Individual-specific Fixed Effects Research Articles

Moment Conditions for Dynamic Panel Logit Models with Fixed Effects

Abstract This paper investigates the construction of moment conditions in discrete choice panel data with individual specific fixed effects. We describe how to systematically explore the existence of moment conditions that do not depend on the fixed effects, and we demonstrate how to construct them when they exist. Our approach is closely related to the numerical “functional differencing” construction in Bonhomme (2012), but our emphasis is to find explicit analytic expressions for the moment functions. We first explain the construction and give examples of such moment conditions in various models. Then, we focus on the dynamic binary choice logit model and explore the implications of the moment conditions for identification and estimation of the model parameters that are common to all individuals.

Approximate functional differencing

Inference on common parameters in panel data models with individual-specific fixed effects is a classic example of Neyman and Scott’s (Econometrica 36:1–32, 1948) incidental parameter problem (IPP). One solution to this IPP is functional differencing (Bonhomme in Econometrica 80(4):1337–1385, 2012), which works when the number of time periods T is fixed (and may be small), but this solution is not applicable to all panel data models of interest. Another solution, which applies to a larger class of models, is “large-T” bias correction [pioneered by Hahn and Kuersteiner (Econometrica 70(4):1639–1657, 2002) and Hahn and Newey (Econometrica 72(4):1295–1319, 2004)], but this is only guaranteed to work well when T is sufficiently large. This paper provides a unified approach that connects these two seemingly disparate solutions to the IPP. In doing so, we provide an approximate version of functional differencing, that is, an approximate solution to the IPP that is applicable to a large class of panel data models even when T is relatively small.

Excess healthcare costs of psychological distress in young women: Evidence from linked national Medicare claims data.

The prevalence of mental health disorders in young adults is increasing, yet there is limited empirical evidence on its economic consequences. We contribute to the literature by estimating the healthcare costs of psychological distress using panel data of young women (aged 18-23years with a 5-year follow-up) from the Australian Longitudinal Study on Women's Health and linked administrative data from Medicare Australia. Our empirical strategy is based on the classical two-part model of healthcare costs with individual specific fixed-effects. We complement our analysis with a test for selection on unobservables to address potential concerns of endogeneity. We find that young women with psychological distress have 15% higher annual healthcare costs (excluding hospital costs) than women with no psychological distress. A large proportion of these costs is driven by the use of antidepressants and the services of psychiatrists and psychologists. We further find that women with psychological distress have higher out-of-pocket costs on these mental health related services compared to non-mental health specific services. Additionally, we show that the effect of psychological distress on healthcare costs is highest during the first 6months of onset, which gradually decreases afterwards. The findings justify the importance of policy initiatives towards early prevention and treatment of psychological distress, especially among young women.

Income windfalls and overweight: evidence from lottery wins

We examine the impact of an income windfall (from a lottery win) on an individual's overweight and Body Mass Index (BMI). We use longitudinal data from the United Kingdom, where a large proportion of the population plays the lottery, and retrieve income effect estimates using time and individual specific fixed effects alongside a set of relevant controls. Although our findings suggest any income windfall may lead to a contemporaneous increase in overweight, we document evidence that a £1,000 win reduces the probability of overweight in the range by up to 3 percentual points 12 months after the lottery win. Estimates are heterogeneous across working hours and educational attainment. A £1000 lottery win reduces the risk of overweight among low educated individuals by 4.5–5 percentual points (pp)12 months after the lottery win.

Health Outcomes Before, During and After the Great Recession in Canada: A Longitudinal Analysis

Using panel data from Canadian National Population Health Survey (2006-2011), this study examined what happened to individuals’ self-reported health during and in the aftermath of great recession of 2008-2009. The study used different methods such as ordered probit method, ordinary least square method and panel fixed effects method. Once individual specific fixed effects are controlled for, the results show that great recession had no statistically significant detrimental impact on self-reported health status. The results further show that health status in the aftermath of great recession did not significantly differ from the health status before the great recession.

