Dean Hyslop

Identification and Estimation of Dynamic Binary Response Panel Data Models: Empirical Evidence Using Alternative Approaches

We examine the roles of sample initial conditions and unobserved individual effects in consistent estimation of the dynamic binary response panel data model. Different specifications of the model are estimated using female welfare and labor force participation data from the Survey of Income and Program Participation (SIPP). These include alternative random effects models, in which the conditional distributions of both the unobserved heterogeneity and the initial conditions are specified, and fixed effects conditional logit models that make no assumptions on either distribution. There are several findings. First, the hypotheses that the sample initial conditions are either exogenous or in equilibrium are rejected by the data. Misspecification of the initial conditions results in drastically overstated estimates of the state dependence and understated estimates of the short- and long-run effects of children on labor force participation. The fixed effects conditional legit estimates are similar to the estimates from the random effects model that is flexible with respect to both the initial conditions and the correlation between the unobserved heterogeneity and the covariates. Heterogeneity appears to explain about 50% and 70% of the overall persistence in welfare and labor force participation, respectively. In addition, for female labor force participation, there is evidence that fertility choices are correlated with both unobserved heterogeneity and pre-sample participation histories.

The Relationship between Individual Labour Market Outcomes, Household Income and Expenditure, and Inequality and Poverty in New Zealand from 1983 to 2003

This paper investigates the relationship between individual labour market outcomes, household income and expenditure, and inequality and poverty in New Zealand using detailed data from the 1983/84 – 2003/04 Household Economic Survey (HES). We begin by discussing and summarising measures of income and expenditure calculated from the HES. We next examine the relationship between individual labour market outcomes, and household income and expenditure for households with different characteristics. In particular, we focus on how this relationship varies over the life cycle and over time. This analysis is then extended to examine inequality in labour market outcomes, household income and household expenditure. In the last section, we examine the determinants of the levels and changes in poverty in New Zealand using both expenditure and income data. We also apply a newly developed methodology for combining income and expenditure data to produce poverty indicators.

Worker-Firm Heterogeneity and Matching: An Analysis Using Worker and Firm Fixed Effects Estimated from LEED

This paper uses Statistics New Zealand’s Linked Employer-Employee Data (LEED) over the six year period April 1999-March 2005 to derive and analyse estimates of two-way worker and firm fixed effects components of job earnings rates. The fixed effects estimates reflect the portable earnings premium that each worker receives in whichever firm they work for, and the time-invariant premium that each firm pays to all the workers it employs. Our main estimates use full-time equivalent annual earnings for each job-year observation weighted by its effective employment, which involves about 18.7 million job-year observations for 2.8 million employees and 320,000 firms. Our analysis focuses on three issues. First, how much of the variation in job earnings rates is attributable to observable worker demographic factors (age and sex), unobserved worker effects and unobserved firm effects? We find that worker effects account for about one half, worker demographics one quarter, and firm effects 10-25 percent of the variance in job earnings. Second, how much compositional change occurred during this period of substantial employment growth? As measured by changes in the annual averages, worker and firm effects declined by about 5 and 1 percent, respectively, over the period. Third, what is the aggregate pattern of sorting of workers and firms across jobs? The correlation between worker and firm effects is 0.12, which is higher than international estimates and implies a tendency for high-earning workers to work for high-paying firms. A primary dimension along which sorting occurs is the full-time / part-time employment dimension. The results are qualitatively robust to various sensitivity tests, including unweighted estimation across all jobs, using only workers’ main jobs held in each year, jobs of workers estimated to be employed full-time during the year, and excluding jobs in firms that have a low degree of connectivity to other firms. The estimated correlation between worker and firm effects is higher based on unweighted jobs (0.18) and more-connected firms (0.17), but lower based on main job (0.06) and full-time workers (-0.01).

Firm Productivity Growth and Skill

This paper examines the relationship between firm multifactor productivity growth (mfp) and changing skill levels of labour in New Zealand, over the period 2001-12, using longitudinal data from Statistics New Zealands Longitudinal Business Database (LBD) and Integrated Data Infrastructure (IDI). We estimate that the average skill of workers declined by 1.8% over the period, reflecting strong employment growth for workers with lower than average skill levels. The net decline was the combined effect of a 3.6% decline in unobserved skill outweighing a 1.8% increase in observed skill, resulting in 1.8% slower estimated skill-adjusted labour growth (13.3%) than the 15.0% growth in full-time equivalent (FTE) employment. Mirroring the skill-dilution, skill-adjusted mfp growth averaged 0.24% per annum over the period compared to 0.14% pa for unadjusted growth. The patterns were stronger over the pre-GFC period to 2008, during which adjusted and unadjusted mfp grew 0.57% pa and 0.42% pa respectively. Our analysis of the effect of changing skill on mfp growth finds that the impact of skill adjustment was almost entirely due to changing skill composition within continuing firms.

An Analysis of New Zealand’S Business Demography Database

Given that a future linked employer-employee database may use Business Demography data as a backbone, what are the basic qualities of this dataset? What are the observed patterns in job creation and destruction in New Zealand? What are the general patterns in business growth in New Zealand? This analysis uses employer-based data from Statistics New Zealand’s Business Demography Statistics (BDS) database for the period 1994-2001.

