Instrumental Variable Regression Model Research Articles

Do remittances improve income inequality? An instrumental variable quantile analysis of the Kenyan case

The impact of remittances on income inequality constitutes a keenly debated topic in the development literature. Yet, a consensus still has not evolved on the issue. This paper explores the argument that the adverse distributional impact of remittances obtained by a number of studies could partially be due to the failure to control for existing differentials in the ability to migrate. We test the impact of remittances on household expenditures using data from the Kenyan Migration Household Survey and employing an instrumental variable quantile regression model to control for the unequal access to migration of rich and poor households. Our results indicate that while remittances increase household expenditure at all levels of the expenditure distribution, the impact is unambiguously greatest for poorer households. Hence, remittances, in and of themselves, improve both poverty and the distribution of income. This suggests that if remittances are to provide an impetus to development, recipient economies need to alleviate the credit constraints that restrict access to migration for the poor and the ability to send money home once the access bar has been overcome.

Implementing Corporate Social Responsibility: Empirical Insights on the Impact of the UN Global Compact on Its Business Participants

The implementation of corporate social responsibility (CSR) is crucial for the legitimacy of an organization in today’s globalized economy. This study aims to enrich our knowledge of the implementation of the largest voluntary CSR initiative—the UN Global Compact (UNGC). Drawing on insights from stakeholder, network, and institutional theory, I derive a positive impact of UNGC participation duration on the implementation level of the UNGC principles, despite potential weaknesses in the initiative’s accountability structure. Moreover, I scrutinize the validity of the newly introduced UNGC “Differentiation Programme” before applying this framework in the empirical analysis. Results from ordinal, linear, and instrumental variable regression models suggest that, contrary to claims made by UNGC critics, the duration of UNGC participation does affect the level of UNGC implementation. However, this effect appears to be much smaller than previous practitioner studies have suggested. Moreover, strong local UNGC networks affect the implementation level of the UNGC positively. Their hypothesized moderating role between UNGC participation duration and UNGC implementation level, however, is only significant in networks with activities of high quality rather than high quantity.

Incremental impact of body mass status with modifiable unhealthy lifestyle behaviors on pharmaceutical expenditure.

Overweight/obesity is a growing health risk in Korea. The impact of overweight/obesity on pharmaceutical expenditure can be larger if individuals have multiple risk factors and multiple comorbidities. The current study estimated the combined effects of overweight/obesity and other unhealthy behaviors on pharmaceutical expenditure. An instrumental variable quantile regression model was estimated using Korea Health Panel Study data. The current study extracted data from 3 waves (2009, 2010, and 2011). The final sample included 7148 person-year observations for adults aged 20 years or older. Overweight/obese individuals had higher pharmaceutical expenditure than their non-obese counterparts only at the upper quantiles of the conditional distribution of pharmaceutical expenditure (by 119% at the 90th quantile and 115% at the 95th). The current study found a stronger association at the upper quantiles among men (152%, 144%, and 150% at the 75th, 90th, and 95th quantiles, respectively) than among women (152%, 150%, and 148% at the 75th, 90th, and 95th quantiles, respectively). The association at the upper quantiles was stronger when combined with moderate to heavy drinking and no regular physical check-up, particularly among males. The current study confirms that the association of overweight/obesity with modifiable unhealthy behaviors on pharmaceutical expenditure is larger than with overweight/obesity alone. Assessing the effect of overweight/obesity with lifestyle risk factors can help target groups for public health intervention programs.

The Effects of Medicaid Eligibility on Mental Health Services and Out‐of‐Pocket Spending for Mental Health Services

Millions of low-income Americans will gain health insurance through Medicaid under the Affordable Care Act. This study assesses the impact of previous Medicaid expansions on mental health services utilization and out-of-pocket spending. Secondary data from the 1998-2011 Medical Expenditure Panel Survey Household Component merged with National Health Interview Survey and state Medicaid eligibility rules data. Instrumental variables regression models were used to estimate the impact of expanded Medicaid eligibility on health insurance coverage, mental health services utilization, and out-of-pocket spending for mental health services. Person-year files were constructed including adults ages 21-64 under 300 percent of the Federal Poverty Level. Medicaid expansions significantly increased health insurance coverage and reduced out-of-pocket spending on mental health services for low-income adults. Effects of expanded Medicaid eligibility on out-of-pocket spending were strongest for adults with psychological distress. Expanding Medicaid eligibility did not significantly increase the use of mental health services. Previous Medicaid eligibility expansions did not substantially increase mental health service utilization, but they did reduce out-of-pocket mental health care spending.

