient conditional moment estimator in small samples compared to two candidates of the PML class depends on the type of regressors and the structure of the equation system. An empirical application to the demand for health care illustrates that this estimation technique is a powerful tool in applied econometric work.
Keywords: seemingly unrelated regressions; count data; conditional moment estimation; optimal instruments; pseudo maximum likelihood; demand for health care
EC74 EUROPEAN MONETARY UNION
The Changing Role of the German Bundesbank. An Empirical Analysis of German Monetary Policy before and after 1983
Katarina JUSELIUS,
European University Institute, Italy
Abstract not received from the author
EC74 EUROPEAN MONETARY UNION
Symmetry and Asymmetry of Supply and Demand Shocks in the European Union: A Dynamic Analysis
Laurence BOONE,
CEPII, France
The empirical literature on Optimal Currency Areas has tested the degree of symmetry between European countries in a purely static fashion. On the other hand, the literature of convergence has focused on nominal and real variables, but has ignored the issue of structural asymmetries in Europe. This paper is an attempt to reconcile the two approaches. It tries to assess the degree of convergence between structural shocks of the European countries. The methodology is that of time-varying parameter estimation: hence, a dynamic measure of the correlation between supply and demand shocks is obtained and we can assess to what extent European economies have converged in structural terms. New evidence, in particular regarding Spain and Italy is provided. Important policy implications are derived, not only in terms of the adjustment costs induced by a monetary union, but also for policy coordination in Europe.
Keywords: Convergence, Kalman Filter, Optimal Currency Area, VAR
EC74 EUROPEAN MONETARY UNION
Exchange Rate Policy in Anticipation of EMU: A Variance Decomposition Analysis for the Drachma/ECU and the Drachma/Dollar Exchange Rates
Athanasios PAPADOPOULOS,
University of Crete, Greece
George Zis
This paper addresses three gaps in the literature of the exchange rate policy for the Greek economy. First, it tests the monetary approach to the exchange rate determination and investigates the sources of the exchange rate fluctuations in Greece. Secondly, the post 1974 experience and the aspects of the exchange rate policy, especially with the use of the most recent data are evaluated. Finally, the methods and estimation techniques applied in the empirical part of the paper, have not been employed for the case of Greece. The estimation techniques applied are integrating the Hendry and Mizon (1993) approach with the standard VAR modelling.Then, a small VAR model is applied with the utilisation of two different sets of macroeconomic data. The results for the two specifications are not analogous. For the Drachma-ECU case, variance decomposition and impulse responses of the model, provided empirical results that are consistent with the theory. For the Drachma-Dollar case the results were not convincing. Variance decomposition analysis displayed that exchange rates variations are largely due to their own innovations.
Keywords: Exchange rate, cointegration, VAR, Greece
EC75 INTRADAY RETURNS IN FINANCIAL MARKETS
Analysis of Order Queues
Gaelle LE FOLLE
Laboratoire de Finance Assurance, CREST, France
Christian Gouriéroux, Bertrand Muller
The bid and ask curves give the demand and supply prices as functions of the instantaneous traded volume. The aim of the paper is the definition, interpretation and intra-day analysis of such bid and ask curves. We provide some illustration from the Paris Bourse quote by quote data.
Keywords: tick by tick data, market microstructure, order book, bid-ask spread
EC75 INTRADAY RETURNS IN FINANCIAL MARKETS
A Bayesian Analysis of Stock Return Volatilty and Trading Volume
Rob BAUER,
Erasmus University Rotterdam, Netherlands
Ronald Mahieu
The relationship between stock return volatility and trading volume is analysed by using the modified mixture model (MMM) framework proposed by Andersen (1996). This theory postulates that price changes and volumes are driven by a common latent infoemation process, which is commonly interpreted as the volatility. Using GMM estimation Andersen finds that the persistence in this latent process falls when a bivariate model of returns and volume, i.e. the MMM, is estimated instead of a univariate model for returns. This empirical finding is inconsistent with the MMM. As oppposed to Andersen's study we apply recently developed simulation techniques based on Markov Chain Monte Carlo (MCMC). A clear advantage of MCMC methods is that estimates of volatility are readily available for use in, for example, dynamic portfolio allocation and option pricing applications. Using Andersen's data for IBM we find that the persistence of volatility remains high in the bivariate case. This suggests that the choice of the estimation technique could be important in testing the validity of the MMM.
Keywords: Modified Mixture Model, MCMC, Stock Mark
EC75 INTRADAY RETURNS IN FINANCIAL MARKETS
An Investigation of Long Range Dependence in Intra-day Foreign Exchange Volatility
Marc HENRY,
LSE, U.K.
Richard Payne
A comprehensive set of estimates of long memory in the volatility of three intra-day foreign exchange data series is presented. Robust semiparametric methods are used. Deseasonalizing procedures are proposed and permit the use of fully parametric methods which provide efficient tests of long memory. The hypothesis of long range dependence in the raw returns is rejected. In the volatility series, however, there is evidence of a long range dependent component, a finding which is significant and consistent across currencies. Furthermore, the hypothesis of integrated volatility is strongly rejected in favour of a covariance stationary alternative, with evidence that previous findings of near-integrated volatility are due to the omission of long-range dependent components.
