The Impact of Young Adult Dependent Coverage Expansions: A Debate of Methods

SHADAC Staff

July 08, 2015

Why the Interest?

The original research by Monheit et al. was supported by the State Health Access Reform Evaluation (SHARE) grant program (located a SHADAC), and both the article by Monheit et al. and the article by Burgdorf were published in HSR special/theme issues funded by SHARE.

In the June 2015 issue of Health Services Research (HSR), a set commentaries addressed two previous HSR articles, the first by Alan Monheit, Joel Cantor, Derek DeLia, and Dina Belloff, of Rutgers University Center for State Health Policy (hereafter "Monheit et al."), and the second by James Burgdorf, of RAND Corporation.

In this blog entry we provide an overview of the debate, which has to do with evaluating the impact of state-level (pre-ACA) young adult dependent coverage expansions.

The Original Analysis: Monheit et al.

The original paper in question, “How Have State Policies to Expand Dependent Coverage Affected the Health Insurance Status of Young Adults?” was published in 2011 and presented an analysis that used the Current Population Survey (CPS) Annual Demographic Supplement from 2000-2008 to examine the impact of state policies expanding parental health insurance coverage for young adults.

The authors used a difference-in-differences (DD) model to estimate that state dependent coverage policies in 19 early-adopting states[i] resulted in small increases in health insurance coverage through a parent’s employer-sponsored insurance (ESI) plan. These gains ranged from a 1.52 percentage point increase in parental ESI coverage for young adults aged 19 to 29, to a 3.84 percentage point increase in parental ESI coverage for individuals aged 19 to 25 and residing with their parents. The authors observed that these increases were largely offset by declines in employer-sponsored insurance (ESI) in the young adults’ own name, leading to no significant changes in the overall uninsurance rate for young adults.

The Critique: Burgdorf

In the December 2014 theme issue of Health Services Research, James Burgdorf published an article entitled “Young Adult Dependent Coverage : Were the State Reforms Effective?” In the article, Burgdorf replicated the Monheit et al. study using a comparable DD model and similar data and found a similar increase in health insurance coverage among individuals ages 19 to 29.

Burgdorf’s analysis took a more granular look at the state-level dependent coverage provisions by differentiating between parental and spousal dependent coverage. Using the model in this way, he found that spousal coverage increased by 1.37 percentage points, driving much of the overall dependent coverage increase found among 19 to 29 year-olds. Due to the non-increase in marriage rates in this age group over the time frame, states Burgdorf, the increase in spousal coverage is implausible.

Burgdorf’s model also demonstrates that state dependent coverage expansions predict being unmarried and living with one’s parents, suggesting that the model might be influenced by factors other than the expansions. After performing several falsification tests, Burgdorf concluded that the model used in the Monheit et al. study was flawed and therefore not accurate in measuring the net effect of dependent coverage expansions on health insurance coverage of young adults.

First Commentary: Monheit et al.

The Monheit team provided a response to Burgdorf’s claim. They re-analyzed the data, this time excluding married young adults from the analysis. The team found statistically significant increases in dependent coverage among unmarried young adults ages 19 to 25 living with their parents (+4.19 pp), among all unmarried young adults ages 19 to 25 (+2.48 pp), and among unmarried young adults ages 19 to 29 living with their parents (+2.87 pp). The authors also highlighted recent peer-reviewed work done by Depew (2015) supporting the team’s 2011 findings. Finally, Monheit et al. suggested that the increase in spousal coverage found by Burgdorf might be the result of coding errors in the CPS that did not accurately distinguish between parental and spousal coverage.

Second Commentary: Burgdorf

Burgdorf responded in turn. He argued that a model indicating an increase in young adult dependent coverage that is driven by an implausible increase in spousal coverage is problematic and consequently the results of the model, taken as a whole, do not support the conclusion of the Monheit team’s analysis. Burgdorf added that the existence of other research with similar results does not validate the particular model used by Monheit’s team or the findings generated by the model. Burgdorf does not accept that the increase in spousal coverage is driven by any potential coding errors in the data-- pointing out that significant systematic coding errors would be required to cause the increase in question. Burgdorf concludes that the model should be reconsidered as an accurate tool to estimate changes in dependent coverage.

[i]Early-adoption defined as states that implemented dependent coverage expansions by January 2008. These include Colorado, Delaware, Florida, Idaho, Illinois, Indiana, Maine, Maryland, Massachusetts, Minnesota, Missouri, Montana, New Hampshire, New Jersey, New Mexico, Rhode Island, South Dakota, Texas, Utah, Virginia, and West Virginia.