Supplement to Tests Based on T-Statistics for IV Regression with Weak Instruments

This paper considers tests of the parameter of an endogenous variable in an instrumental variables regression model. The focus is on one-sided conditional t-tests. Theoretical and numerical work shows that the conditional 2SLS and Fuller t-tests perform well even when instruments are weakly correlated with the endogenous variable. When the population F-statistic is as small as two, their power is reasonably close to the power envelopes for similar and non-similar tests which are invariant to rotation transformations of the instruments. This finding is surprising considering the bad performance of two-sided conditional t-tests found in Andrews, Moreira, and Stock (2007). We show these tests have bad power because the conditional null distributions of t-statistics are asymmetric when instruments are weak. Taking this asymmetry into account, we propose two-sided tests based on t-statistics. These novel tests are approximately unbiased and can perform as well as the conditional likelihood ratio (CLR) test.

Efficient Estimation in Models with Independence Restrictions

Unconditional or conditional independence restrictions are used in many econometric models to identify their parameters. However, there are few results about efficient estimation procedures for finite-dimensional parameters under these independence restrictions. This paper computes the efficiency bound for finite-dimensional parameters under independence restrictions, and proposes an estimator that is consistent, asymptotically normal and which achieves the efficiency bound. The estimator is based on a growing number of zero-covariance conditions that are asymptotically equivalent to the independence restriction. The results are illustrated with examples, including an instrumental variables regression model and partially linear regression models. A small Monte-Carlo study is performed to investigate the estimator’s small sample properties and to quantify the efficiency gains that can be made by using the proposed efficient estimator.

Choosing a Set of Instruments among Many and Possibly Invalid Instruments

We consider the instrument selection problem in instrumental variable regression model when there is a large set of instruments with potential invalidity. Existing methods for instrument selection are based on a priori assumptions of an instrument's validity. Moreover, existing approaches based on consistent moment selection or first-order asymptotics can lead to a large finite sample bias with many instruments. We derive higher-order mean square error (MSE) approximation of a two-stage least squares (2SLS) estimator allowing locally invalid instruments. Based on the approximation to the higher-order MSE, we propose an invalidity-robust instrument selection criterion (IRC) that captures two sources of finite sample bias at the same time: bias from using many instruments and bias from invalid instruments. We show that our criterion is an asymptotically unbiased estimator of the higher-order MSE approximation. In various simulation experiments, our criterion performs well in terms of lower bias and better coverage properties of the estimator compare with other existing selection methods. We also apply our criterion to the empirical application in Card (1995).

Endogenous social capital and self-rated health: Results from Canada’s General Social Survey

In this study we analyze data from Statistics Canada’s General Social Survey, a cross-sectional and nationally representative survey of Canada’s population, to assess the impact of three dimensions of social capital on self-rated health. We measure these dimensions, which consist of social networks and social support, civic participation and social participation, with a comprehensive set of 5 indicators. To avoid reverse causality due to the cross-sectional nature of the data, we employ an instrumental variable simultaneous equations bivariate probit regression model. Our findings indicate that all of the tested dimensions of social capital have a positive and significant impact on self-rated health. These findings suggest that social capital plays an important role in enhancing the health of Canada’s population. Endogeneity was detected in all of our estimations. Consequently, this study also has important methodological implications in that we demonstrate that relying solely on naive estimations leads to biased results.

Robustness of Bootstrap in Instrumental Variable Regression

This paper studies robustness of bootstrap inference methods for instrumental variable (IV) regression models. We consider test statistics for parameter hypotheses based on the IV estimator and generalized method of trimmed moments (GMTM) estimator introduced by Čížek (2008, 2009), and compare the pairs and implied probability bootstrap approximations for these statistics by applying the finite sample breakdown point theory. In particular, we study limiting behaviors of the bootstrap quantiles when the values of outliers diverge to infinity but the sample size is held fixed. The outliers are defined as anomalous observations that can arbitrarily change the value of the statistic of interest. We analyze both just- and overidentified cases and discuss implications of the breakdown point analysis to the size and power properties of bootstrap tests. We conclude that the implied probability bootstrap test using the statistic based on the GMTM estimator shows desirable robustness properties. Simulation studies endorse this conclusion. An empirical example based on Romer's (1993) study on the effect of openness of countries to inflation rates is presented. Several extensions including the analysis for the residual bootstrap are provided.