EC76 SEASONALITY (B)
Likelihood Analysis of Seasonal Cointegration
Soren JOHANSEN,
European University Institute, Italy
Ernst Schaumburg
The vector autoregressive model for seasonal cointegration is analysed. The general error correction model is discussed and conditions are found under which the process is integrated of order 1 at seasonal frequency and exhibits cointegration. Under these conditions a representation theorem for the solution is given expressed in terms of seasonal random walks. Finally the asymptotic properties of the likelihood ratio test for cointegrating rank is given, and it is shown that the estimated cointegrating vectors are asymptotically mixed Gaussian. The results resembles the result for cointegration at zero frequency but expressed in terms of a complex Brownian motion. Tables are provided for asymptotic inference under various assumptions on the deterministic terms.
Keywords: Autoregressive process, Granger's theorem, Error correction model, Complex Brownian motion
EC76 SEASONALITY (B)
Common Cycles in Seasonal Nonstationary Time Series
G. CUBADDA,
University of Rome La Sapienza, Italy
This paper extends the notion of common cycles to quarterly time series having unit roots both at the zero and seasonal frequencies. It is shown that common cycles are present in the Hylleberg-Engle-Granger-Yoo decomposition of these series when their seasonal differences are codependent of order at most three. The effects of seasonal adjustment on codependence are also examined. Inference for common cycles analysis among series which are cointegrated at the zero and seasonal frequencies is derived from existing statistical methods for codependence. Finally, the methodology is applied to study seasonal and non-seasonal comovements between components of consumption in Italy.
Keywords: Codependence, Common Cycles, Seasonal Cointegration, Seasonal Adjustment, Generalized Method of Moments
EC76 SEASONALITY (B)
On the Practical Problems of Computing Seasonal Unit Root Tests
A. TAYLOR,
University of York, U.K.
In this paper we explore the practical problems that can arise in computing the regression based testing procedure for unit roots in quarterly time-series processes. We use the illustrative case study of UK real consumers' expenditure on non-durable goods together with a Monte Carlo study. The evidence presented in this paper suggests that such regression basedtests are highly sensitive to both the form of deterministic components included in the test regression and to the lag structure adopted, the latter of which will tend to be under-fitted when pursuing conventional information based rules such as the it Schwarz BIC criterion, or sequential significance tests on the lagged dependent variables.
Keywords: seasonal unit roots, HEGY tests, model selection
EC76 SEASONALITY (B)
Testing for Integration Using Evolving Trend and Seasonal Models: A Bayesian Approach
Gary KOOP,
University of Edinburgh, U.K.
Herman van Dijk
In this paper, we make use of state space models to investigate the presence of stochastic trends in economic time series. A model is specified where such a trend can enter either in the autoregressive representation or in a separate state equation. Tests based on the former are analogous to Dickey-Fuller tests of unit roots, while the latter are analogous to KPSS tests of trend-stationarity. We use Bayesian methods to survey the properties of the likelihood function in such models and to calculate posterior odds ratios comparing models with and without stochastic trends. In addition, we extend these ideas to the problem of testing for integration at the seasonal frequencies and show how are techniques can be used to carry out Bayesian variants of HEGY test or the Canova-Hansen test.
Keywords: Bayesian, seasonality, unit roots, Gibbs sampler, state space models
EC77 BINARY DATA MODELS
Predictive Performance of the Binary Logit Model in Unbalanced Samples
J.S. CRAMER,
Erasmus University of Rotterdam, The Netherlands
In a binary logit with unequal sample frequencies the less frequent outcome always has lower estimated prediction probabilities than the other. We show why this is so and propose a new measure of fit that is independent of the sample composition. We also consider the effect of unequal sample shares on the detection of outliers.
Keywords: logit, fit, outliers
EC77 BINARY DATA MODELS
Semiparametric Binary Choice Model Estimation with Mismeasured Regressors and Heteroskedastic Errors
Arthur LEWBEL,
Brandeis University, U.S.A.
Consider the standard binary choice model where y is one if a linear latent variable is positive, and zero otherwise. The distribution of the latent variable error e is unknown and may be correlated with the regressors or be conditionally heteroscedastic, e.g., the regressors could be measured with error. A simple root n consistent, asymptotically normal semiparametric estimator of the latent variable coefficients is proposed. A consistent estimator of the conditional error distribution is also provided. The identification assumptions are that e is uncorrelated with some instruments and that one regressor has some special properties.
