German stock market dynamics

This article estimates generalized ARCH (GARCH) models for German stock market indices returns, using weekly and monthly data, various GARCH specifications and (non)normal error densities, and a variety of diagnostic checks. German stock return series exhibit significant levels of second-order dependence. Our results clearly demonstrate that for both weekly as well as monthly return series the Student-t distribution is superior to the standard normal distribution. In particular, the estimated GARCH-t models appear to be reasonably successful in accounting for both observed leptokurtosis and conditional heteroskedasticity from German stock return movements.     

Economies of scale in public education: an econometric analysis

This article investigates the sources of scale economies in the production of public education. The relationship between the average cost of producing educational output and school characteristics including school and district size is estimated using a neoclassical cost function. The empirical analysis used panel data from Utah school districts and estimates the function using the covariance and error component models after making necessary corrections for heteroskedasticity and autocorrelation. The uncorrected fixed effects model generates a significant negative coefficient on district size in both the cost and expenditure functions; the coefficient on number of students has the hypothesized sign but is not significant in either equation. After making various corrections for autocorrelation and heteroskedasticity, the coefficients have the correct signs and are significant in all equations. Thus, it is concluded that scale economies arise from both sources but that the evidence is stronger for district size.     

Multivariate tests of the capm under heteroskedasticity: A note

This note extends the multivariate testing procedure to the case where heteroskedasticity is present. Previous tests of the CAPM relied on the market model. However, a substantial body of literature indicates that the error term in the market model is heteroskedastic. Failing to correct for heteroskedasticity can lead to biased estimates of the variance-covariance matrix and hence, incorrect statistics for hypothesis testing. In this note, a Wald test with a variance-covariance matrix corrected for heteroskedasticity is derived to test the CAPM. Using monthly data from 1926 to 1994, the adjusted test overwhelmingly rejects the zerobeta version of the CAPM for the 14 subperiods and for the entire sample period.     

Tests for additive heteroskedasticity: Goldfeld and Quandt revisited

A number of new tests for heteroskedasticity have recently become available. Using Monte Carlo methods this paper explores the small sample properties of some of these tests in the context of additive heteroskedasticity. Lagrange multiplier and Wald tests (and variants thereof) are found to be inferior to the likelihood ratio and Goldfeld and QuandtF tests. This is a reconfirmation of the conclusions obtained byGoldfeld/Quandt [1972] in their study of additive heteroskedasticity. The paper also contains some new results onAmemiya's GLS estimator of the additive heteroskedastic structure.     

Job changing decisions of faculty members: a case study of a model with multiple selection

A model of faculty job search behaviour must explain the two decisions of whether to search for another position and of whether to accept the new position if an offer is received. Developing a theoretical model of this process is not diffcult, but estimation difficulties caused by the self-selection inherent in the data subsamples of non-searchers, searchers who stay, and searchers who leave must be considered. To estimate the structural determinants of these decisions we make assumptions regarding the covariance matrix of the error terms of the structural equations that allow us to use a straightforward extension of earlier approaches to this problem. The particular variables that correspond to the theoretical measures in the model are not common to most data bases on faculty. Thus, we were forced to use a rather old data base to estimate the model, which reduces the usefulness of the estimates for current policy purposes. Nonetheless, the estimates do shed light on the efficacy of the model and illustrate the issues involved in the estimation problems.     

Small sample properties of Canonical Cointegrating Regressions

Monte Carlo simulations are performed to examine small sample properties of Canonical Cointegrating Regressions (CCR). The first data generation process is designed to generate both cointegrated and non-cointegrated systems with normal disturbances. If the near-observational equivalence of the stationary and the integrated processes is not significant, both powers and empirical sizes of CCR tests are acceptable. The second data generation process is based on the error correction model. Cointegrated systems with various fat-tailed disturbances are generated and analyzed. The empirical sizes of CCR tests with studentt disturbances and GARCH disturbances are found to be reasonable under certain restrictions. The last data generation process is a generalized least squares (GLS) process that incorporates heteroskedasticity into the error correction model. Again, the empirical sizes of CCR tests are reasonable.     