When Do Dictators Decide to Open Trade Regimes?-Inequality and Trade Openness in Authoritarian Countries

This paper investigates how authoritarian leaders employ trade openness as a response to rising inequality. Based on the Heckscher–Ohlin model of international trade and models of democratic transition, I argue that unskilled laborers in authoritarian regimes can benefit from engaging in international trade, thus becoming more compliant to the authoritarian rules as their countries integrate into the world economy. Therefore, dictators in labor-abundant countries expand trade to neutralize democratization threats initiated by rising inequality. My argument uses supporting data from around eighty authoritarian regimes during the period from 1963 to 2003. I address endogeneity problems with dynamic panel data and instrumental variable regression models in this paper. My analyses suggest that economic globalization helps strengthen authoritarian regimes.

Endogenous social capital and self-rated health: Results from Canada’s General Social Survey

In this study we analyze data from Statistics Canada’s General Social Survey, a cross-sectional and nationally representative survey of Canada’s population, to assess the impact of three dimensions of social capital on self-rated health. We measure these dimensions, which consist of social networks and social support, civic participation and social participation, with a comprehensive set of 5 indicators. To avoid reverse causality due to the cross-sectional nature of the data, we employ an instrumental variable simultaneous equations bivariate probit regression model. Our findings indicate that all of the tested dimensions of social capital have a positive and significant impact on self-rated health. These findings suggest that social capital plays an important role in enhancing the health of Canada’s population. Endogeneity was detected in all of our estimations. Consequently, this study also has important methodological implications in that we demonstrate that relying solely on naive estimations leads to biased results.

Tests based on t-statistics for IV regression with weak instruments

This paper considers tests of the parameter of an endogenous variable in an instrumental variables regression model. The focus is on one-sided conditional t-tests. Theoretical and numerical work shows that the conditional 2SLS and Fuller t-tests perform well even when instruments are weakly correlated with the endogenous variable. When the population F-statistic is as small as two, their power is reasonably close to the power envelopes for similar and non-similar tests which are invariant to rotation transformations of the instruments. This finding is surprising considering the bad performance of two-sided conditional t-tests found in Andrews et al. (2007). We show these tests have bad power because the conditional null distributions of t-statistics are asymmetric when instruments are weak. Taking this asymmetry into account, we propose two-sided tests based on t-statistics. These novel tests are approximately unbiased and can perform as well as the conditional likelihood ratio (CLR) test.

Medicaid Lapses and Low-Income Young Adults’ Receipt of Outpatient Mental Health Care After an Inpatient Stay

This study examined low-income young adults' use of outpatient mental health services after an inpatient mental health stay, with a focus on Medicaid enrollment lapses and public mental health safety-net coverage. The sample included 1,174 young adults ages 18 to 26 who had been discharged from inpatient psychiatric care in a mid-Atlantic state. All were enrolled in Medicaid at the time of discharge, and all were eligible for continued outpatient public mental health services regardless of Medicaid enrollment. Administrative claims data were used to examine outpatient mental health clinic use, psychotropic medication possession, inpatient readmission, and emergency department admission during the 365-day period after the index discharge. The main independent variable was a lapse in Medicaid enrollment. An instrumental-variables regression model was used to minimize estimation bias resulting from unmeasured confounding between lapses and service use. Nearly a third (30%) of the young adults had an enrollment lapse. In instrumental-variables analysis, those whose coverage lapsed were less likely than those who had continuous Medicaid coverage to have at least two clinic visits (38% versus 80%); they also had a lower average psychotropic medication possession ratio (25% versus 55%). Age-related Medicaid enrollment lapses were common in this sample of young adults and were associated with receipt of less clinical care postdischarge despite continued eligibility for public services. States should examine opportunities to assist young adults with serious mental health problems who are aging out of Medicaid enrollment categories for children.

Immigration and the Changing Nature of Homicide in US Cities, 1980–2010

Previous research has neglected to consider whether trends in immigration are related to changes in the nature of homicide. This is important because there is considerable variability in the temporal trends of homicide subtypes disaggregated by circumstance. In the current study, we address this issue by investigating whether within-city changes in immigration are related to temporal variations in rates of overall and circumstance-specific homicide for a sample of large US cities during the period between 1980 and 2010. Fixed-effects negative binomial and two-stage least squares (2SLS) instrumental variable regression models are used to analyze data from 156 large US cities observed during the 1980–2010 period. Findings from the analyses suggest that temporal change in overall homicide and drug homicide rates are significantly related to changes in immigration. Specifically, increases in immigration are associated with declining rates for each of the preceding outcome measures. Moreover, for several of the homicide types, findings suggest that the effects of changes in immigration vary across places, with the largest negative associations appearing in cities that had relatively high initial (i.e., 1970) immigration levels. There is support for the thesis that changes in immigration in recent decades are related to changes in rates of lethal violence. However, it appears that the relationship is contingent and varied, not general.

