BORROWING TO SAVE? UNINTENDED CONSEQUENCES OF AUTOMATIC ENROLLMENT

Transcription

1 BORROWING TO SAVE? UNINTENDED CONSEQUENCES OF AUTOMATIC ENROLLMENT John Beshears, Harvard James J. Choi, Yale David Laibson, Harvard Brigitte C. Madrian, Harvard William L. Skimmyhorn, West Point October 16, 2017

2 Standard Disclaimer 2 The views expressed herein are those of the authors and do not reflect the views or position of SSA, the United States Military Academy, the Department of the Army, the Department of Defense, any agency of the federal government, Harvard, Yale, or the NBER.

3 Additional Disclaimer 3 PRELIMINARY

4 Preview of results At four years of tenure, automatic enrollment increases cumulative employee plus employer contributions by ~6% of starting salary on average 95% CI [5.3%, 6.6%] Over the same horizon, automatic enrollment does not have a statistically significant effect on the change in debt excluding auto loans and first mortgages 95% CI [-1.7%, 2.4%] The effect of automatic enrollment on auto loans and first mortgages is positive, but offset by the accompanying purchase of an asset

5 Chetty et al. (2014): Identification from 1% mandatory contribution

6 Savings Plan Participation and 6 Automatic Enrollment 100% 80% 80% 94% 74% 90% 86% 86% 60% 40% 37% 20% 0% Before Automatic Enrollment Before AE (All) Low income (<$20K) High Income (>$80K) After Automatic Enrollment Young (<20) Older (50-59) Male Female Source: Madrian (2001)

7 Automatic Enrollment Has Dramatically Increased Savings Plan Participation in the US 7 75% 50% 25% 0% Plans with automatic enrollment Plans with participation >80% Participation rate for employees with 0-1 years tenure Source: Vanguard, How America Saves: 2013

8 8 Automatic Enrollment is Dramatically Increasing Savings Plan Participation in the UK Mandatory automatic enrollment being phased in from by firm size Opt-out rate of only 9-10% ~31 percentage point increase in workplace savings plan participation from

9 Workplace Pension Plan Participation in the UK Source: Automatic enrolment, Commentary and analysis: April 2016 March The Pensions Regulator.

10 OregonSaves Auto-IRA Program Mandatory automatic enrollment being phased in from by firm size Pilot started July 2017 To date, opt-out rate of only 23%

11 Where Does the Increased Savings 11 Come From? Reduced Consumption? Reduced Saving Elsewhere? Increased Debt?

12 Where Does the Increased Savings 12 Come From? Today s research question: Increased Debt? To what extent is the increased saving induced by automatic enrollment offset by higher levels of debt?

13 Research Question: 2 13 To what extent does preretirement leakage out of the retirement system offset the positive impact of automatic enrollment on savings?

14 Related literature Blumenstock et al. (2017) find positive but not statistically significant effects of automatic enrollment on total savings in Afghanistan, with large standard errors from self-reported data and small sample sizes Chetty et al. (2014) find in Danish data that a 1% increase in mandatory contributions to retirement accounts leads to a ~0.8% increase in total savings

15 Chetty et al. (2014) Two natural experiments: Analyze Danes switching between jobs with different mandatory pension contribution amounts n Typically 2/3 of contribution paid by employer and 1/3 paid by employee Analyze implementation of additional 1% mandatory employee contribution for employees earning > DKr 34,500 (~$5,500)

16 Chetty et al. (2014): Identification from Job Switchers

17 Chetty et al. (2014): Danish Effect Are Persistent

18 Contribution Rates under Automatic Enrollment Are Less Persistent Fraction of new hires still at default contribution rate

19 Distribution of Savings Balances by 19 Income Quintile (2007 SCF, $2013) $70,000 $60,000 $50,000 $40,000 $30,000 $20,000 $10,000 $0 Value in Savings Accounts Stra+fied by Income Quin+le, 2007 SCF (2013 $) Income Quin+le 10th Percen3le of Savings 25th Percen3le of Savings 50th Percen3le of Savings 75th Percen3le of Savings 90th Percen3le of Savings

