The information on people tapping 401(k) funds prior to retirement was obtained from two questions, one asking whether a person or spouse had taken out a loan from a 401(k) plan in the last 12 months and the other asking about hardship loans over the same time period. The survey did not ask for information about people choosing to cash out their retirement plans when leaving their employer. Hence, answers to these questions may understate leakages from 401(k) plans. Also, the narrow time frame of both questions, transactions in the last 12 months, may lead to an understatement of total leakages from 401(k) plans.
The survey also included substantial information on debts and assets for respondents. Information on consumer debt was obtained from several questions asking whether respondents had unpaid credit card balances at the end of the month, student loans, medical debt and automobile loans.
The information on leakages from 401(k) plans and consumer debt was used to create a 2x2 contingency table describing the relationship between tapping 401(k) plans and consumer credit (Table 1).
The data in this contingency table can be used to calculate the conditional probabilities. P(Taps 401k)/Has some consumer credit) = 1348/5803 or 0.2323 and P(Tap 401(k)/no consumer credit) = 79/2107 or 0.0375.
These conditional likelihoods can be used to obtain the conditional odds: 0.3026 (0.2323/(1 − 0.2323)) and 0.0390 (0.0375/(1 − 0.0375)) for tapping 401(k) plans conditional on having and not having consumer debt. The ratio of these odds gives a statistic called the odds ratio, which is 7.76. The same odds ratio could be obtained from a simple linear logistic regression model where tapping 401(k) plan is the dependent variable and having some consumer credit is the independent variable.Footnote 10
The odds of a person tapping 401(k) funds through a loan or hardship distribution is 7.76 times larger for a person with consumer credit than for a person without consumer credit.
Multivariate logistic regression models
The existence of consumer debt is of course not the only variable impacting the decision of whether a person takes out a 401(k) loan or a hardship distribution. The tendency to tap a 401(k) plan prior to retirement can be impacted by a variety of characteristics of the respondent including age, education, income, number of children, homeownership, as well as the financial situation of the borrower. The National Finance Capability Study has information on respondent characteristics and finances that allow us to create a more complete model of the decision to tap 401(k) plans through loans or hardship distributions. The multivariate model has the advantage of holding constant other factors that also influence the decision to take out a 401(k) loan or a hardship distribution.
The odds ratios for a multivariate logistic regression model for 401(k) loans and hardship withdrawals are presented in Table 2.
The odds ratios for each explanatory variable in the logistic regression model are presented in the second column of the table. The odds ratio compares the odds of tapping a 401(k) for the included category to the odds for the omitted category. An odds ratio greater than/less than 1.0 indicates respondents with the included characteristic are more/less likely to take out a 401(k) loan or hardship distribution than respondents with a characteristic in the omitted group.
The odds ratios for the three financial distress variables in this model—some consumer credit, bad credit rating, and underwater mortgage—support the view that people in financial distress are highly likely to tap 401(k) funds prior to retirement.
The odds of people with some credit tapping their 401(k) loan are 4.6 times higher than for people with no consumer credit.
The odds of a person with bad credit tapping their 401(k) loans are 2.2 times higher than the odds for person with good credit.
The odds of a person with home value less than their mortgage tapping their 401(k) plan are 13.1 time higher than for a person not in this situation.
The logistic regression model was used to generate the likelihood of tapping a 401(k) plan prior to retirement for three individuals one with no consumer credit,—one with consumer credit and a good credit rating and one with consumer credit and a poor credit rating.Footnote 11 The choice of the three types of individuals stems from the tendency for respondents with poor credit ratings tend to have some consumer loans (Table 3).
These figures indicate people with debt and a good credit rating are 4 times more likely to take out a 401(k) loan or hardship distribution than a person with no consumer debt in the last 12 months. People with debt and a poor credit rating are 7.4 times more likely to use a 401(k) loan or a hardship distribution in the last 12 months.