Maximum likelihood estimation of dynamic panel threshold models

Threshold estimation methods are developed for dynamic panels with individual specific fixed effects covering short time periods. Maximum likelihood estimation of the threshold and slope parameters is proposed using first difference transformations. Threshold estimate is shown to be consistent and its asymptotic distribution is nonstandard when the number of individuals grows to infinity for a fixed time period; the slope estimates are consistent and asymptotically normally distributed. The method is applied to a sample of 74 countries and 11 periods of 5-year averages to determine the effect of inflation rate on long-run economic growth.

International Real Estate Review

In this study, we estimate the housing wealth effect of households with different income levels. Since we expect the housing wealth effect to vary based on the different income levels, we use the threshold estimation technique developed in Hansen (1999) instead of imposing an exogenous criterion to divide the sample by income level. This econometric technique is developed for panels with individual-specific fixed effects. Therefore, we apply this econometric method on the findings in the existing literature to estimate the housing wealth effect, while considering the heterogeneity in different income categories. We obtain individual level data from the 2012 to 2016 Korea Household Finance and Welfare Survey (KHFWS) and find statistically significant threshold income levels, thus indicating households show different behaviors based on the threshold income. We provide the groundwork for future research to identify the target group who maximizes their wealth effect, which has housing policy implications.

Good boss, bad boss, workers’ mental health and productivity: Evidence from Japan

It is widely believed that the relationship between a supervisor and his/her employees greatly affects employees’ well-being and/or productivity. However, only a few papers in the economics literature analyze how supervisors influence employees’ well-being and enhance productivity. This paper uses longitudinal data of employees with information about their immediate bosses’ ability, management skills, and characteristics (such as competency, communication skills, and personality traits) to investigate the influence of supervisors on employees. The main findings are as follows. First, even after controlling for individual-specific fixed effects and other job characteristics, such as those proposed in the job strain model, we find that supervisors’ good communication with staff and competency in managerial tasks significantly improve employees’ mental health. Second, we find that good communication between the boss and his/her staff enhances the latter’s productivity and lowers presenteeism. Third, supervisors’ bad communication and low competency increase the probability of quitting. Fourth, good communication partially depends on boss–staff compatibility, which is governed in part by their combined personality traits.

Participation in the Supplemental Nutrition Assistance Program and maternal depressive symptoms: Moderation by program perception

RationalePrevious studies have observed an association between participation in the Supplemental Nutrition Assistance Program (SNAP) and depression, which is contrary to SNAP's potential to alleviate food insecurity and financial strain. ObjectiveThis study investigated the impact of change in SNAP participation status on maternal depression, and whether perceptions of government assistance moderate this association. MethodsData were from the Fragile Families and Child Wellbeing Study (FFCWS). Logistic regression models with individual-specific fixed-effects, were fit to SNAP-eligible mothers who changed SNAP participation and depression status (N = 256) during waves 2 to 4. Perceptions of government assistance were defined as feelings of humiliation or loss of freedom and tested for interactions with SNAP participation. ResultsPerceptions of government assistance moderated the association between SNAP participation and depression (p-interaction = 0.0208). Those with positive perceptions of welfare had 0.27 (95% CI = 0.08 to 0.89) times lower odds of depression when enrolled vs. not enrolled in SNAP. Among those with negative perceptions of welfare, SNAP enrollment was not associated with depression (OR = 1.13; 95% CI = 0.85 to 1.51). ConclusionEvidence suggests that SNAP mental health benefits may be context specific. SNAP's capacity to improve mental health may depend on individual perceptions of government assistance. More research is needed to determine whether interventions aimed at mitigating negative perceptions of programs like SNAP could ameliorate poor mental health among program participants.

Revisiting Cheerful Jane and Miserable John: the impact of income, good health, social contacts and education declines with increasing subjective well-being

ABSTRACTThis short note seeks to replicate the quantile regression analysis in Binder and Coad (2011), but taking into account individual-specific fixed effects (FE; using the British Household Panel Survey data set). It finds declining effects of the four main variables of interest (health, social life, income, education) over the quantiles of the subjective well-being distribution, with attenuated effect sizes for the FE model. Equivalized log income has a negative impact on subjective well-being throughout the distribution. Apart from a number of robustness checks, existing research is extended by looking into the quantile effects of the above variables on a set of domain satisfactions.