The effect of trauma center care on pediatric injury mortality in California, 1999 to 2011

Trauma centers (TCs) have been shown to decrease mortality in adults, but this has not been demonstrated at a population level in all children. We hypothesized that seriously injured children would have increased survival in a TC versus nontrauma center (nTC), but there would be no increased benefit from pediatric-designated versus adult TC care. This was a retrospective study of the unmasked California Office of Statewide Health and Planning Department patient discharge database (1999-2011). DRG International Classification of Diseases-9th Rev. (ICD-9) diagnostic codes indicating trauma were identified for children (0-18 years), and injury severity was calculated from ICD-9 codes using validated algorithms. To adjust for hospital case mix, we selected patients with ICD-9 codes that were capable of causing death and which appeared at both TCs and nTCs. Instrumental variable (IV) analysis using differential distance between the child's residence to a TC and to the nearest hospital was applied to further adjust for unobservable differences in TC and nTC populations. Instrumental variable regression models analyzed the association between mortality and TC versus nTC care as well as for pediatric versus adult TC designations, adjusting for demographic and clinical variables. Unadjusted mortality for the entire population of children with nontrivial trauma (n = 445,236) was 1.2%. In the final study population (n = 77,874), mortality was 5.3%, 3.8% in nTCs and 6.1% in TCs. IV regression analysis demonstrated a 0.79 percentage point (95% confidence interval, -0.80 to -0.30; p = 0.044) decrease in mortality for children cared for in TC versus nTC. No decrease in mortality was demonstrated for children cared for in pediatric versus adult TCs. Our IV TC outcome models use improved injury severity and case mix adjustment to demonstrate decreased mortality for seriously injured California children treated in TCs. These results can be used to take evidence-based steps to decrease disparities in pediatric access to, and subsequent outcomes for, trauma care. Therapeutic/care management, level III.

Bayesian Analysis of Instrumental Variable Models: Acceptance-Rejection within Direct Monte Carlo

We discuss Bayesian inferential procedures within the family of instrumental variables regression models and focus on two issues: existence conditions for posterior moments of the parameters of interest under a flat prior and the potential of Direct Monte Carlo (DMC) approaches for efficient evaluation of such possibly highly non-elliptical posteriors. We show that, for the general case of m endogenous variables under a flat prior, posterior moments of order r exist for the coefficients reflecting the endogenous regressors’ effect on the dependent variable, if the number of instruments is greater than m +r, even though there is an issue of local non-identification that causes non-elliptical shapes of the posterior. This stresses the need for efficient Monte Carlo integration methods. We introduce an extension of DMC that incorporates an acceptance-rejection sampling step within DMC. This Acceptance-Rejection within Direct Monte Carlo (ARDMC) method has the attractive property that the generated random drawings are independent, which greatly helps the fast convergence of simulation results, and which facilitates the evaluation of the numerical accuracy. The speed of ARDMC can be easily further improved by making use of parallelized computation using multiple core machines or computer clusters. We note that ARDMC is an analogue to the well-known “Metropolis-Hastings within Gibbs” sampling in the sense that one ‘more difficult’ step is used within an ‘easier’ simulation method. We compare the ARDMC approach with the Gibbs sampler using simulated data and two empirical data sets, involving the settler mortality instrument of Acemoglu et al. (2001) and father's education's instrument used by Hoogerheide et al. (2012a). Even without making use of parallelized computation, an efficiency gain is observed both under strong and weak instruments, where the gain can be enormous in the latter case.

Externalities of openness in innovation

Discussion of open innovation has typically stressed the benefits to the individual enterprise from boundary-spanning linkages and improved internal knowledge sharing. In this paper we explore the potential for wider benefits from openness in innovation and argue that openness may itself generate positive externalities by enabling improved knowledge diffusion. The potential for these (positive) externalities suggests a divergence between the private and social returns to openness and the potential for a sub-optimal level of openness where this is determined purely by firms’ private returns. Our analysis is based on Irish plant-level panel data from manufacturing industry over the period 1994–2008. Based on instrumental variables regression models our results suggest that externalities of openness in innovation are significant and that they are positively associated with firms’ innovation performance. We find that these externality effects are unlikely to work through their effect on the spread of open innovation practices. Instead, they appear to positively influence innovation outputs by either increasing knowledge diffusion or strengthening competition. Our evidence on the significance of externalities from openness in innovation provides a rationale for public policy aimed at promoting open innovation practices among firms.

