VoxEU Column Labour Markets

Retirement age across countries: The role of occupations

In Mexico, the average male worker retires at 75. In Bulgaria, he does so at 58. This column argues that an economy’s composition of occupations matters for its average effective retirement age as the nature of different occupations leads workers to retire at different ages. It suggests the differences in occupational composition explain up to 40% of the observed cross-country variation in retirement age.

Long-standing trends towards earlier retirement and higher life expectancy threaten the sustainability of existing pension systems. What’s more, rising debt levels in the wake of the Great Recession have intensified the need for reforms (see Diamond 2011).

The age of retirement – a widespread concern

Proposals to curb public expenditures by raising minimum retirement ages typically face stiff political resistance. In France, for example, the plan to raise the retirement age by two years provoked “a national strike […] in which more than a million people took to the streets in the biggest show of popular discontent in years” (New York Times 2011).

Such politically controversial reform plans, however, appear modest in comparison with cross-country differences in the effective average retirement age. Figure 1 exhibits cross-country variations in the retirement age of male working populations for OECD countries. The numbers vary widely between the poles of Mexico (75) and Bulgaria (58.2).

Figure 1. Effective age of retirement of male working populations in 2000

Source: OECD, based on national labour force surveys.

A new perspective on retirement ages across countries

The cross-country variation in effective retirement age is usually attributed to institutional differences that affect individuals’ incentives to retire. In a recent paper, we argue that incentives driven by social-security rules are not the sole explanation for the cross-country variation. Instead, we suggest an alternative explanation for the observed variation; the composition of occupations within an economy matters for its average effective retirement age. We show that the differences in occupational composition explain up to 40% of the observed cross-country variation in retirement age in a sample of 38 countries.

Specifically, we infer the average retirement age for 179 different occupations from US employment data. For example, the average age of retirement of psychologists is as high as 71, while airplane pilots retire around the age of 60.2. We then weight these occupational retirement ages with occupational employment shares of a set of other countries to construct a raw predictor for retirement ages across the world. Thus, the construction of the raw predictor is exclusively based on a country’s occupational distribution plus retirement decisions in the US. This raw predictor constitutes a natural benchmark against which we compare the effective retirement age across countries.

A benchmark for the age of retirement

Within OECD countries, the effective retirement age and our raw predictor are compared by computing the differences between both numbers. Figure 2 plots these differences in a bar chart. While workers in Iceland, South Korea and Switzerland retire much later than suggested by their occupational pattern, those in Hungary, Belgium and Bulgaria retire relatively early. Interestingly, the average effective retirement age of males in Spain is almost exactly the same as in the US when accounting for the occupational distribution. Another interesting fact is that the Czech Republic and Poland have virtually identical effective retirement ages yet Czechs retire nearly ten months later than Poles when correcting for the respective occupational composition.

Figure 2. Effective minus predicted age of retirement in 2000

Figure 2, which describes the deviation from the effective retirement age, takes the raw predictor, which bases on US labour market data, at face value. However, we do not view the raw predictor as an absolute measure but rather a relative one. Therefore, we standardise this measure by considering the best linear fit of the effective retirement data.

For an extended sample including 38 countries, Figure 3 provides a graphical representation of this fit, plotting the effective versus the predicted retirement age. Strikingly, the predictions perform well for very diverse countries such as Uganda, Pakistan, Gabon and Iran. Within the full sample of countries, the linear fit in Figure 3 explains about 40% of the cross-country variation in the age of retirement.

Figure 3. The effective vs. predicted age of retirement across 38 countries in 2000

It is further verified that the strong positive correlation between the effective and our raw predictor is preserved when controlling for measures of retirement incentives linked to social-security systems, per-capita GDP, life expectancy, the share of urban population and average schooling in the regression.

These results indicate that occupational distribution is a major determinant of the differences in the age of retirement across countries.

Using the new benchmark in the policy debate

Our methodology provides two measures of a country’s ‘natural’ retirement age. The raw predictor, described in Figure 2 and a linear fit as plotted in Figure 3. Both of these benchmarks provide a yardstick to assess common arguments in on-going political debates. For example, Greece, Spain and Portugal are frequently accused of retiring too early. In particular, the German chancellor Angela Merkel requested that people in Southern Europe should not ‘be able to retire earlier than in Germany’.

This argument seems far-fetched when observing that the effective age of retirement in Greece exceeds that of Germany by about 27 month. Indeed, observing these inconsistencies, the Financial Times Deutschland writes that “Merkel's push for a comparison here is both unnecessary and absurd.”

However, when accounting for differences in occupational compositions, the picture is somewhat altered. Greek excess retirement age over the German one shrinks considerably to less than 20 months (see Figure 2). Moreover, when refining the predicted occupational distribution with a linear fit, as done in Figure 3, the overall picture turns upside down. The German retirement age is exactly on the trend line, while Greece is far below the sample prediction. Indeed, the figure suggests that Greeks retire much earlier than Germans when correcting for the occupational distributions.1 In light of the cross-country differences in occupations, Mrs Merkel’s statements appear potentially unnecessary but certainly less absurd.

The case is, however, different when considering Spain and Portugal. Figure 2 shows that Spaniards work as long as Americans when accounting for the occupational distribution. Figure 3 illustrates that this statement also applies for the fitted line in the extended sample and, indeed, Spaniards do not retire later than Germans. And there is no reason to criticise Portugal. Figures 2 and 3 show that the Portuguese retire very late compared to Americans or, indeed, Germans. Thus, Mrs Merkel does seem to miss the point when criticising Spain and Portugal.

Disclaimer: Views expressed in this article do not necessarily reflect those of the Swiss National Bank.


Diamond, Peter (2011) “Demographic change, retirement and healthcare”, VoxEU.org, 2 September.

New York Times (2011) "France Moves to Raise Minimum Age of Retirement" 15 September.

Sauré, Philip and Hosny Zoabi (2011), “Retirement Age across Countries: The Role of Occupations,” Unpublished Manuscript.

1 The difference exhibited in Figure 3 is somewhat exacerbated by an adapted computation of the effective retirement age which applies to the larger sample relative to the OECD countries. See Sauré and Zoabi (2011).

9,554 Reads