Liquidity in Retirement Savings Systems: An International Comparison

What is the socially optimal level of liquidity in a retirement savings system? Liquid retirement savings are desirable because liquidity enables agents to flexibly respond to pre-retirement events that raise the marginal utility of consumption, like medical emergencies or income shocks.¹ On the other hand, pre-retirement liquidity is undesirable when it leads to under-saving arising from, for example, planning mistakes or self-control problems.²

This paper compares the liquidity that six developed economies have built into their employer-based defined contribution (DC) retirement savings systems. We find that all of them, with the sole exception of the United States, have made their DC systems overwhelmingly illiquid before age 55.

In the United States, employer-sponsored DC account balances can be moved to an Individual Retirement Account (i.e., a “rollover” IRA) once the individual no longer works for the employer, which provides considerable scope for liquidation before the withdrawal-eligibility age of 59½. Pre-eligibility IRA withdrawals may be made for any reason by paying a 10 percent tax penalty, and certain classes of pre-eligibility IRA withdrawals are exempt from this penalty.³

Liquidity generates significant pre-retirement “leakage” in the United States: for every $1 contributed to the DC accounts of savers under age 55 (not counting rollovers), $0.40 simultaneously flows out of the DC system (not counting loans or rollovers).⁴ This amount of leakage may or may not be socially optimal, an issue that is beyond the scope of the current paper.⁵

I. Analytic Framework

We focus on the five highest-GDP developed countries that have English as an official language: the United States, the United Kingdom, Canada, Australia, and Singapore.⁶ We also analyze Germany, the largest developed economy with a substantial pool of DC savings that does not have English as an official language.⁷

We analyze employer-based DC plans instead of defined benefit (DB) plans for three reasons. First, DC plans are gaining assets relative to DB plans in almost all countries around the world, including the six that we study. Second, DC plans already have more than half of retirement wealth in three of the countries that we study: Australia, Singapore, and the United States.⁸ Third, in most circumstances, DC assets are at least as liquid as DB assets, so DC assets are the relevant margin for a household considering liquidating retirement wealth to augment pre-retirement consumption.

There are many ways to measure liquidity, including the actual quantity of liquidations or the marginal price of liquidations. We use the marginal price because statistics on actual liquidations are difficult to obtain. Even if such statistics were readily available, it is unclear how they should be compared across countries. For example, should liquidations be normalized by DC balances, retirement assets, total assets, or GDP? Also, from an economic perspective, the most natural object to study is the marginal price because it summarizes the incentives that consumers face.

Accordingly, we compute the marginal rate of transformation (MRT) between withdrawal-funded consumption at ages when the household is “pre-eligible” for withdrawals and withdrawal-funded consumption at ages when the household is “eligible” to make withdrawals (in all countries that we study, eligibility begins no earlier than 55 and no later than 63): ^9,10,11

$$\mathit{\text{MRT}}=\frac{1-\tau (\mathit{\text{pre}},y)}{\left[1-\tau (\mathit{\text{eligible}},Y)\right]\times R^n}.$$ (1)

In this equation, τ(pre, y) is the marginal tax rate (accounting for penalties and phase-outs of means-tested benefits) on a $1 withdrawal from the DC plan when (i) the household is young enough to be at a pre-eligible withdrawal age and (ii) the household's employment income, y, in the withdrawal year is less than or equal to the household's permanent income, Y. Likewise, τ(eligible, Y) is the marginal tax rate on a $1 withdrawal from the DC plan when (i) the household is old enough to be eligible to make withdrawals and (ii) household earnings in the withdrawal year equal permanent income, Y. Because we are studying a situation in which the household may have a liquidity need at a pre-eligible age, we calculate how the MRT varies as we change y. We assume permanent income is Y = US$60,000, which is approximately the median household income in each of the six countries. For simplicity, we set the gross real interest rate, R, to one (i.e., we set the net real interest rate to zero). Cross-country comparisons are not affected by this interest rate assumption.

We need to make additional demographic assumptions to pin down the household's marginal tax rate. We assume the household is a one-earner married couple with no dependents that rents housing, takes the standard income tax deduction and is not disabled. In the pre-eligible withdrawal state, the earner is any age strictly under 55; in the eligible withdrawal state, the earner is at least 65 years old.

In some situations, withdrawals are completely prohibited in the pre-eligible state. We treat such a ban as a 100 percent marginal tax rate—i.e., τ(pre, y) = 1. High values of the MRT are associated with high levels of liquidity (early withdrawals are potentially encouraged), and low values of the MRT are associated with low levels of liquidity (early withdrawals are discouraged or completely banned).

II. DC Liquidity Across Six Countries

We are now ready to describe the MRT as a function of labor income during the pre-eligible withdrawal year, y, country by country. More detailed analysis and a description of our methodology are provided in the Appendix of Beshears et al. (2015b).