Keywords: Semiparametric, Binary Choice, Measurement Error
EC77 BINARY DATA MODELS
Binary Choice Panel Data Models with Predetermined Variables
Raquel CARRASCO,
CEMFI, Spain
Manuel Arellano
In this paper we present a class of binary choice models for panel data with the following features: (i) The explanatory variables are predetermined but not strictly exogenous. This includes models with lagged dependent variables as well as models with other forms of unspecified feedback. (ii) Individual effects are allowed to be correlated with the explanatory variables. Dependence is specified through the conditional expectation of the effects given the explanatory variables which is let to be nonparametric. We also present a convenient GMM estimator for these models, which iS consistent and asymptotically normal for fixed T as N goes to infinity. We study the finite sample properties of this estimator in a model with a single lagged dependent variable by means of Monte Carlo simulations. Finally, as an empirical illustration, we estimate a female labour force participation equation with predetermined children and individual effects using PSID data.
EC77 BINARY DATA MODELS
Binary Random Effects Models: Generalised Residuals and Inference
Chris ORME,
The University of Manchester, U.K.
This paper considers some aspects of maximum likelihood estimation and inference in binary data (probit or logit) random effects models. Generalised residual type quantities are identified which define certain aspects of estimation. It is shown that certain algebraic equalities exist which fail to hold numerically in any given application due to the use of Gaussian quadrature (or simulation methods) as an approximation to exact integrals. A simulation method for approximating the asymptotically efficient version of various test statistics in this context is also described and illustrated. The cost of obtaining this simulated version can be much less than that of obtaining the fully efficient version and evidence suggests that it performs just as well.
EC78 TESTING FOR UNIT ROOTS
The Order of the Error Term for Moments of the Log Likelihood Ratio Unit Root Test
in an Autoregressive Process
Rolf LARSSON,
Stockholm University, Sweden
This paper investigates the asymptotics of the log likelihood ratio test for a unit root in an autoregressive (AR) process of general order. The main result is that the expectation and variance (in fact, all moments) of the test statistic may, to the order of T-1, where T is the number of observations, be approximated by the expectation and variance of the corresponding test in an AR(1) process. This result has obvious implications for the asymptotics of unit root tests for panels (cf. Im et al (1995)). An explicit formula for the approximation error of a test in an AR(2) process is also given.
Keywords: Approximation error, Unit root test
EC78 TESTING FOR UNIT ROOTS
LM Tests for Unit Roots in the Presence of Missing Observations
C.R. MCKENZIE,
Osaka University, Japan
Hiro Toda
This paper considers the problem of testing for unit roots in the presence of missing observations using the Lagrange Multiplier (LM) testing principle. LM tests are derived for a unit root in a first order autoregressive process. Modifications of the tests to account for serial correlated errors are considered. The small sample size and power properties of the tests are investigated using a Monte Carlo simulation. An illustrative example using data from the Australian foreign exchange market is also presented.
Keywords: Unit roots, missing observations, Lagrange Multiplier test
EC78 TESTING FOR UNIT ROOTS
A Unit Root Test for Sustainability
Patrick FEVE,
CEPREMAP, France
P.Y. Henin, P. Jolivald
In the context of testing for sustainability, the joint dynamics of debt and surplus are ofently ignored. This paper shows that such practise leads to inapropriate statistical inference, but taking account for this joint dynamics induces large power gains. The limit of the statistics are derived under the null of unit root. Methodological issues are also addressed and these constitute another important focus of the paper. Our results favor the use of a bivariate approach using cross-equation restrictions. They also point out the disvantages in the use of cointegration and unit root with covariates approaches. A simulation study and an emprical application based on public debt sustainability illustrate the potential of this approach.
Keywords: Unit root tests, covariates, parameters uncertainty, size distortions, tests for sustainability
EC78 TESTING FOR UNIT ROOTS
A Completely Consistent Approach for Selecting Among I(1) and I (0) in Levels and in Logs
Valentina CORRADI,
School of Arts and Sciences, U.S.A.
Norman Swanson
In macroeconometrics, unit root tests are typically performed using logs. While this is sensible from a theoretical macroeconomic perspective, there is no clear reason, particularly from an empirical perspective, why logs should be used rather than levels. Further, standard unit root tests assume linearity under both the null and the alternative hypothesis. Violation of this linearity assumption can result in severe size and power distortion, both in finite and large samples. Finally, casual inspection of gnp per capita, for example, plotted against its fitted linear deterministic trend (see Figure 1) gives no clear indication that a loglinear specification should be preferred to a linear specification. Thus, it is reasonable to address the problem of data transformation before running a unit root test. In this paper we propose a simple completely consistent procedure for choosing between levels and log-levels specifications in the presence of deterministic and/or stochastic trends. Once we have chosen the proper data transformation, we remain with the standard problem of choosing between I(1) or I(0), either in levels or in logs. Based on a series of Monte Carlo experiments, we show that the frequency of selecting the correct data transformation is close to one, even for very small sample sizes, when our procedure is implemented. Empirical evidence is also presented, and suggests that 11 of the 14 variables in the Nelson and Plosser (1982) dataset are appropriately modeled in levels rather than in logs, when performing unit root tests.
