Gustav Kjellsson

On correcting the concentration index for binary variables.

This article discusses measurement of socioeconomic inequalities in the prevalence of a health condition, in response to the recent exchange between Guido Erreygers and Adam Wagstaff, in which they discuss the merits of their own corrections to the frequently used concentration index. We first reconcile their debate and discuss the value judgments implicit in their indices. Next, we provide a formal definition of the previously undefined value judgment in Wagstaffs correction. Finally, we show empirically that the choice of index matters, as illustrated by comparisons between countries using data from the European Survey of Health, Ageing and Retirement.

Forgetting to remember or remembering to forget: A study of the recall period length in health care survey questions.

Self-reported data on health care use is a key input in a range of studies. However, the length of recall period in self-reported health care questions varies between surveys, and this variation may affect the results of the studies. This study uses a large survey experiment to examine the role of the length of recall periods for the quality of self-reported hospitalization data by comparing registered with self-reported hospitalizations of respondents exposed to recall periods of one, three, six, or twelve months. Our findings have conflicting implications for survey design, as the preferred length of recall period depends on the objective of the analysis. For an aggregated measure of hospitalization, longer recall periods are preferred. For analysis oriented more to the micro-level, shorter recall periods may be considered since the association between individual characteristics (e.g., education) and recall error increases with the length of the recall period.

Lies, Damned Lies, and Health Inequality Measurements: Understanding the Value Judgments

Supplemental Digital Content is available in the text.

A general method for decomposing the causes of socioeconomic inequality in health

We introduce a general decomposition method applicable to all forms of bivariate rank dependent indices of socioeconomic inequality in health, including the concentration index. The technique is based on recentered influence function regression and requires only the application of OLS to a transformed variable with similar interpretation. Our method requires few identifying assumptions to yield valid estimates in most common empirical applications, unlike current methods favoured in the literature. Using the Swedish Twin Registry and a within twin pair fixed effects identification strategy, our new method finds no evidence of a causal effect of education on income-related health inequality.

Lost in Translation: Rethinking the Inequality-Equivalence Criteria for Bounded Health Variables

What change in the distribution of a population’s health preserves the level of inequality? The answer to this analogous question in the context of income inequality lies somewhere between a uniform and a proportional change. These polar positions represent the absolute and relative inequality equivalence criterion (IEC), respectively. A bounded health variable may be presented in terms of both health attainments and shortfalls. As a distributional change cannot simultaneously be proportional to attainments and to shortfalls, relative inequality measures may rank populations differently from the two perspectives. In contrast to the literature that stresses the importance of measuring inequality in attainments and shortfalls consistently using an absolute IEC, this chapter formalizes a new compromise concept for a bounded variable by explicitly considering the two relative IECs, defined with respect to attainments and shortfalls, to represent the polar cases of defensible positions. We use a surplus-sharing approach to provide new insights on commonly used inequality indices by evaluating the underpinning IECs in terms of how infinitesimal surpluses of health must be successively distributed to preserve the level of inequality. We derive a one-parameter IEC that, unlike those implicit in commonly used indices, assigns constant weights to the polar cases independent of the health distribution.

Breaking bad habits by education : smoking dynamics among Swedish women

In a dynamic Two-Part Model (2 PM), we find the effect of previous smoking on the participation decision to be decreasing with education among Swedish women, i.e. more educated are less state dependent. However, we do not find an analogous effect of education on the conditional intensity of consumption.

Measuring Health Inequalities Using the Concentration Index Approach

This article briefly reviews the recent discussion on how to use different versions of the concentration index to measure socioeconomic health inequalities; the appropriateness of the different versions depends on the measurement properties of the health variable and the intended value judgment. Using the European Survey of Health, Ageing and Retirement, the article also empirically illustrates the guidelines for practitioners.

Socioeconomic distribution of GP visits following patient choice reform and differences in reimbursement models: Evidence from Sweden

OBJECTIVE This study aims to analyse changes in the socioeconomic distribution of GP visits following primary care patient choice reform, and to compare their magnitude and direction in pure capitation, versus capitation/activity-based mixed, provider reimbursement settings. METHODS We compute absolute and relative concentration indices using total population registry data from three Swedish counties (N∼3.6 million) two years pre, to two years post, reform. We decompose the indices by the contribution of first, non-recurrent and recurrent visits, and compare their changes in the different provider reimbursement settings. RESULTS In all three counties, the number of visits increased for all population groups. Increases were larger, and distributional changes more pro-poor, in the county with mixed reimbursement. Visit increases were mostly driven by recurrent and, especially, non-recurrent, visits, which were increasingly pro-poor in all counties in absolute, but not in relative, terms. First visits either became decreasingly pro-poor, or did not change significantly. Exclusion of high users removed the pro-poor patterns in the two counties with pure capitation. CONCLUSIONS The reform led to increased access to GP visits, but implied small changes in their socioeconomic distribution. In combination with provider reimbursement models with incentives for higher visit volumes, changes were more pro-poor over time, but it is not clear whether this was at the expense of reduced visit length or content.

Extending decomposition analysis to account for unobserved heterogeneity and persistence in health behavior: Income‐related smoking inequality among Swedish women

This article suggests an enrichment of the standard method for decomposition of the concentration index to account for unobserved heterogeneity and persistence in health behavior. As the underlying regression model in the decomposition, this approach uses a dynamic random-effect probit that both consider individual heterogeneity, using a Mundlak type of specification, and applies a simple solution to account for smoking persistence. I illustrate the suggested approach using a panel of Swedish women in Statistics Swedens Survey of Living Conditions for one vital health-related behavior, smoking. The results highlight the importance of persistence and show that education and living in a single household are the main drivers of income-related smoking inequality.