20 Distribution of Checking Balances by 20 Income Quintile (2007 SCF, $2013) $70,000 $60,000 $50,000 $40,000 $30,000 $20,000 $10,000 $0 Value in Checking Accounts Stra+fied by Income Quin+le, 2007 SCF (2013 $) Income Quin+le 10th Percen3le of Checking 25th Percen3le of Checking 50th Percen3le of Checking 75th Percen3le of Checking 90th Percen3le of Checking

21 Distribution of Home Equity by Income 21 Quintile (2007 SCF, $2013) $1,000,000 Value in Home Equity Stra+fied by Income Quin+le, 2007 SCF (2013 $) $800,000 $600,000 $400,000 $200,000 $ Income Quin+le 10th Percen3le of Home Equity 25th Percen3le of Home Equity 50th Percen3le of Home Equity 75th Percen3le of Home Equity 90th Percen3le of Home Equity

22 22 Setting: Thrift Savings Plan Participation of U.S. Army Civilian Employees Employer contributions Non-contingent contribution: 1% of pay Match: 100% on first 3% of pay, 50% on next 2% of pay Opt-in Automatic Enrollment Contribution rate: 3% Asset allocation: U.S. Treasury Fund August 1, 2010

23 Data 23 U.S. Dept. of Defense Payroll from Compensation TSP contributions Employee demographics Credit Bureau Credit records from June and December Note: No data (yet) from the Thrift Savings Plan on actual savings plan balances, asset allocation, loans, or withdrawals.

24 Empirical Strategy 24 Compare outcomes for two one-year hire cohorts at equivalent levels of tenure Before AE Hires August 1,2009 July 31, 2010 After AE Hires August 1, 2010 July 31, 2011 August 1, 2010

25 One-Year Cohort Comparison 25 Before AE After AE Avg. annualized starting salary $56,959 $55,810 Avg. age at hire % male 61.2% 61.5% % with credit report within 6- months before hire 83.0% 83.1% Credit score before hire Sample size N=32,088 N=26,826

26 Main Outcomes 26 Ratio of cumulative TSP contributions since hire (employee + employer) to first-year pay Excludes capital gains Excludes withdrawals and loans Ratio of change in debt to first-year pay Difference between cumulative TSP contributions relative to first-year pay and the change in debt relative to first-year pay Change in credit score Change in first mortgage and auto debt as a fraction of first-year pay

27 Primary Debt Measure: Non-collateralized Debt Debt included: Non-HELOC revolving debt n Credit cards, personal lines of credit Student loans Home equity line of credit (HELOC) Second mortgages Other installment debt n Retail installment loans n Personal installment loans Accounts in external collections Residual debt n Charge cards (e.g. American Express) Debt excluded: First mortgage loans Auto loans

28 Timing of Outcomes 28 N T=1mo. T=7 mos. T=13 mos. T=19 mos. 12/31/09 J T=5 mos. T=11 mos. T=17 mos. 12/31/09 06/30/10 12/31/10 06/30/11 12/31/09 06/30/10 N T=1 mo. T=7 mos. 12/31/09 06/30/10 12/31/10 J T=5 mos.

29 Timing of Outcomes 29 N T=1mo. T=7 mos. T=13 mos. T=19 mos. 12/31/09 J T=5 mos. T=11 mos. T=17 mos. Credit Outcomes Contributions 12/31/09 06/30/10 N T=1 mo. T=7 mos. 12/31/09 06/30/10 12/31/10 Credit Outcomes Contributions J T=5 mos.