Keywords: completely consistent procedure, deterministic trend, integratedness, Nelson-Plosser data, nonlinear transformation
EC79 WAGE DETERMINATION
Wages and Asymmetric Information in the Labor Market
Marc VAN AUDENRODE
Université Laval, Canada
Jonathan Leonard
In this paper, we examine both the persistence of displaced worker's wages and their effect on the probability of finding a new job. We use a new database looking at the post-displacement experience of a sample of Belgian workers who lost their jobs because of a sizable reduction in their firm's workforce. We decompose past wages into a market return to human capital, a firm-specific component (the "firm effect"), and an individual component. We develop an information model of wages and test its predictions. These predictions are validated by the evidence on subsequent wages. We find support in favor of an asymmetric information model of labor markets.
Keywords: Wages, Asymmetric information, Firm level data
EC79 WAGE DETERMINATION
Does Wage Drift Offset the Central Settlements? Evidence from the Nordic Countries.
Steinar HOLDEN,
University of Oslo, Norway
The system of wage setting in the Nordic countries is often regarded as highly centralised. However, a large part of the increase in money wages is the result of negotiations at firm level, and this wage drift may offset the outcome of the central agreement. This paper presents evidence from the four major Nordic countries, suggesting that there is no or very little such offsetting effect. Yet the institutional system of wage setting may induce nominal rigidities at the central negotiations, that may prevent wage restraint in times of low inflation and low productivity growth.
Keywords: Wage drift, wages, centralization
EC79 WAGE DETERMINATION
Investigating Rationality in Wage Setting
Sonia BHALOTRA,
University of Bristol, U.K.
This paper performs a direct test of the efficiency wage hypothesis. It also investigates the extent to which deviations from the optimal wage are small enough to be consistent with near-rational behaviour. In this manner, the commonly used statistical metric is complemented with an economic metric. The analysis uses a panel of data from Indian manufacturing. Generalised method of moments estimates are consistent with both the consumption wage and the relative wage having positive efficiency effects. The profit function is, however, fairly flat around the optimum, a deviation of 10% from the efficiency wage resulting in a profit loss of at most 1%. The finding of efficiency wage effects in a labour-surplus economy has interesting policy implications.
Keywords: efficiency wages, near-rationality, panel data, India
EC79 WAGE DETERMINATION
A Simple Test of the Shirking Model
Alan MANNING,
LSE, U.K.
Jonathan Thomas
Traditionally in the literature on local public goods it is assumed that each local public good is a selection from a convex space. In this paper existence of equilibrium is shown for a class of finite models where local public goods are selections from abstract, possibly non-convex, commodity spaces. Consumers are free to migrate between regions. Equilibrium are supported by a system of personalized valuations. It is demonstrated that consumers not only must face a system of personalized valuations in equilibrium but must also, in general, face a different system of personalized valuations out of equilibrium. These equilibria are shown to lie in the core.
EC80 EMPIRICAL GAME THEORETIC MODELS
Testing Game-Theoretic Models of Price Fixing Behaviour
V.A. HAJIVASSILOU,
LSE, U.K.
Abstract not received from the author
EC80 EMPIRICAL GAME THEORETIC MODELS
Estimating Demand for Local Telephone Service with Asymmetric Information and Optional Calling Plans
Eugenio MIRAVETE,
INSEAD, France
This paper studies the theoretical and econometric implications of agents' uncertainty about their future consumption when a monopolist offers them either a unique, mandatory nonlinear tariff, or a choice in advance among a menu of optional nonlinear schedules. In this model, agents' uncertainty is resolved through individual and privately known shocks on their types. In such a situation the principal may screen agents according to their ex-ante or ex-post type, by offering either a menu of optional tariffs or a standard nonlinear schedule. The theoretical implications of the model are used to evaluate the tariff experiment run by South Central Bell in two cities of Kentucky in 1986. The econometric approach explicitly accounts for the existence of asymmetric information between local telephone users and the regulated monopolist. The empirical evidence suggests that there exists a significant asymmetry of information between consumers and the monopolist under both tariff regimes. However, type shocks are not statistically significant, and therefore most of consumers differences are captured through the design of the optional plans instead of using volume discounts. Finally, according to the structural estimates, I conclude that the regulatory constraint was not effective, and that the introduction of optional calling plans was not welfare improving although it increased the local monopolist's expected profits.
EC80 EMPIRICAL GAME THEORETIC MODELS
Hierarchical Dynamic Game Approach Between Stock Markets and An Influential Securities Company: Japanese Stock Market and Nomura Securities Co
Yukio ITO,
Osaka University of Economics, Japan
The purpose of this paper is to investigate applicability and empiricab ility for the Hierarchical (Stackleberg) Dynamic Games between Japanese stock market and Nomura Securities Co. as a securies comapny under the external environment determined by the central (Japanese) government r egulation and the central bank (Bank of Japan)'s interest rate and loa n policies. The empirical invstigation will be shown by policy simulati ons created by the various financial conditions and the targets in the stock market and information of the financial states of Nomura Securite s Co. obtained from the B/S and P/L.