Cyclical Earnings Variation and the Composition of Employment

This paper uses Statistics New Zealand’s Linked Employer-Employee Database (LEED) to assess the extent and impact of such changes in the employment composition of workers and firms over this period. LEED provides comprehensive coverage of all wage and salary employment since 1999. It enables longitudinal linking of both workers and firms, and also of the jobs that link them. As the LEED data do not directly measure hours worked or hourly wages, we construct a measure of the full-time equivalent (FTE) annual earnings rate associated with each job observed. Our analysis uses a linear model that regresses log (FTE annual earnings) on worker demographics (sex and age) and aggregate male and female time effects, and also controls for the effects of constant unobserved worker and firm specific factors. We use the estimates from this model, together with the employment transition patterns of workers and firms over the period, to investigate in detail the effects of compositional change on average earnings.

Job Mobility, and Wage Dynamics

Matched employer-employee data research has found that workers f wages are affected by the characteristics of the firms they work in, and that higher skilled workers tend to be employed by higher paying firms. This paper examines the contribution of workers f job mobility to their wage dynamics. We focus on the possible trade-off between moving to a better paying firm and losing a firm-tenure specific component of earnings, and examine what types of workers benefit from changing firms, rather than staying with their existing employer. Our analysis provides four main findings. First, although the raw earnings gains to jobmovers and stayers are about the same, we find that, after controlling for observable differences, job-movers have about 1.3 percent lower annual earnings growth than nonmovers. Second, we estimate that job-movers gain 0.3 percent per year on average from moving to higher paying firms, but lose 1.6 percent in transitory earnings associated with changing jobs. The gains from moving to better firms are larger for both younger and new entrant workers, while the transitory earnings losses are smaller. We interpret these findings as being due to an earnings growth trade-off for workers between moving to a higher paying firm and losing their tenure-related earnings at their existing firm. Third, we estimate that, on average, workers gain (almost) all of the change in firm earnings premiums when they change jobs. However, such gains are not equally shared by all workers. In particular, our estimates suggest that it is the higher ability workers (as measured by the estimated worker earnings premiums) whose earnings gain (or lose) the most from moving to a firm with higher (or lower) earnings premiums. Finally, we find that workers f earnings also benefit on average from a change in the average earnings of their co-workers. Controlling for other factors, we estimate that a 1 standard deviation change in the estimated average peer earnings is associated with about 0.25 percent change in a worker fs earnings on average.

The Specification of Dynamic Discrete-Time Two-State Panel Data Models

This paper examines dynamic binary response and multi-spell duration model approaches to analyzing longitudinal discrete-time binary outcomes. Prototypical dynamic binary response models specify low-order Markovian state dependence and restrict the effects of observed and unobserved heterogeneity on the probability of transitioning into and out of a state to have the same magnitude and opposite signs. In contrast, multi-spell duration models typically allow for state-specific duration dependence, and allow the probability of entry into and exit from a state to vary flexibly. We show that both of these approaches are special cases within a general framework. We compare specific dynamic binary response and multi-spell duration models empirically using a case study of poverty transitions. In this example, both the specification of state dependence and the restrictions on the state-specific transition probabilities imposed by the simpler dynamic binary response models are severely rejected against the more flexible multi-spell duration models. Consistent with recent literature, we conclude that the standard dynamic binary response model is unacceptably restrictive in this context.

Skill Upgrading in New Zealand, 1986–2001

Educational attainment increased markedly in New Zealand between 1986 and 2001, while the income premia for higher qualifications first increased and then stabilised or decreased over the 1990s. We first document the growth in qualification-based skills and then examine its contribution to average income growth and the relationship with relative demand changes. Of the 15 per cent increase in real average incomes between 1986 and 2001, upskilling accounted for 25 per cent, while 70 per cent was due to income growth across all qualifications. The pattern of qualification employment share and relative income changes provides evidence of changing demand for skills within detailed industry–occupation cells.

Effects of results-based financing of maternal and child health services on patient satisfaction in Afghanistan

Objective To examine the impact of a results-based financing programme on patient satisfaction in Afghanistan. Methods We analysed data collected from over 3000 patients from a stratified sample of 112 health facilities (56 results-based financing and 56 non-results-based financing) in 11 out of the 34 provinces of Afghanistan over a three-year period. The 112 facilities were part of 442 primary care facilities that were stratified on facility type and randomly assigned to the results-based financing (intervention) and non-results-based financing (control) groups in the 11 provinces. Data were obtained from the National Health Service Performance Assessment, a baseline survey of patients conducted in 2010, and two follow-up surveys in 2011 and 2012. Measurement of patient satisfaction used a questionnaire comprising 11 questions concerning overall satisfaction, and satisfaction with specific aspects of care, from provider behaviour (e.g. respectfulness, communication) to cleanliness, opening hours and waiting times of facilities. We used a regression adjusted difference-in-differences estimator to investigate the treatment effect on patient satisfaction at facility level. Results There were some improvements in patient satisfaction in results-based financing facilities but these were not statistically significant. In 2011, 2 out of 11 aspects of satisfaction showed 5% improvement: satisfaction with the way health care providers explained the patient’s condition and treatment. In 2012, patient satisfaction was higher for almost all aspects of care in results-based financing facilities, including the two aspects on health care provider respectfulness and facility opening hours, which had been 5% lower in results-based financing facilities than in non-results-based financing facilities in 2011. Overall satisfaction was almost 10% higher in results-based financing facilities than in non-results-based financing facilities in 2012. Conclusion Even if not statistically significant, observed trends suggest some positive impacts of the results-based financing programme on patient satisfaction.