LIML estimation of import demand and export supply elasticities

I apply limited-information maximum likelihood (LIML) to estimate import demand and export supply elasticities for a range of eurozone countries. The results highlight inconsistencies in the parameters estimated by LIML relative to an estimator that is robust to heteroskedasticity. The nature of the structural equations reveals complications generated by the limiting behavior of the parameters that can be replicated in finite samples. The results of simulations underscore improvements in parameter estimates in a three-dimensional panel, suggesting that the problem of limiting behaviour can be overcome in larger dataset/panels.     

Few and Not So Far Between: A Meta-analysis of Climate Damage Estimates

Given the vast uncertainty surrounding climate impacts, meta-analyses of global climate damage estimates are a key tool for determining the relationship between temperature and climate damages. Due to limited data availability, previous meta-analyses of global climate damages potentially suffered from multiple sources of coefficient and standard error bias: duplicate estimates, omitted variables, measurement error, overreliance on published estimates, dependent errors, and heteroskedasticity. To address and test for these biases, we expand on previous datasets to obtain sufficient degrees of freedom to make the necessary model adjustments, including dropping duplicate estimates and including methodological variables. Estimating the relationship between temperature and climate damages using weighted least squares with cluster-robust standard errors, we find strong evidence that duplicate and omitted variable biases flatten the relationship. However, the magnitude of the bias greatly depends on the treatment of speculative high-temperature (>4 \(^{\circ }\)C) damage estimates. Replacing the DICE-2013R damage function with our preferred estimate of the temperature–damage relationship, we find a three- to four-fold increase in the 2015 SCC relative to DICE, depending on the treatment of productivity. When catastrophic impacts are also factored in, the SCC increases by four- to five-fold.     

Estimating Euler equations

In this paper we consider conditions under which the estimation of a log-linearized Euler equation for consumption yields consistent estimates of the preference parameters. When utility is isoelastic and a sample covering a long time period is available, consistent estimates are obtained from the log-linearized Euler equation when the innovations to the conditional variance of consumption growth are uncorrelated with the instruments typically used in estimation. We perform a Monte Carlo experiment, consisting in solving and simulating a simple life cycle model under uncertainty, and show that in most situations, the estimates obtained from the log-linearized equation are not systematically biased. This is true even when we introduce heteroskedasticity in the process generating income. The only exception is when discount rates are very high (e.g. 47% per year). This problem arises because consumers are nearly always close to the maximum borrowing limit: the estimation bias is unrelated to the linearization and estimates using non-linear GMM are as bad. Across all our situations, estimation using a log-linearized Euler equation does better than nonlinear GMM. Finally, we plot life cycle profiles for the variance of consumption growth, which, except when the discount factor is very high, is remarkably flat. This implies that claims that demographic variables in log-linearized Euler equations capture changes in the variance of consumption growth are unwarranted.     

A panel stationarity test with gradual structural shifts: Re-investigate the international commodity price shocks

This paper proposes a simple panel stationarity test which takes into account structural shifts and cross-section dependency. Structural shifts are modelled as gradual/smooth process with a Fourier approximation. The so-called Fourier panel stationarity test has a standard normal distribution. The Monte Carlo simulations indicate that (i) if the error terms are i.i.d, the test shows good size and power properties even in small samples; and (ii) if the error terms are serially correlated, the test has reasonable size and high power. We re-examine the behavior of the international commodity prices and find out an evidence on the persistence of shocks.     

A pseudo-panel data model of household electricity demand

We study the dynamic behaviour of household electricity consumption on the basis of four large independent surveys conducted in the province of Québec from 1989 to 2002. The latter region displays some rather unique features such as the very extensive use of electricity for space heating in a cold climate and the wide range of energy sources used to meet space heating requirements. We adopt Deaton (1985) approach to create 25 cohorts of households that form a pseudo-panel. The cohorts have on average 131 households. The model error terms allow for group heteroskedasticity and serial correlation. Short-run and long-run own and cross-price elasticities are statistically significant. Electricity and natural gas are estimated to be substitutes while electricity and fuel oil are complements, as it may occur in the Quebec context. The estimate of the income elasticity is not significant. Comparisons with related studies are provided.     