The impact of economic downturn on mental health in Canada

Purpose – The purpose of this paper is to use longitudinal Canadian data from the National Population Health Survey (1994-2006) to examine the impact of provincial unemployment rate on mental health as measured by the short form depression scale. Design/methodology/approach – To control for the unobserved individual specific factors, the study utilized individual-specific fixed-effects model. Findings – The study found that, for the overall model, provincial unemployment rate has a significant positive impact on depression. The study further examined the impact of unemployment rate on depression for a number of sub-groups based on gender, age, marital status, and education. The results suggest that the impacts of unemployment rate on depression are heterogeneous across different sub-groups. Practical implications – The results of this study have important policy implications. Previous studies suggest that mental stress may lead to risky health behaviours such excessive drinking, substance use, and smoking. These risky health behaviours may have long term health consequences in terms of chronic conditions such as heart disease, cancer, etc. Thus policy makers may consider taking appropriate steps to provide mental health support during the period of recession. Such support may also be helpful for the unemployed individuals who are too depressed to search for job. Originality/value – Previous studies on this issue may suffer from potential bias since they omitted unobserved individual specific factors from the estimating equations. This paper has taken the opportunity of utilizing longitudinal Canadian Population Health Survey and adopts an individual specific fixed effects method to estimate the effects of macroeconomic conditions on mental health. All of the studies reviewed here used data from the USA. So far no study has examined the impact of unemployment rate on mental health using Canadian data. It is interesting to conduct a study using Canadian data since there are important differences between Canada and the USA with respect to labour market policies and health care systems.

Obesity and happiness: does gender matter?

This study used longitudinal Canadian data from the National Population Health Survey (NPHS), covering a period of 1994-2006, to examine the impact of obesity on happiness. In order to control for unobserved individual-specific fixed effects, this study utilized the fixed effects method. The study found that obesity significantly reduces happiness for females, while it has an insignificant impact on happiness for males. The study further found that the negative impact of obesity on happiness is concentrated among females in prime working age group and also among females with a higher level of education.

Estimation of Nonlinear Dynamic Panel Data Models with Individual Effects

This paper suggests a generalized method of moments (GMM) based estimation for dynamic panel data models with individual specific fixed effects and threshold effects simultaneously. We extend Hansen’s (Hansen, 1999) original setup to models including endogenous regressors, specifically, lagged dependent variables. To address the problem of endogeneity of these nonlinear dynamic panel data models, we prove that the orthogonality conditions proposed by Arellano and Bond (1991) are valid. The threshold and slope parameters are estimated by GMM, and asymptotic distribution of the slope parameters is derived. Finite sample performance of the estimation is investigated through Monte Carlo simulations. It shows that the threshold and slope parameter can be estimated accurately and also the finite sample distribution of slope parameters is well approximated by the asymptotic distribution.

Household Debt and Adult Depressive Symptoms

This study uses data from waves 1 (1987-1989) and 2 (1992-1994) of the National Survey of Families and Households in the United States and a series of standard ordinary least squares (OLS) regressions and OLS regressions with individual-specific fixed effects to estimate associations of particular types and levels of debt with adult depressive symptoms. Results suggest that household debt is positively associated with greater depressive symptoms. However, this association is driven by short-term (unsecured) debt; we find little evidence of associations with depressive symptoms for mid- or long-term debt. The link between short-term debt and depression is generally robust to alternative specifications of our models, including whether debt is defined in absolute or relative terms. Furthermore, these associations are particularly concentrated among 51 to 64 year old adults and those with a high school education or less. These findings suggest that short-term debt may have an adverse influence on psychological wellbeing, particularly for those who are less educated or are approaching retirement age.

