The effect of automatic enrolment on debt

John Beshears, James Choi, David Laibson, Brigitte C. Madrian, William Skimmyhorn 17 August 2019

a

A

Automatically enrolling people into defined contribution pension plans has become increasingly common. The Plan Sponsor Council of America (2018) reports that 60% of the US 401(k) plans in its survey automatically enrol employees. The UK, New Zealand, and Turkey have national pension schemes that mandate automatic enrolment. Automatic enrolment has spread rapidly because relative to enrolment systems that require participants to opt in to contributing, the opt-out mechanism of automatic enrolment increases both the fraction of employees who contribute to the savings plan and (if the default contribution rate is set high enough) the average contribution rate to the plan (Madrian and Shea 2001, Choi et al. 2002 and 2004, Beshears et al. 2008).

But does automatic enrolment increase economic security in retirement? The assumption among its advocates has been that the incremental contributions it induces are financed mostly by decreased consumption (Thaler 1994, Beshears et al. 2006). But if incremental contributions are funded by slower growth in the balances of other asset accounts or faster growth in debt balances, that would at least partially undo the intended benefit of automatic enrolment.

A natural experiment to study debt effects

In Beshears et al. (2019), we study how automatic enrolment affects the debt side of the household balance sheet by linking payroll records to individual credit reports. We utilise a natural experiment that saw automatic enrolment introduced at a 3% of income default contribution rate for newly hired civilian employees of the US Army on 1 August 2010. Importantly, employees hired prior to 1 August 2010 were never subject to automatic enrolment.

We identify the effect of automatic enrolment by comparing savings and debt outcomes for the 32,072 employees hired in the year prior to the adoption of automatic enrolment to savings and debt outcomes for the 26,802 employees hired in the year after, while controlling for shocks that affect everyone equally within a given period. Gelman et al. (forthcoming) find that on the day before payday, the median US federal government employee in their sample has liquid assets (checking plus savings account balances) that can cover only five days of spending. Our new-hire sample’s average income of about $56,000 is much lower than Gelman et al.’s sample average income of $89,804, so our study sample is unlikely to have more liquidity (Kaplan et al. 2014). If automatic enrolment does not reduce consumption expenditures, the additional retirement plan contributions it induces are likely funded by borrowing rather than decumulation of liquid assets.

Findings from the natural experiment

We first confirm that automatic enrolment increases contributions to the retirement savings plan. On average, at 43-48 months after hire, automatic enrolment at a 3% default contribution rate increases cumulative employee contributions plus employer matching contributions (since hire) by 4.1% of first-year annualised salary.

Our main results are for credit scores and debt balances excluding auto debt and first mortgages. At 43-48 months after hire, automatic enrolment’s effect on credit scores is a minuscule 0.001 standard deviation increase, with a 95% confidence interval of [-0.02 standard deviations, 0.03 standard deviations] that is extremely tight around zero. Derogatory debt balances that have been passed to an external collection agency decrease insignificantly by 0.1% of first-year salary, with a 95% confidence interval of [-0.3%, 0.1%] – a further indication that automatic enrolment does not increase financial distress. Total debt balances excluding auto and first mortgage debt fall by 0.6% of first-year salary, with a 95% confidence interval of [–2.4%, 1.2%]. In sum, our evidence does not support the hypothesis that automatic enrolment increases financial distress and costly borrowing. 

Figure 1 Effect of automatic enrolment on Vantage credit score

Notes: Point estimates and 95% confidence intervals are shown. To interpret the magnitudes, note that the standard deviation of Vantage credit scores is 95.

Figure 2 Effect of automatic enrolment on debt balances excluding first mortgages and auto loans divided by annualised first-year pay

Note: Point estimates and 95% confidence intervals are shown.

 Our results on auto loans and first mortgages are less conclusive for two reasons. 

  • First, because these types of debt are usually taken on to finance the acquisition of an asset, any increases in their balances have ambiguous implications for net worth – assets typically increase along with liabilities. 
  • Second, because of the high variance of first mortgage balances, we estimate effects on first mortgages with little statistical precision despite our relatively large sample size. We find no significant increase in either kind of debt balance in our main regression specification, although the point estimates are positive. At 43-48 months of tenure, the point estimate of the auto debt balance effect is 1.1% of income (95% confidence interval = [–0.1%, 2.3%]), and the point estimate of the first mortgage balance effect is 2.2% of income (95% confidence interval = [–5.1%, 9.5%]). However, we note that the auto and first mortgage debt effects are positive and statistically significant in some alternative (non-benchmark) statistical specifications.