Keywords: hierarchical dynamic game, securities company, simulation
EC81 FINITE SAMPLE THEORY
Conditional Inference in Possibly Unidentified Structural Equations
G. FORCHINI,
University of Southampton, U.K.
G. Hillier
The single structural equation model may involve weak instruments, or, in the extreme, be completely unidentified. These possibilities cannot be ruled out a priori, and this casts doubt on standard inference procedures for such models. This paper argues that inference about structural parameters should be made conditional on identification tests. The consequences of conditioning for the properties of ordinary least squares and two stage least squares estimators are analysed in finite samples for the general case of an equation containing an arbitrary number of endogenous variables. The results obtained clearly support the argument for conditional inference in such models.
Keywords: conditional inference, structural models, unidentified models, precision
EC81 FINITE SAMPLE THEORY
A Missing Dimension in Econometrics
Asatoshi MAESHIRO,
University of Pittsburgh, U.S.A.
Two major inference problems in classical econometrics are the problem of nonspherical disturbances and that of contemporaneous correlation between the disturbances and some elements of the right-hand variables. To solve these problems, a researcher frequently transforms the model, implicitly or explicitly, and estimate the transformed model. Using a model of autocorrelated disturbances and a model of contemporaneous correlation between the disturbances and a right-hand variable, the paper demonstrates that if the transfromation reduces the variation in the values of the transformed variables and/or increases multicollineartiy, the resluting estimator can systematically be less efficient than the OLS estimator when the sample size is small.
Keywords: nonspherical disturbances, contemporaneous correlation
EC81 FINITE SAMPLE THEORY
Necessary and Sufficient Assumptions for Exact Distributional Inference in Linear Regression
Paul BEKKER,
University of Groningen, The Netherlands
The paper provides a unified approach to nonparametric inference in the linear model. A simple equality shows the equivalence of assumptions about the random behavior of disturbances and regressors on the one hand and exact distributional inference about regression coefficients on the other. The result is applied to cases that are closely related to nonparametric tests. Also normal inference based on Student's t is described within this unified framework. The result is related to early work on errors-in-variables models. Finally, the paper briefly discusses nonparametric misspecification tests.
Keywords: Linear model, permutation tests, exact inference
EC81 FINITE SAMPLE THEORY
Markovian Processes, Two-Sided Autoregressions and Exact Inference for Stationary and Stationary and Nonstationary Autoregressive Processes
Olivier TORRES,
Université de Lille 3, France
Jean-Marie Dufour
This paper presents procedures for making finite-sample inference in stationary and nonstationary autoregressive (AR) models. The method is based on special properties of Markov processes and a split-sample technique. In the context of a linear regression model with AR(1) errors, we show how these results can be used to simplify the distributional properties of the model. This leads to a new model which has the form of a two-sided autoregression. We show how to derive tests and confidence sets for the mean and/or autoregressive parameters of the model. We also develop a test on the order of an autoregression. We show that a combination of subsample-based inferences can improve the performance of the procedure.
Keywords: Markov processes, finite-sample inference, split-sample technique
EC82 DURATION MODELS
Stratified Partial Likelihood Estimation
I. TUNALI,
Koc University, Turkey
G. Ridder
When multiple durations are generated by a single unit, they may be related in a way that is not fully captured by a vector of regressors. The omitted variables might vary over the durations and be correlated with the variables in the regression component. We propose a Stratified Partial Likelihood Estimator designed for this situation and develop a specification test for detecting unobserved unit-specific effects. We show how second stage estimates of a common time-dependence function can be formed without compromising the consistency of the regression coefficients and propose a method analysing the unit-specific effects. Keywords: counting process, Cox partial likelihood, stratification
EC82 DURATION MODELS
Combing Micro and Macro Unemployment Duration Data
Bas VAN DER KLAAUW
Erasmus University Rotterdam, The Netherlands
Gerard van den Berg
We combine micro and macro unemployment duration data to study the effects of the business cycle on the outflow from unemployment. We allow the business cycle to affect the exit probabilities of all unemployed workers, and we also allow it to affect the composition of the inflow into unemployment, which may lead to different aggregate exit probabilities as well. We estimate the model using (micro) survey data and (macro) administrative data from France. The inflow composition is estimated along with the other parameters. To deal with differences between the micro and macro unemployment definitions, we develop a Bayesian estimation method with a classical interpretation.
Keywords: unemployment composition, heterogeneity, duration dependence, business cycle, seasons
EC82 DURATION MODELS
Asympotic Expansions of Residual Based Diagnostics in Log-Linear Duration Models
Simon PETERS,
University of Manchester, U.K.