Housing wealth effect: Evidence from threshold estimation

This paper attempts to estimate the housing wealth effect of households in different income levels. To endogenously split the sample by income levels, we use the threshold estimation technique, developed in Hansen (1999), for non-dynamic panels with individual-specific fixed effects. The data are drawn from the Panel Study of Income Dynamics (PSID), during the waves of 2001, 2003, and 2005. We find two significant threshold income levels of $74,046, and $501,000. Housing wealth has a significant effect on consumption with a coefficient of 0.010602, if income is below $74,046. It is also significant and equals 0.028224 if income is between $74,046 and $501,000. For incomes above $501,000, the coefficient is not statistically significant.

A Penalty Function Approach to Bias Reduction in Nonlinear Panel Models with Fixed Effects

We consider estimation of nonlinear panel data models with individual specific fixed effects. Estimation of these models is complicated since estimation of the fixed effects when the time dimension is short generally results in inconsistent estimates of all model parameters. We present a penalized objective function that reduces the bias in the resulting point estimates. The penalty function is simple to construct and requires no modification for models with multiple individual specific parameters. We illustrate the approach through a series of simulations that suggest the approach is effective in reducing bias and in an empirical study of insider trading activity.

Diferenciais salariais inter-regionais, interindustriais e efeitos fixos individuais: uma análise a partir de Minas Gerais

Este artigo tem como objetivo identificar se os diferenciais de salário entre os trabalhadores de diferentes regiões do Brasil e de diversos ramos industriais persistem após o controle pelas características não-observáveis destes indivíduos, que são fixas ao longo do tempo. Para atingir este objetivo, usamos a base de dados da Raismigra (MTE) para acompanhar os trabalhadores que estavam formalmente empregados na indústria de transformação de Minas Gerais entre 1999 e 2001, usando a abordagem de dados em painel. Os principais resultados obtidos indicam que os diferenciais de salário entre regiões e entre ramos de atividade persistem após o controle pela heterogeneidade dos trabalhadores, mas sua importância diminui significativamente.

A Penalty Function Approach to Bias Reduction in Non-linear Panel Models with Fixed Effects

In this paper, we consider estimation of nonlinear panel data models that include individual specific fixed effects. Estimation of these models is complicated by the incidental parameters problem; that is, noise in the estimation of the fixed effects when the time dimension is short generally results in inconsistent estimates of the common parameters due to the nonlinearity of the problem. We present a penalty for the objective function that reduces the bias in the resulting point estimates. The penalty function involves only cross-products of scores and the hessian matrix and so is simple to construct in practice. The form of the penalty also provides interesting intuition into how the bias reduction is working. We present simulation results that suggest that the penalized optimization approach may substantially reduce the bias in nonlinear fixed effects models.

Dynamic panel cointegration approaches to the estimation of money demand functions

This paper investigates properties of the money demand functions of Group of Six (G6) member countries (Canada, France, Italy, Japan, U.K., and U.S.) using the estimation and inference techniques of panel cointegration. Empirical analyses are conducted by estimating money demand equations of G6 countries individually and as a whole, allowing heterogeneity in individual specific fixed effects across countries through dynamic, nonstationary panel estimation techniques. By using recently developed panel cointegration techniques, the paper contributes to the literature of money demand studies by improving the power performance of the relevant estimation and inference procedures. It reports fully modified OLS (FMOLS) estimation results of the money demand model for different data frequencies, to find varying signs and magnitudes of real income, interest rates and inflation elasticities of money demand for G6 nations.

The Impact of the Source of Changes in Marginal Tax Rates on Participation in Individual Retirement Accounts

This paper uses a series of two-year panels of tax return data to estimate the effects of two sources of tax rate changes on the participation in Individual Retirement Accounts (IRAs). This paper uses a panel logit approach to control for individual specific fixed effects, which may also influence IRA participation behavior. This paper examines participation during the years of open eligibility for IRAs, as well as examining the impact of the 1986 tax reform on participation. A key finding of this paper is that taxpayers' IRA participation decisions are more sensitive to changes in tax rates due to changes in taxable income than to direct changes in the tax tables.