"Choose the Plan That's Right for You": Individuation, Risk, and Social Stratification in U.S. Employer-Sponsored Health Insurance

Friday, June 24, 2016: 2:30 PM-4:00 PM
210 South Hall (South Hall)
Adam Goldstein, Harvard University, Cambridge, MA; Harvard University, Cambridge, MA

This study examines the distributional consequences of U.S. employers’ efforts to devolve responsibility for managing their employees’ medical insurance risk. Out-of-pocket (OOP) medical expenditure burden among those with employer-sponsored insurance has grown significantly over the past decade. We ask to what extent does making employees choose from among insurance plans with uneven levels of cost-sharing alter the social stratification of OOP burden? How do differing institutional arrangements at the employer-level differentially distribute burdens among employees? One question of particular interest is whether the devolution of choice results in increased inequality, over and above the effects of employer-mandated increases in cost-sharing?

To answer these questions we analyze data from a large matched employer-employee insurance claims database covering 44 million persons from 2007-2014 (OptumInsight). The Optum database is unique insofar as it contains socio-economic information on enrollees, including income, wealth, education, and race. This permits us to ask novel questions about how social stratification plays out in insurance markets. It also allows us to understand how patterns of household-level financial burden are shaped by employer-level shifts in benefit provision.  

Theoretical Background

The recent diffusion of consumer choice-based architecture in U.S. health insurance reflects broader processes of individuation in American economic life. Weakening buffers between individuals and market forces have left households exposed to ever-greater economic risks. At the same time, transformations in systems of social provision mean that employees and households are expected to take more direct responsibility for managing risks across domains such as retirement planning and medical insurance.  Increasingly, benefit provision is accompanied by demands to “choose the plan that’s right for you.” In the case of medical insurance, this means learning to be one’s own actuary by choosing from among insurance plans with varying risk and cost profiles.

Efforts to devolve the management of risk carry significant implications for the dynamics of social stratification. In particular, they imply that the processes by which risks and costs are distributed are functions not only of broad categorical criteria (e.g. citizenship, employment, age, etc.), but also increasingly mediated through individuals’ own choices and actuarial abilities. Offering employees the choice of a high-deductible plan is touted as a means to reduce costs while matching consumers to plans that better meet their needs. However, the prospect of lower monthly premiums can lure individuals into plans that carry greater economic risks, often with severe financial consequences. Moreover, individuals in different social positions have uneven levels of resources to draw on when navigating complex decisions such as choosing insurance plans. This implies that individuation may exacerbate social inequality in expenditure burden.


Empirically, the analysis examines both distributional and social inequality. By distributional inequality, we mean the total variance in the distribution of OOP among enrollees. By social inequality, we mean the degree to which OOP burden is unequally distributed across socio-economic status indicators such as educational attainment, occupation, wealth, and race/ethnicity. 

The Optum data includes all persons enrolled in health plans through a very large U.S. commercial insurer from 2007-2014. This amounts to 44.5 million cumulative person records, with 18-19 million person records per year, and average duration of 2.7 years at the individual level. The insurer’s large size, geographic scope, and diversification across business segments make the Optum sample closely representative of the commercially insured U.S. population.

Distributional inequality is measured using standard measures of dispersion within firms. Social inequality is conceived as the conditional association between SES measures and OOP burden. The primary employer-level variable of interest is an indicator for whether the firm offers traditional plans exclusively, a high-deductible plan exclusively, or a choice of plans that include both traditional and HDHP options. We are not interested in the sheer number of choices. Rather, we are interested in whether employees face a choice between plans with uneven levels of cost-sharing. 

Empirical Tests

1) The first set of analyses focus on within-firm effects. Our analytical approach leverages the longitudinal structure of the data. When firms which initially offer a low deductible plan switch to offering a choice between plans, what is the effect on distributional inequality and social disparities among employees?

 Hypothesis 1: Firm transitions to choice will be associated with increased variance in the OOP burden among employees within firms.

 Hypothesis 2: Firm transitions to choice will have a more deleterious effect on the OOP burdens of lower-SES enrollees than higher-SES enrollees.

2) To what extent does devolved choice increase social inequality, over and above the effects of employer-mandated increases in cost-sharing? To answer this question we utilize a difference-in-difference matching strategy. This approach combines a difference-in-difference model with employer-level matching to adjust for pretreatment observables. 

Hypothesis 3: Firms transitioning to choice will exhibit greater relative growth in social disparities in OOP burden compared to firms transitioning to HDHPs.

3)    Mechanisms and Follow-up Tests -- Depending on the results above, several additional questions follow.

  1. To what extent do social disparities account for observed change in distributional inequality? In other words, what portion of change in the total OOP variance within firms is attributable to change in between-group variance? Variance decomposition techniques will allow us to quantify the relative salience of sociologically-defined axes of stratification.
  2. If the devolution of choice increases social inequality, to what extent is the individual selection process the operative mechanism? To get traction on this we will estimate individual-level models: Conditional on having selected into an HDHP, do members of certain groups tend to see greater growth in OOP compared to the previous year? In other words, does the lure of the lower-premium higher-risk plan end up burdening some social groups disproportionately? 
  3. If devolution of choice results in greater social disparities, to what extent is this attenuated after controlling for health/prior year utilization? In other words, does choice heighten inequality primarily by shifting OOP burden onto the less healthy? Or are there residual SES differences?
  4. If social inequality does not grow, is this accounted for by decreased utilization? To answer this question, we will simulate OOP under the counterfactual scenario where healthcare utilization was held constant at the previous year’s level.