Chris Orme
This paper obtains the O(1/n) correction for asymptotically efficient conditional moment tests, in general. In doing so, the relevant expressions are given in matrix-vector form. These are then applied to arbitrary residual-based test statistics of the type often used in log-linear duration (location-scale) models; e.g., testing for non-normal skewness and/or kurtosis in the normal linear model, or testing for neglected heterogeneity in a Weibull model. It is shown that such test statistics admit a correction which depends upon regression design only through the number of non-constant regressors present in the null model. Monte Carlo evidence is presented on the efficacy of employing such corrections.
EC82 DURATION MODELS
Conducting Inference in Semiparametric Duration Models Under Inequality Restrictions on the Shape of the Hazard Implied by Job Search Theory
Charles ROMEO,
Rutgers University, U.S.A.
In this paper we develop the technical apparatus to efficiently conduct inferences regarding monotonicity in a semiparametric duration model. In process, we advance the state of the art of Bayesian estimation in two respects. First, we develop a systematic approach to adjusting proposal distributions, turning the adjustment process into a nonlinear least squares problem. This greatly improved the numerical efficiency of our importance sampling and the GibbsMetropolis estimation procedures. Second, we developed an efficient approach to conducting inference under linear hypotheses in noulinear models.Our approximate posterior probabilities estimates indicate that the data strongly support the hypothesis of a monotonically decreasing hazard, while the probability that the hazard increases monotonically is estimated to be negligible.
Keywords: Search theory, Bayesian inference, Monte Carlo integration
EC83 FIRM BEHAVIOUR
Non Linear Response of Firm Investment to Q: Testing a Model of Convex Adjustment Costs
Plutarchos SAKELLARIS,
University of Maryland, U.S.A.
Stephen Barnett
Abel and Eberly (1994) study optimal investment behavior in the presence of flow fixed costs, proportional costs, and convex costs. A clear prediction is that investment will alternate between regimes of insensitivity and responsiveness to q separated by unknown threshold levels of q. At the firm level, we find evidence for different regimes of sensitivity to q but not for a regime of zero sensitivity. Our finding that investment has a non-linear relationship to q is important because it implies an elasticity of aggregate investment to q (and fundamentals) that is high and variable over time.
Keywords: Investment,Tobin's Q, Adjustment Costs, Nuisance Parameters
EC83 FIRM BEHAVIOUR
Box-Cox Quantile Regression and the Distribution of Firm Sizes
Jose MACHADO,
Universidade Nova de Lisboa, Portugal
Jose Mata
Using the Box-Cox Quantile Regression model, we estimate the effect of selected industry attributes on the location, scale, skewness and kurtosis of the conditional size distributions of firms in Portuguese manufacturing during the 1980s. We find that a number of industry attributes affect the size of firms in the same direction, but the effects of these variables are much greater at the largest quantiles. We also observe that, over time the distribution shifted towards smaller firms, owing mainly to the way the economy responds to industry characteristics rather than to changes of the level of these characteristics.
Keywords: Box-Cox transformation, Regression Quantiles, Consistent Covariance Estimation, Firm Size Distribution, Small Firms
EC83 FIRM BEHAVIOUR
Firm Exit When Expectations About Present Values are Based on a Vector Autoregressive Prediction Model
A. RAKERUD,
University of Oslo, Norway
Each firm in a panel data set is faced with the problem whether to close down or to continue production for another year. The profit in the case that continuation is chosen is determined by the outcome of a vector autoregressive process with random components. The decision problem of the firm is that of optimal stopping, where the stopping rule depends on the 'option value' of continuing; the firm closes down if the option value falls short of a stochastic treshold. The inference problem consists in estimating the VAR model, estimating the distribution of the random effects, and estimating the stopping rule - the latter being of primary interest.
Keywords: Firm exit, Optimal Stopping, VAR models, Option Value, Random Components, Maximum likelihood
EC83 FIRM BEHAVIOUR
External Economies at the Firm Level: Evidence from Swedish Manufacturing
T. LINDSTRÖM,
Uppsala University, Sweden
An Examination of Swedich manufacturing firm-level data on output and factor inputs from 1979 through 1994 reveals that external economies associated with aggregate manufacturing activity appear to be present. Moreover, internal returns to scale are approximately constant. However, there are other conflicting interpretations, such as procyclical unmeasured factor utilization rates, common productivity shocks, and misspecifications, implying that no definitive conclusions can be made at this stage.
EC84 MODELLING EXCHANGE RATES
Long-Horizon Exchange Rate Predictability
Jeremy BERKOWITZ
Federal Reserve Board, U.S.A.
Lorenzo Giorgianni
Abstract not received from the author
EC84 MODELLING EXCHANGE RATES
Foreign Exchange Mark Efficiency, Uncovered Interest Rate Parity and Monetary Policy Feedback: An Application of the Cointegrated VAR
Jonathan RUBIN,
University of Copenhagen, Denmark
Abstract not received from the author
EC84 MODELLING EXCHANGE RATES
Exchange Rate Target Zones: A New Approach
Feike DROST,
Tilburg University, The Netherlands
Frank de Jong, Bas Werker
We propose a new model for an exchange rate target zone that captures most stylized facts from the existing target zone models while remaining analytically tractable. The model is based on a modified two-limit version of the Cox, Ingersoll, and Ross (1985) model. The exchange rate is kept within the band by decreasing the variance as the exchange rate approaches the band-boundaries. Parity adjustments are also discussed. The models are estimated by conditional GMM. Moreover, prices of currency options are obtained. Throughout, we apply the theory to EMS exchange rate data. We show that, after the EMS crisis of 1993, currencies remain in an implicit target zone which is narrower than the officially announced target zones.
