Most workers begin claiming Social Security benefits before they reach normal retirement age. In 2004 (the most recent year for which detailed data is available) two-thirds of men who started claiming benefits had not reached normal retirement age; and almost half (49 percent) were 62-year-olds. Of the women who began collecting benefits in 2004, more than 70 percent had not reached normal retirement age.⁵

An individual who receives Social Security before reaching normal retirement age and continues to work is subject to an earnings test. That is, some Social Security benefits are withheld if the "retiree" earns more than a certain amount of wage income. The amount withheld depends on two thresholds, which change each year. The first threshold applies to all early retirees in 2006. Social Security withholds one dollar of an early retiree's benefits for every two dollars earned above $12,480.⁶ The second threshold applies to early retirees during the calendar year in which they reach their normal retirement age. In 2006, Social Security withholds one dollar of benefits for every three dollars earned above $33,240.⁷

"Seniors who receive early Social Security retirement benefits but keep working are subject to the earnings test."

It is important to note that the benefits withheld by the earnings test are gradually restored over time after the retiree reaches the normal retirement age. The Social Security benefit is recalculated and the individual receives a higher payment for the rest of his or her life. The adjustments used to restore benefits account for the time value of money (the fact that a dollar of benefits today is more valuable than a dollar of future benefits) and for the fewer remaining expected years of life as one delays receiving benefits. The implicit adjustment for the time value of money is a discount rate close to the real interest rate the government pays to borrow money — roughly 2 percent to 3 percent per year. Also, the implicit mortality adjustment is close to the life expectancy of the average individual. On the average, the benefits a retiree will receive equal the amount he or she would have gotten if the earnings test had not been applied — plus interest and an adjustment that takes into account the likelihood of dying at any age from 62 to 70. [See the sidebar: "The Earnings Test at Work."]

However, many retirees may not view these adjustments as an even trade for three reasons. First, those who begin claiming benefits at early ages are more likely to discount the value of a dollar received in the future at a rate much higher than the government's borrowing rate. Not only can individuals not borrow at the government's borrowing rate, they often borrow at rates much higher. If the rate at which benefits are "restored" is less than the retiree's personal discount rate, the earnings test will be perceived as a net tax.⁸ Second, individuals who retire early are likely to be less healthy than average and, as a result, their life expectancy may be lower than the average for all retirees. Finally, risk-averse individuals — those who value a certain amount of money in their pocket today over the promise of an uncertain larger payday in the future — are more likely to claim benefits early.⁹ For these reasons, early retirees are likely to perceive they are worse off because their benefits are withheld. Thus it is not surprising that a majority of retirees choose to begin receiving benefits early and many of them stop working.

Researchers have often viewed the withholding of benefits due to the earnings test as a tax, but it is not a pure tax since the benefits are later restored. However, depending on an individual's own subjective discount rate, mortality expectations and risk preferences, claiming benefits early may be a natural response to the scheduled adjustments for early retirement.

"The earnings test withholds $1 of benefits for each $2 earned above a certain threshold."

Effect on Labor Market Participation. The decision to retire is often considered a decision to both stop working and to begin claiming Social Security benefits; but, as the previous discussion showed, these are two separate choices. To what extent does the earnings test affect an individual's decision to stop working or reduce hours of work? Researchers have reached different conclusions.

According to Jonathan Gruber of the Massachusetts Institute of Technology and Peter Orszag of the Brookings Institution, the earnings test has a small labor supply effect.¹⁰ However, economists Alan L. Gustman and Thomas L. Steinmeier estimate that the earnings test reduces the share of married men between 62 and the normal retirement age who work full-time by 4 percentage points.¹¹

"The earnings test reduces the percentage of men earning more than the threshold amount."

The earnings test originally applied to all beneficiaries who received wage income, regardless of age. However, beginning in 2000 the test was abolished for all retirees who have reached their normal retirement age.¹² This change allows some conclusions to be drawn about the effect of the earnings test on the work habits of elderly Americans. Since the earnings test had previously applied to beneficiaries ages 65 to 69 and still applies to those ages 62 to 65, comparing labor market earnings before and after the 2000 legislation allows one to infer how workers and beneficiaries responded to the change.¹³

Figure II shows the percentage of retirees in two age groups who had earnings above the earnings test threshold in the four years before and after the 2000 reform.¹⁴ The first group consists of retirees who were ages 63 and 64 during the four years before or after the legislation. The earnings test applied to these workers in both periods; thus they can serve as a "control" group. The other group consists of retirees who were 66 and 67 in the four years before or after the reform; they faced the earnings test before, but not after.

Among people unaffected by the policy change (younger retirees), the percentage with earnings above the threshold fell, consistent with the trend toward less work by this age group. However, among people affected by the policy change (older retirees) the percentage with earnings above the threshold rose. This difference in behavior strongly suggests there has been a labor supply response to the repeal of the earnings test.

Another comparison confirms the previous result, focusing on retirees with above-average earnings. As Figure III shows, the percentage of older retirees with above-average earnings rose after the earnings test was eliminated for them; but the percentage of younger retirees (for whom the earnings test continues to apply) with above-average earnings actually fell.

These results imply that eliminating the earnings test for all workers will increase earnings and work by early benefit claimants.¹⁵