30 Equalizing The Number of Paydays At end of nth calendar month of tenure, not all employees have had the same number of paydays (and thus TSP contribution opportunities) Within cohort, some employees hired earlier in their calendar month of hire Across cohorts at some tenure months, one cohort will have had more paydays because of where biweekly paydays fall with respect to calendar month boundaries

31 Equalizing the Number of Paydays Adjustment to cumulative contributions at tenure n Let x pre = # of paychecks at end of tenure month n of pre-ae cohort member hired at beginning of their hire calendar month Let x post = # of paychecks of post-ae cohort member hired at beginning of their hire calendar month Let y = min{x pre, x post } Scale each individual s cumulative contributions to approximate what they would have been if they had had y paychecks by the end of month n n Keep contributions through months n 1 as is n Adjust by multiplying only month n s contributions by appropriate factor

32 Alternative Empirical Strategy 32 Compare outcomes for two one-month hire cohorts at equivalent levels of tenure Before AE Hires July 1,2010 July 31, 2010 After AE Hires August 1, 2010 August 31, 2010 August 1, 2010 Pros: reduce potential effect of secular trends across cohort Cons: Smaller samples, different outcome tenures

33 One-Month Cohort Comparison 33 Before AE After AE Avg. annualized starting salary $56,981 $53,787 Avg. age at hire % male 63.2% 65.2% % with credit report within 6- months before hire 82.8% 83.2% Credit score before hire Sample size N=2,432 N=3,414

34 Ratio of Cumulative TSP Contributions to Initial Pay Cumulative TSP Contributions Divided by Initial Pay 80% 60% 40% 20% 0% Pre-AE Post-AE 5.2%ñ 5.5%ñ 4.5%ñ 2.6%ñ Tenure (months)

35 Ratio of Cumulative TSP Contributions to Initial Pay Cumulative TSP Contributions Divided by Initial Pay 80% 60% 40% 20% 0% 10 th Percentile Pre-AE Post-AE Tenure (months)

36 Ratio of Cumulative TSP Contributions to Initial Pay Cumulative TSP Contributions Divided by Initial Pay 80% 60% 40% 20% 0% 25 th Percentile Pre-AE Post-AE Tenure (months)

37 Ratio of Cumulative TSP Contributions to Initial Pay Cumulative TSP Contributions Divided by Initial Pay 80% 60% 40% 20% 0% 50 th Percentile Pre-AE Post-AE Tenure (months)

38 Ratio of Cumulative TSP Contributions to Initial Pay Cumulative TSP Contributions Divided by Initial Pay 80% 60% 40% 20% 0% 90 th Percentile Pre-AE Post-AE Tenure (months)

39 Assessing the Impact of Automatic Enrollment on Cumulative TSP Contributions Regression framework Outcome: cumulative TSP contributions (at different tenure levels) Key variable of interest: indicator for being hired after automatic enrollment (at different tenure levels) Control variables: Log deflated starting salary Age quadratic Gender (binary) Education (categories) Job type (categories) College major (categories) Race/ethnicity (categories)

40 Regression Estimating equation y iτ = s Control variables X i Tenure categories Log deflated starting salary Age quadratic Gender (binary) Educational categories Job type categories College major categories Race categories [I(τ T s )(α s + β s X i + λ s PostAE i )]+ε iτ

41 The Estimated Impact of Automatic Enrollment on Cumulative TSP Contributions 8% Estimated Impact of AE on Cumulative TSP Contributions Divided by Initial Pay 6% 4% 2% 0% 5.9% -2% Tenure (months)

42 CFPB Consumer Credit Trends Data: 42 Credit Cards

43 CFPB Consumer Credit Trends Data: 43 Auto Loans

44 Ratio of Non-collateralized Debt Divided by Initial Pay 45% Non-collateralized Debt Divided by Initial Pay 40% 35% 30% Pre-AE Post-AE Time

45 Ratio of Auto Debt Divided by Initial Pay 30% Auto Debt Divided by Initial Pay 25% 20% 15% Pre-AE Post-AE Time

46 Assessing the Impact of Automatic Enrollment on Credit Outcomes Regression framework Outcome: change in debt relative to debt held at hire (at different tenure levels) Key variable of interest: indicator for being hired after automatic enrollment (at different tenure levels) Control for time trends in credit availability using the experience of hire cohorts outside of the 1-year pre-ae and post-ae hire cohorts