Invest in the best or compensate the weak?

PurposeThe paper aims to test whether a firm's provision of training depends on the intake quality of trainees. While a firm may just treat each trainee equally, independent of his or her intake quality, firms may alternatively also provide more training to less able individuals or focus on the most able ones. The authors develop a theoretical framework that illustrates under what circumstances a firm chooses a particular training strategy.Design/methodology/approachThe paper uses representative administrative survey data for more than 1,400 Swiss establishments. To test their theoretical predictions about a firm's training strategy, the authors apply multivariate and instrumental variable (IV) regression models. In addition, they use case study evidence from a large Swiss retailer, allowing them to analyze how different instructors in a specific firm react when confronted with apprentices of different intake qualities.FindingsThe authors find that a firm's training strategy depends on a trainee's intake quality and the expected net costs of a particular training occupation. Although firms generally provide less training to less qualified trainees, the authors find that a firm is willing to compensate low‐ability trainees with additional training when training is on average profitable in the short run.Practical implicationsWhen training regulations force firms to follow an investment‐oriented training strategy (net costs in the short run), then low‐ability trainees will not receive additional instruction time and the dropout risk increases.Social implicationsGenerating a regulatory framework that allows firms to achieve a net benefit from work‐based training is crucial for low‐ability trainees to have the opportunity to receive additional training investments that compensate for a lack of competences when training commences.Originality/valueThis paper provides novel and direct empirical evidence on how firms adjust their training provision when faced with heterogeneity in the ability of their trainees.

Causal inference of the effect of adiposity on bone mineral density in adults

The causal effect of adipose tissue on bone mass and the direction of its net influence have not been directly assessed in adult humans. Using the Mendelian randomization analysis, we assessed the causality of adiposity in measurements of bone mass in adult males and females. Subjects consisted of 2154 adults aged 25-54years from a cross-sectional cohort of the employees of the Electricity Generating Authority of Thailand. Body composition was determined after at least 3h of fasting using multifrequency bioelectrical impedance analysis. Bone mineral density (BMD) was assessed by dual energy X-ray absorptiometry. A polymorphism in the fat mass and obesity-associated gene (FTO rs9939609) was used as an instrument in the Mendelian randomization analysis. The genotype distribution of the FTO rs9939609 polymorphism was 61·1% TT, 33·9% AT and 5·0% AA. The average body mass index (BMI), body fat mass and percentage body fat were 23·9kg/m(2) (SD=3·6), 17·9kg (SD=6·6) and 26·8% (SD=7·2), respectively. The FTO rs9939609 polymorphism was significantly correlated with BMI (coefficient=0·673kg/m(2) , P<0·001), body fat mass (coefficient=0·948kg, P<0·001) and percentage body fat (coefficient=0·759%, P<0·01). An instrumental variable (IV) regression model, using BMI as the intermediate phenotype, suggested that FTO was a strong IV. Also, the FTO-BMI polymorphism was significantly associated with total hip and femoral neck BMD but was not correlated with total spine BMD, with estimated correlation coefficients of 0·0189 (95% CI: 0·0046, 0·0332), 0·0149 (95% CI: 0·0030, 0·0268) and 0·0025 (95% CI: -0·0131, 0·0136)g/cm(2) , respectively. The variances of BMDs explained by the FTO-BMI were 19·0%, 21·3% and 1·1%, respectively. Similar trends were also observed for the FTO-body fat mass and FTO-percentage body fat correlations. Mendelian randomization analysis suggests that adiposity might be causally related to BMD at the femur but not at the spine.

Hospital readmission among elderly patients

This study investigates the incidence and determinants of hospital readmissions among elderly patients in Norway. The analyses are based on registered data on inpatient admissions to public hospitals from 1999 to 2006. During this period, mean length of stay in hospital decreased, while readmission rates increased. Probit and instrumental variable regression models are applied for the analyses. The results indicate that longer length of stay in the hospital is associated with lower probability of readmission. A patient's age, comorbidities, and complexity of the treatment procedure are positively associated with readmissions, while higher number of diagnostic procedures negatively affects hospital readmission. Finally, patients discharged to institutions are more likely to be readmitted to the hospital.