Job Lock: Evidence from a Regression Discontinuity Design

• Author: Susan Gates,Robert W. Fairlie,Kanika Kapur
• Date: 01 January 2016
• DOI: http://doi.org/10.1111/irel.12127
• Published date: 01 January 2016

Job Lock: Evidence from a Regression

Discontinuity Design*

ROBERT W. FAIRLIE, KANIKA KAPUR, and SUSAN GATES

Employer-provided health insurance may restrict job mobility, resulting in “job

lock.”Previous research on job lock finds mixed results using several methodolo-

gies. We take a new approach to examine job lock by exploiting the discontinuity

created at age 65 through the qualification for Medicare. Using a novel procedure

for identifying age in months from matched monthly Current Population Survey

data and a relatively unexplored administration measure of job mobility, we com-

pare job mobility among male workers in the months just prior to turning age 65

to job mobility in the months just after turning age 65. We find no evidence that

job mobility increases at the age 65 threshold when Medicare eligibility starts.

We also do not find evidence that other factors such as retirement, reduction in

hours worked, Social Security eligibility, pension eligibility, and sample changes

confound the results on job mobility in the month individuals turn 65.

Introduction

The predominant source of health insurance in the United States for work-

ing-age adults is employer-provided health insurance. More than three quarters

of full-time workers between the ages of 18 and 64 have health insurance

through employers. In comparison, only 6 percent have health insurance pur-

chased through the individual market and another 6 percent have public insur-

ance (Fronstin 2010a). A potential cost of this reliance on employer-provided

health insurance is the nonportability of insurance across employers, which

potentially results in “job lock.”That is, workers may be reluctant to leave their

jobs to search for new ones when otherwise optimal because of the possible

*The authors’affiliations are, respectively, University of California, Santa Cruz, Santa Cruz, California

and RAND, Santa Monica, California. Email. rfairlie@ucsc.edu; University College Dublin, Dublin, Ireland.

Email: kanika.kapur@ucd.ie; and RAND, Santa Monica, California. Email: sgates@rand.org.

JEL: J60, I13.

This research was supported by the Kauffman-RAND Institute for Entrepreneurship Public Policy through

a grant from the Ewing Marion Kauffman Foundation. The authors thank David Card, Jonathan Leonard,

Ken Couch, Elise Atkins, Paul Devereux, Alex Resch, as well as seminar participants at UC Berkeley,

UCSC, and the ASHE Meetings, for useful comments and suggestions.

INDUSTRIAL RELATIONS, Vol. 55, No. 1 (January 2016). ©2015 Regents of the University of California

Published by Wiley Periodicals, Inc., 350 Main Street, Malden, MA 02148, USA, and 9600 Garsington

Road, Oxford, OX4 2DQ, UK.

92

loss of coverage due to pre-existing condition exclusions, waiting periods on

new jobs, loss of particular insurance plans, and disruption in the continuity of

care with their healthcare providers.

1

Changing jobs during a health-plan year

may also result in the loss of any earned credit toward plan deductibles and

unused balances in health reimbursement accounts. The heavily debated Afford-

able Care Act (ACA) of 2010 attempts to increase coverage and portability of

health insurance with key provisions that went into effect in 2014.

Although the theoretical justifications for the existence of job lock are

unambiguous, there is no consensus in the large and growing empirical litera-

ture on the size of job lock (Gruber and Madrian 2004). A number of studies

have observed large and statistically significant estimates of job lock, often

finding that health insurance reduces job mobility by 25 to 50 percent (Adams

2004; Anderson 1997; Bansak and Raphael 2008; Boyle and Lahey 2010;

Bradley et al. 2007; Bradley, Neumark, and Barkowski 2012; Cooper and

Monheit 1993; Gruber and Madrian 1994; Madrian 1994; Rashad and Sarpong

2008; Stroupe, Kinney, and Kniesner 2001, Tunceli et al. 2009;). Other studies

have found no effects or smaller effects that in some cases attain statistical sig-

nificance for specific demographic groups (Berger, Black, and Scott, 2004;

Dey and Flinn 2005; Gilleskie and Lutz 2002; Hamersma and Kim 2009;

Holtz-Eakin 1994; Kapur 1998; Penrod 1994, Slade 1997; Spaulding 1997).

The existing literature generally employs one of three main identification

strategies to estimate job lock. The most common approach taken in the previ-

ous literature is to use a difference-in-difference strategy that compares the

probability of job mobility of otherwise similar employees who differ only in

how much they value their current employer’s health insurance policy.

2

Examples of measures of the demand for health insurance coverage used in

the literature include pregnancy of spouse, family size, family health status,

and availability of alternative health insurance coverage.

3

1

Federal regulations (such as the Health Insurance Portability and Accountability Act [HIPAA] and the

Consolidated Omnibus Budget Reconciliation Act [COBRA]) aim to reduce these problems; however, indi-

viduals still have to either pay for potentially expensive continuation coverage if they terminate employment

or face new pre-existing condition exclusions and waiting periods if they are uninsured for more than 63

days between jobs. COBRA is expensive, with premiums averaging $4824 for employee-only coverage and

$13,375 for family coverage in 2009, and this expense may be responsible for low take-up rates of less than

10 percent (Fronstin 2010b). Take-up rates increased when legislation allowed for a temporary COBRA

subsidy from 2009 to 2010; however, this subsidy has now expired (Bovbjerg et al. 2009).

2

See Madrian (1994); Kapur (1998); Bradley et al. (2007); and Bradley, Neumark, and Barkowski

(2012) for a few examples, and Gruber and Madrian (2004) for a review of the literature.

3

In their review of the literature, Gruber and Madrian (2004) discussed some concerns with this

approach taken in many previous studies. A common measure of alternative health insurance coverage is

spousal coverage, but this may be endogenous to the job mobility decision (Royalty and Abraham 2006).

Other measures used to proxy health insurance demand are family health status and expected medical utiliza-

tion. These may have relatively low power and/or be difficult to measure.

Job Lock: A Regression Continuity Design /93