Keywords: target zones, currency option pricing
EC84 MODELLING EXCHANGE RATES
Detecting Undeclared Target Zones within the European Monetary System
Giuseppe CAVALIERE,
University of Copenhagen, Denmark
International agreements, such as the European Monetary System, require the exchange rates to lie inside official target zones. In particular, when the bands are sufficiently large, monetary Authorities could intervene in order to defend an undeclared target zone which could be different from the institutional target zone. In order to evaluate Central Banks interventions, the existence of such undeclared target zones is investigated by developing a statistical procedure based on the analysis of the exchange rate autoregressive structure. An empirical investigation of the European Monetary System shows that it is common for Central Banks to defend undeclared bands.
Keywords: target zones, central banks intervention, reflected Brownian motion
EC85 WELFARE PROGRAMMES
There's No Such Thing as a Free Lunch: Evidence from the Effect of In-Kind Transfers
Ian WALKER,
Keele University, U.K.
Paul Bingley
Several commodities are provided at zero or subsidised prices directly to those who are intended to consume them in an attempt to circumvent agency problems associated with providing a cash transfer to household heads. In addition, some countries provide extensive in-kind transfers for the specific needs of particular client groups such as the elderly, the disabled, and dependent children. This paper is concerned with three UK programmes: a school nutrition programme that provides children from poor households with meals at school at zero cost that ineligible children have the option of buying; free milk to poor households with pre-school children; and a further free milk programme for pre-school children in day-care. We investigate the relationship between household food and milk expenditures, prices, incomes, demographic variables and programme participation. We find evidence that is consistent with in-kind public transfers being offset by the actions of altruistic parents which undermines the case for such transfers.
Keywords: In-kind transfers, programme participation, child welfare
EC85 WELFARE PROGRAMMES
Sickness Duration and Disability Eligibility
Espen BRATBERG,
University of Bergen, Norway
Work, rehabilitation, and disability entrance are alternative destinations in a multistate, multiple episode model for sickness duration, using Norwegian register data. In the observation period the rules for disability benefits eligibility were tightened, thus we can also assess the effect of a "natural experiment". The effect of previous earnings on transitions back to work is statistically significant and positive, but of modest size. Individuals enter disability retirement and rehabilitation with higher rates in second spells of sickness. The results regarding the natural experiment indicate that the duration of all long term sickness spells increased after the reform, suggesting unintended costs.
Key words: sickness duration, multiple episode model, natural experiment, disability pensions
EC85 WELFARE PROGRAMMES
The Effects of Cutbacks in Swedish Sickness Benefits
David EDGERTON,
Lund University, Sweden
Since 1990 sickness benefits have been reduced in Sweden. In this paper we use data from the labour force survey to study the introduction of a qualification day in 1993. Our main centre of interest is the distributional effects caused by this cutback, and to this end we analyse a large number of interactions between background variables and the reform dummy. Considerable attention is thus paid to variable subset selection in binary choice models. Our results show that sick leave incidence is greater for those with job security, but that this difference has decreased after the cutbacks.
Keywords: binary choice models, variable subset selection
EC85 WELFARE PROGRAMMES
Land, Welfare and Efficiency in Rural China
Robin BURGESS,
LSE, U.K.
Abstract not received from the author
EC86 LIMITED DEPENDENT VARIABLE MODELS
Optimal Selection for Direct Mail Using a Discriminant Function Based on Response Amount Modelling
Heik VAN DER SCHEER
University of Groningen, The Netherlands
Pieter Otter
We present several models for profit maximization when households are selected from a mailing list for a direct mail campaign when the response elicited from the campaign can varies over households. The relevant decisions taken by the household are (a) whether to respond, and, in the case of response, (b) the amount of purchase. We show how the selection rule can be improved when the distribution of the decision variable is used as a discriminant function. We empirically evaluate the models using a data set from a charitable foundation, and we show that modeling the decisions by separate structures yields considerably higher profits. Keywords: direct marketing, discriminant analysis
EC86 LIMITED DEPENDENT VARIABLE MODELS
Estimation of a Latent Linear Model Based on the Rank Statistics of the Dependent Variable
Fred JOUNEAU,
CORE, Belgium
L. Broze
In this paper we study a new type of latent model in which only the rank statistics of the dependent variable is observed. This problem appears naturally in the microeconometric literature, in particular in the case of the parametric estimation of a production function when the output is poorly observed. A full information approach seems difficult. So we consider another model which describes part of the information of the first one. This second model is Probit model with serial correlations. The inferential problems (test and estimation) have been studied in the literature but not from this viewpoint. We show by simulations that the proposed estimators behave nicely even in relatively small samples. We also perform the estimation of a production function on a real data set.