Figure 3 Effect of automatic enrolment on auto loan balances divided by annualised first-year pay

 

Note: Point estimates and 95% confidence intervals are shown.

Figure 4 Effect of automatic enrolment on first mortgage balances normalised by annualised first-year pay

Note: Point estimates and 95% confidence intervals are shown.

Employees whose contributions are increased by automatic enrolment typically have few financial assets outside the plan. If automatic enrolment does increase auto and first mortgage debt, the increases are likely due to increases in the value of the assets acquired rather than a previously induced spend-down of non-retirement financial assets. Therefore, the short-run effects on net worth of increases in these types of debt will be muted. (To be certain of this, we would need data on non-retirement-plan assets).

Conclusion

We provide strong evidence against the hypothesis that automatic enrolment increases financial distress and debt, excluding auto loans and first mortgages. In our benchmark analyses, we also do not find statistically significant evidence that automatic enrolment increases auto debt or first mortgages. However, we sometimes find positive and significant effects on these latter categories of debt in alternative specifications, and we estimate the first mortgage effect with little precision. 

Future research should consider how automatic enrolment contemporaneously affects non-retirement assets, so that we have a comprehensive picture of effects on the total household balance sheet. More research remains to be done on other margins of crowd-out as well. A recent paper (Choukhmane 2019) finds that automatically enrolled workers in the UK contribute less than non-automatically enrolled workers to their next employer’s pension plan if that subsequent employer does not use automatic enrolment. This crowd-out effect is absent if the next employer automatically enrols its employees. In ongoing work, we are investigating how automatic enrolment affects withdrawals from the savings plan before retirement.

References

Beshears, J, J J Choi, D Laibson, and B C Madrian (2006), “Retirement Saving: Helping Employees Help Themselves”, Milken Institute Review (September): 30-39.

Beshears, J, J J Choi, D Laibson, and B C Madrian (2008), “The Importance of Default Options for Retirement Saving Outcomes: Evidence from the United States”, in S J Kay and S Tapen (eds), Lessons from Pension Reform in the Americas, Oxford: Oxford University Press: 59-87.

Beshears, J, J J Choi, D Laibson, B C Madrian, and W L Skimmyhorn (2019), “Borrowing to Save? The Impact of Automatic Enrollment on Debt”, NBER Working Paper 25876.

Choi, J J, D Laibson, B C Madrian, and A Metrick (2002), “Defined Contribution Pensions: Plan Rules, Participant Decisions, and the Path of Least Resistance”,in J Poterba (ed.) Tax Policy and the Economy 16, 67-114.

Choi, J J, D Laibson, B C Madrian, and A Metrick (2004), “For Better or for Worse: Default Effects and 401(k) Savings Behavior”, in D A Wise (ed.), Perspectives on the Economics of Aging, Chicago: University of Chicago Press: 81-121.

Choukhmane, T (2019), “Default options and retirement savings dynamics”, Yale University mimeo.

Gelman, M, S Kariv, M D Shapiro, D Silverman, and S Tadelis, “How Individuals Respond to a Liquidity Shock: Evidence from the 2013 Government Shutdown”, Journal of Public Economics (forthcoming).

Kaplan, G, G L Violante, and J Weidner (2014), “The Wealthy Hand-to-Mouth”, Brookings Papers on Economic Activity, 2014(1): 77-138.

Madrian, B C, and D F Shea (2001), “The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior”, Quarterly Journal of Economics, 116: 1149-1187.

Plan Sponsor Council of America (2018), 60th Annual Survey of Profit Sharing and 401(k) Plans. Chicago: Plan Sponsor Council of America.

Thaler, R H (1994), “Psychology and Savings Policies”, American Economic Review, 84(2): 186-192.

a

A

Topics:  Financial markets Microeconomic regulation

Tags:  pensions, automatic enrolment, retirement savings plans

Terrie F. and Bradley M. Bloom Associate Professor of Business Administration, Harvard Business School

Professor of Finance, Yale School of Management

Robert I. Goldman Professor of Economics, Harvard University

Dean and Marriott Distinguished Professor, Marriott School of Business, Brigham Young University 

Assistant Professor of Economics and Finance, William and Mary

Events

CEPR Policy Research