Robust Inference by Sub-sampling

This paper provides a simple technique of carrying out inference robust to serial correlation, heteroskedasticity and spatial correlation on the estimators which follow an asymptotic normal distribution. The idea is based on the fact that the estimates from a larger sample tend to have a smaller variance which can be expressed as a function of the variance of the estimator from smaller subsamples. The major advantage of the technique other than the ease of application and simplicity is its finite sample performance both in terms of the empirical null rejection probability as well as the power of the test. It does not restrict the data in terms of structure in any way and works pretty well for any kind of heteroskedasticity, autocorrelation and spatial correlation in a finite sample. Furthermore, unlike theoretical HAC robust techniques available in the existing literature, it does not require any kernel estimation and hence eliminates the discretion of the analyst to choose a specific kernel and bandwidth. The technique outperforms the Ibragimov and Müller (2010) approach in terms of null rejection probability as well as the local asymptotic power of the test.

     

MODELING INTERNATIONAL CONSUMPTION PATTERNS

This article addresses a number of key problems commonly confronted in the literature on international demand analysis. These include data issues and requirements, multistage budgeting, outliers, group heteroskedasticity, and model selection. A two-stage demand system is fit to International Comparison Programme data for 114 countries for nine aggregate categories and eight food sub-categories of goods. Outliers are identified and omitted from the sample. Parameter estimates for the two stages are obtained with a maximum-likelihood procedure that corrects for group heteroskedasticity. Country-specific income and own-price elasticities are calculated and indicate that poor countries are more responsive to changes in income and prices than rich countries. We also find evidence for the strong version of Engel's law; when income doubles, the budget share of food declines by approximately 0.10.     

Errors in variables and spatial effects in hedonic house price models of ambient air quality

In the valuation of the effect of improved air quality through the estimation of hedonic models of house prices, the potential “errors in variables” aspect of the interpolated air pollution measures is often ignored. In this paper, we assess the extent to which this may affect the resulting empirical estimates for marginal willingness to pay (MWTP), using an extensive sample of over 100,000 individual house sales for 1999 in the South Coast Air Quality Management District of Southern California. We take an explicit spatial econometric perspective and account for spatial dependence and endogeneity using recently developed Spatial 2SLS estimation methods. We also account for both spatial autocorrelation and heteroskedasticity in the error terms, using the Kelejian–Prucha HAC estimator. Our results are consistent across different spatial weights matrices and different kernel functions and suggest that the bias from ignoring the endogeneity in interpolated values may be substantial. This paper is part of a joint research effort with James Murdoch (University of Texas, Dallas) and Mark Thayer (San Diego State University). Their valuable input is gratefully acknowledged. The research was supported in part by NSF Grant BCS-9978058 to the Center for Spatially Integrated Social Science (CSISS), and by NSF/EPA Grant SES-0084213. Earlier versions were presented at the 5th International Workshop on Spatial Econometrics and Statistics, Rome, Italy, May 2006, the 53th North American Meetings of the Regional Science Association International, Toronto, ON, Nov. 2006, the 2007 Meetings of the Allied Social Science Assocations, Chicago, IL, Jan 2007, and at departmental seminars at the University of Illinois. Comments by discussants and participants are greatly appreciated. A special thanks to Harry Kelejian for his detailed and patient clarification of the HAC estimator. The usual disclaimer holds.     

On the median regression for SURE models with applications to 3-generation immigrants data in Sweden

In this paper we generalize the median regression method to be applicable to system of regression equations, in particular SURE models. Giving the existence of proper system wise medians of the residuals from different equations, we apply the weighted median regression with the weights obtained from the covariance matrix of the equations obtained from ordinary SURE method. The benefit of this model in our case is that the SURE estimators utilise the information present in the cross regression (or equations) error correlation and hence more efficient than other estimation methods like the OLS method. The Seemingly Unrelated Median Regression Equations (SUMRE) models produce results that are more robust than the usual SURE or single equations OLS estimation when the distributions of the dependent variables are not normally distributed or the data are associated with outliers. Moreover, the results are also more efficient than is the cases of single equations median regressions when the residuals from the different equations are correlated. A theorem is derived and indicates that even if there is no statistically significant correlation between the equations, using SUMRE model instead of SURE models will not damage the estimation of parameters.     