47 The Estimated Impact of Automatic Enrollment on Non-collateralized Debt Divided by Initial Pay Estimated Impact of AE on Non-collateralized Debt Divided by Initial Pay 8% 6% 4% 2% 0% -2% 0.3% Tenure (months)

48 The Impact of Automatic Enrollment on Net Wealth Accumulation Step 1: TSP Contributions 8% Estimated Impact of AE on Cumulative TSP Contributions Divided by Initial Pay 6% 4% 2% 0% 5.9% -2% Tenure (months)

49 The Impact of Automatic Enrollment on Net Wealth Accumulation Step 2: Non-collateralized Debt Estimated Impact of AE on Non-collateralized Debt Divided by Initial Pay 8% 6% 4% 2% 0% -2% 0.3% Tenure (months)

50 The Impact of Automatic Enrollment on Net Wealth Accumulation Step 3: Subtract Debt from Contributions Estimated Impact of AE on Net Wealth Accumulation Divided by Initial Pay 8% 7% 6% 5% 4% 3% 2% 1% 0% 5.6% Tenure (months)

51 Then Impact of Automatic Enrollment on Vantage Scores Estimated Impact of AE on Vantage Scores Tenure (months)

52 The Impact of Automatic Enrollment on Auto Debt Divided by Pay 8% Estimated Impact of AE on Auto Debt Divided by Initial Pay 6% 4% 2% 0% -2% 1.7% Tenure (months)

53 The Impact of Automatic Enrollment on First Mortgage Debt Divided by Pay 20% Estimated Impact of AE on First Mortgage Debt Divided by Initial Pay 15% 10% 5% 0% -5% 9.4% Tenure (months)

54 Interpreting the Impact of Automatic Enrollment on Auto and Mortgage Debt Starting point: Debt is used to acquire an asset. If the increase in debt is offset by an equivalent increase in the value of the underlying asset, no change in net worth n Potential mechanism: greater TSP balances may allow individuals to qualify for a larger loan (consistent with debt effects taking some time to show up) Caveats: If individuals purchase larger autos/homes post-ae, they will have a larger consumption stream going forward and correspondingly greater depreciation è net worth will decline over time è our measure of the impact of AE on net worth will be inflated If increased debt results from lower non-retirement savings and not from purchasing an asset of greater value, then our measure of the impact of AE on net worth is inflated A larger mortgage (or a new mortgage on the extensive margin) may create more forced savings over time è our measure of the impact of AE on net worth will be too low going forward in time

55 The Impact of Automatic Enrollment on Population Subgroups Total TSP contributions Debt excl. first mortgage, auto Net wealth Vantage score Auto debt Mortgage debt Salary < $34K 0.089** (0.006) (0.031) 0.089* (0.032) (3.388) 0.049* (0.022) (0.117) months of tenure Age < 30 HS only 0.055** (0.057) (0.017) 0.070** (0.019) (2.872) (0.014) (0.086) 0.069** (0.003) (0.024) 0.047** (0.014) (1.904) 0.037** (0.011) 0.170* (0.059) Vantage < 620 Black Hispanic 0.083** (0.004) (0.029) (0.029) (3.125) 0.037* (0.018) (0.091) 0.088** (0.007) (0.033) 0.096** (0.033) (4.059) (0.022) (0.122) 0.083** (0.011) (0.043) (0.044) (7.427) (0.035) (0.197)

56 Conclusions Four years after hire, automatic enrollment increases cumulative (employee plus employer) TSP contributions by ~6% of starting salary The change in non-collateralized debt is small in magnitude and not statistically significant There may be meaningful increases in auto and first mortgage debt at longer time horizons (2-4+ years) Not clear what impact this has on net wealth No evidence of increased financial distress as measured by Vantage scores and collections Results highlight the importance of viewing multiple portions of (if not entire) household balance sheet

57 Coming Attractions 57 Additional data on Army civilians Administrative data from the TSP n Balances n Loans n Withdrawals Credit bureau data for four additional firms that implemented automatic enrollment Range of default contribution rates What will the effects be when default contribution rates are higher? What will the effects be if participation rates are lower before automatic enrollment?