A. Germany, Singapore, and the United Kingdom

In Germany, Singapore, and the United Kingdom, early withdrawals are banned: MRT = 0 for all y.¹² Only disabled¹³ or terminally ill individuals may receive payments (an allowance that exists in all six countries). Singapore carves out some additional exceptions: a portion of DC balances may be used for medical expenses, a home purchase (which must be repaid with interest if the home is sold), and education (which must be repaid with interest in 12 years).¹⁴

B. Canada and Australia

In Canada¹⁵ and Australia, the MRT = 0 under normal circumstances,¹⁶ but DC balances become liquid in the event of adverse transitory labor income shocks.

Canada (Ontario)

Employer-based DC plan balances cannot be accessed before the eligibility age unless a household's expected income in the 12-month period following the application for withdrawal falls below US$32,428.¹⁷ Therefore, MRT = 0 at our hypothetical household's normal level of income: US$60,000. Once income in the pre-eligible withdrawal year falls below US$32,428, the MRT jumps from 0 to 1.11. The MRT increases with further declines in income, y, because the marginal tax rate in the pre-eligible year falls while the marginal tax rate in the eligible year is held fixed. Means-tested benefit programs generate (local) nonmonotonicities in the marginal tax rate that feed through to the MRT. As income approaches zero, the MRT plateaus at a peak value of 1.50 (see Figure 1). Hence, the Canadian DC system has the intuitive property that, for a typical household, DC withdrawals are barred when income is near its normal level but are encouraged (MRT > 1) when income declines substantially.

Figure 1. Marginal Rate of Transformation (MRT) for Canada.

Notes: This figure reports the MRT for a household in Ontario, Canada with assets from a DC Registered Pension Plan that have been rolled over to a locked-in retirement account. For more details (including an explanation for each discontinuity) see the Appendix of Beshears et al. (2015b).

Australia

In Australia, the MRT = 0 as long as the household remains employed, no matter how low income falls. However, if the household receives income support from the government for at least 26 weeks (e.g., unemployment benefits), the household becomes eligible for DC withdrawals.^18,19 Hence, Australia also has a rising MRT as income in the pre-eligible year declines if low income in the pre-eligible year is due to a long unemployment or underemployment spell and the household receives government benefits as a result (see Figure 2).

Figure 2. Marginal Rate of Transformation (MRT) for Australia.

Notes: This figure reports the MRT for a household in New South Wales, Australia, with DC assets in a superannuation fund. We assume that the reduction in employment income is due entirely to an unemployment spell. Hence, an x percent reduction in income is engendered by x percent of 52 weeks of unemployment. We also assume that the household receives unemployment benefits throughout the unemployment spell. For more details (including an explanation for each discontinuity) see the Appendix of Beshears et al. (2015b).

C. United States

In contrast, even at a normal level of income, the US DC system is liquid. Workers can roll over balances from a previous employer's DC plan into an IRA and then liquidate those balances under any circumstances with a maximum tax penalty of 10 percent. For instance, if our hypothetical household lived in Texas, its MRT with pre-eligible income equal to permanent income would be

$$\begin{array}{cc}\mathit{\text{MRT}}& =\frac{1-\tau (\mathit{\text{pre}},y)}{1-\tau (\mathit{\text{eligible}},Y)}\hfill \\ & =\frac{1-0.1-0.15}{1-0.15}=0.88.\hfill \end{array}$$ (2)

As pre-eligible income falls below its normal level, the MRT tends to rise (as in Canada and Australia) due to falling marginal tax rates in the pre-eligible withdrawal year. As pre-eligible income approaches zero, the MRT eventually exceeds one (see Figure 3). Hence, like the Canadian and Australian systems, the US MRT increases as income falls transitorily, but the rise is much more muted in the United States: the MRT increases from 0 to 1.50 in Canada, from 0 to 1 in Australia, and from 0.88 to 1.06 in the United States.

Figure 3. Marginal Rate of Transformation (MRT) for the United States.

Notes: This figure reports the MRT for a household in Texas, with some DC assets that have been or can be rolled over to an IRA. For more details (including an explanation for each discontinuity) see the Appendix of Beshears et al. (2015b).

III. Conclusions

The six countries that we study fall into three groups. In Germany, Singapore, and the United Kingdom, withdrawals from employer-based DC plans are essentially banned no matter what kind of transitory income shock the household realizes.

By contrast, in Canada and Australia, liquidity in employer-based DC plans is sharply state-contingent. For a household that normally earns US$60,000, DC accounts are completely illiquid unless annual income falls substantially, at which point the DC assets may be accessed. Canadian workers who temporarily have very low income face strong incentives to withdraw their DC balances (MRT = 1.50).

The United States stands alone in the high degree of liquidity in its DC system. Penalties for early withdrawals are relatively low, and early withdrawals are slightly subsidized as income falls transitorily.

This cross-country heterogeneity begs the question of why the United States has chosen a different path from its peers, a question we leave to future research.