Aggregation and the Endogeneity Problem

This paper investigates the combined effects of the aggregation and endogeneity problems when there exists a correlation between explanatory variables and the error term in the micro equation. Comparing OLS estimates on the basis of an asymptotic-square-bias criterion, the analysis indicates that aggregate equations are in general subject to less endogeneity due to the synchronization effect. The trade-off between aggregation and endogeneity biases is examined and the conditions under which the aggregate equation is superior to the micro equation are derived in terms of the relative sizes of parameters. [210]     

Poverty Mapping with Aggregate Census Data: What is the Loss in Precision?

Spatially disaggregated maps of the incidence of poverty can be constructed by combining household survey data and census data. In some countries (notably China and India), national statistics agencies are reluctant, for reasons of confidentiality, to release household‐level census data, but they are generally more willing to release aggregated census data, such as village‐ or district‐level means. This paper examines the loss in precision associated with using aggregated census data instead of household‐level data to generate poverty estimates. The authors show analytically that using aggregated census data will result in poverty rates that are biased downward (upward) if the rate is below (above) 50%, and that the bias approaches zero as the poverty rate approaches zero, 50%, and 100%. Using data from Vietnam, it is found that the mean absolute error in estimating district‐level poverty rates is 2.5 percentage points if the census data are aggregated to the enumeration‐area level means, and 3–4 percentage points if the data are aggregated to commune or district level. Finally, the authors propose a method for reducing the error using variances calculated from the census. When this approach is applied to the Vietnam data, this method can cut the size of the aggregation errors by around 75%.     

Estimating VARs with long stationary heterogeneous panels: A comparison of the fixed effect and the mean group estimators

The most promising candidates for estimating vector autoregressive models with long, stationary, possibly heterogeneous panel data sets (panel-VARs) are the fixed effect (FE) and the mean group (MG) estimators. With a view to providing guidance to applied researchers on how to pool in a panel-VAR, this paper compares the performance of the FE and the MG estimators both asymptotically and in Monte Carlo simulations. The main results of the analysis suggest the use of both estimators in applied work. If FE and MG estimates give similar estimates, the FE estimator should be used because it is more efficient. If they differ, the MG estimator should be used only if the panel is sufficiently long — say, twice as long as usually recommended in the dynamic panel data literature. If FE and MG estimates differ and the panel is not long enough, neither is generally a desirable alternative and other estimation techniques are needed.     

Random forests and selected samples

This paper presents a procedure for recovering causal coefficients from selected samples that uses random forests, a popular machine-learning algorithm. This proposed method makes few assumptions regarding the selection equation and the distribution of the error terms. Our Monte Carlo results indicate that our method performs well, even when the selection and outcome equations contain the same variables, as long as the selection equation is nonlinear. The method can also be used when there are many variables in the selection equation. We also compare the results of our procedure with other parametric and semiparametric methods using real data.     

The Chinese Stock Market Does not React to the Japanese Market: Using Intraday Data to Analyse Return and Volatility Spillover Effects

In this paper, we use high‐frequency data to explore the effects of return and volatility spillover during periods in which trading hours in China and Japan overlap. Specifically, we utilize 5‐min returns to estimate fractionally integrated asymmetric power autoregressive conditional heteroskedasticity and fractionally integrated exponential generalized autoregressive conditional heteroskedasticity models, then use the models' standardized residuals to employ a cross‐correlation function approach that tests for the degree to which the Chinese and Japanese markets affect each other. Results indicate a unidirectional influence of the Chinese stock market on Japanese markets in terms of return. This result is likely attributable to restrictions on foreign investment in the Chinese market and the lack of diversified international portfolios among individual Chinese investors.