EC86 LIMITED DEPENDENT VARIABLE MODELS
Stopped Sum Models for Health Care Demand
Joao SANTOS
ISEG, Portugal
Frank Windmeijer
It is generally accepted that the demand for certain types of health care services depends on decisions of both the individual and the health care provider. In this paper it is argued that the frequently used hurdle models are inadequate to describe this type of demand. Alternatively, it is suggested that under very mild assumptions the conditional expectations both of the number of illness spells and of the demand in each spell can be estimated by the generalized method of moments using only data on the total demand. The results of the paper are illustrated with an empirical application.
Keywords: Two-Part Decision Process, Stopped-Sum Distributions, Health Care Demand, GMM
EC86 LIMITED DEPENDENT VARIABLE MODELS
Outliers, Perfect and Deleting Values in Logistic Binary Choice Models
Pau SALSAS,
Estadistica i Economia Espanyola, Spain
Montserrat Guillen, Ramon Alemany
We develop a procedure for detecting outliers in the context of binary choice models. The basic idea is analogue to the approach used when dealing with nonignorable nonresponse in contingency tables. The concept of perfect value fit and deleting value are defined for logistic models. Iterative and analytic solutions are presented respectively. Our detection diagnostics behave better than traditional statistics for two empirical applications: the analysis of incidents in pre- Challenger and the choice between fixed and adjustable mortgage rates.
Keywords: logit models, outliers, penalised likelihood
EC87 FINANCIAL ECONOMETRICS (B)
Nonlinear Principal Components and Inference on a Conditional Expectation Operator
with Applications to Diffusion Processes
Serge DAROLLES,
CREST, France
Jean-Pierre Florens, Eric Renault
The conditional expectation operator T of any function of given has recently been put forward as a convenient object to build inference procedures about diffusion processes. We argue in this paper that such procedures which are based on the spectral anal ysis of T*T should be viewed as a natural nolinear extension of the principal components analysis (PCA). This analogy is fruitful, not only for the interpretation, but also for the practical implementation of nonlinear PCA as the limit (in the spirit of o rthogonal series methods of estimation) of the standard canonical correlation analysis. Moreover, nonlinear PCA is shown to be also relevant for other classes of applications : nonparametric regression and instrumental variables, conditional heteroskedast icity and postrerior expectation.
Keywords: Diffusion processes, Orthogonal series, Dynamic factor models, Canonical correlation
EC87 FINANCIAL ECONOMETRICS (B)
Beyond the Sample: Extreme Quantile and Probability Estimation with Applications to Financial Data
Casper DE VRIES
Erasmus University of Rotterdam, The Netherlands
Jon Danielsson
Many economic problems are extremal problems like the Value at Risk (VaR) Problem which studies the probability of a large loss. Often the VaR is computed from worst case analysis or by assuming normality. This paper advocates a semi parametric analysis of the tail probabilities with better statistical properties. The main econometric innovation is the consistent estimator of the point where the tail starts. The estimator requires a subsample bootstrap procedure; a full sample bootstrap would fail. This facilitates an asymptotically normal estimate of the VaR probabilities. The alternative VaR methods are compared on real and simulated data.
Keywords: Value-at-Risk, Subsample Bootstrap, Extremes, Semi-parametrics
EC87 FINANCIAL ECONOMETRICS (B)
An Empirical Investigation of the Domestic Pricing Error in an Integrated World
Clemens KOOL,
Universiteit Maastricht, The Netherlands
Kees Koedijk, François Nissen, Peter Schofman
Increasing capital market integration has important implications for the calculation of the cost of capital. In an integrated world, the cost of capital should be determined using the International Capital Asset Pricing Model (ICAPM) rather than the domestic Capital Asset Pricing Model (CAPM). We investigate this issue with an international asset pricing model that explicitly allows for deviations from Purchasing Power Parity (PPP). The pricing error when using the domestic CAPM rather than the ICAPM is zero if diversifiable domestic risk is orthogonal to both the global market portfolio return and foreign currency changes. We use Hansen's (1982) Generalized Method of Moments to test for orthogonality and implement the test for over 3000 individual stocks in 10 different countries. We are unable to reject the hypothesis that the global market portfolio and the foreign currencies affect the cost of capital of an individual firm through the effect of the global market on the risk premium of the local market instead of through the global beta of the firm.
Keywords: asset pricing, cost of capital, purchasing power parity, international integration
EC87 FINANCIAL ECONOMETRICS (B)
Stylized Facts of Daily Return Series and the Hidden Markov Model
Timo TERÄSVIRTA,
Stockholm School of Economics, Sweden
Tobias Rydén, Stefan Åsbrink
Abstract not received from the author