VoxEU Column Politics and economics Welfare state and social Europe

Drivers of pension reform measures in the OECD

Population ageing means that many current pension regimes are unsustainable, but the timing of pension reform measures is a political as well as an economic decision. This column uses new data on OECD pension reforms since 1970 to show that their timing has not been driven by projected demographic developments or political change, but by the state of the economy at the time when reforms were legislated. Pension systems have expanded more frequently during booms, and have contracted during economic downturns.

There is a public debate about the ageing population and its implications for the sustainability of existing pension arrangements. It is, however, not easy to get politicians to agree on reforms to make pensions more sustainable. This is not surprising. These measures inevitably create winners and losers. Worse, the winners from pension reform tend to be the young and future cohorts, while the losers would be current voters. Reform is therefore often only possible when the economic circumstances are sufficiently bad that it is obvious that something must be done.

Research on crisis-induced reform (e.g. Rodrik 1996, Abiad and Mody 2005) finds that a country needs to find itself in a crisis before it is able to adopt structural reform measures. The experience of the Netherlands is an example. After a stagnant decades‐long debate, the pressure of the Global Crisis meant that only a few weeks in 2012 were needed to decide on a schedule to gradually increase the public pension retirement age.

A new dataset of pension reform measures

In a recent paper, we present a dataset of all pension reform measures between 1970 and 2013 in 24 countries that have been OECD members since the 1970s (Beetsma et al. 2017). We use a narrative approach to identify, date, and classify legislative changes in pension systems, using documents from the International Social Security Association, the OECD and the International Labor Organization. Eventually, we will use the dataset to investigate the determinants of pension reform measures, based on the information available when they are legislated. Because of this, we use the date of the year the measures were legislated, which may differ from their year of implementation. For example, in the Netherlands, increases in the retirement age tend to be announced well before they actually take place. We include all reform measures that we can identify from our sources, and focus on those that we expect to affect the public budget in some way.

We classify the consequential reforms into four categories:

  • Coverage measures: These expand the coverage of a pension arrangement, for example by loosening the eligibility criteria.
  • Generosity and adequacy measures: These expand the generosity of the pension system, for example by raising the benefit level.
  • Financial and fiscal sustainability measures: These enhance the financial sustainability of the pension arrangement, for example by reducing benefits or by raising the retirement age.
  • Work incentives measures: These enhance work incentives, for example by introducing bonuses for working after the minimum age at which pension benefits can be collected.

Measures in the 'Coverage' and 'Generosity and adequacy' categories are measures that expand pension arrangements, with payment from the public purse, while measures in 'Financial and fiscal sustainability' and 'Work incentives' categories are measures that contract pension arrangements.

Figure 1 shows the numbers of reform measures in each category, for each sample year. We can make three observations. First, the number of reform measures has increased over time. Second, early in our sample the reform measures are predominantly expansionary, and later mostly contractionary. Finally, the number of measures fluctuate with the business cycle, especially later in the sample. There is a clear spike in contractionary reforms after the dot-com recession and the subprime crisis.

Figure 1 Numbers and types of pension reform measures over time

Source: Beetsma et al. (2017).

Table 1 shows the numbers of reform measures in each category over the full sample period, and the equally sized sub-periods 1970–1991 and 1992–2013. There is a balance of expansionary measures between 1970 and 2013, although the second sub-period featured slightly more of them. By contrast, confirming the pattern from Figure 1, there are far more contractionary measures after 1992. Only a quarter of the measures categorised as 'Financial and fiscal sustainability' and only a tenth of the measures categorised as 'Work incentives' took place between 1970 and 1991.

Table 1 Number of (country, year) combinations by category and reform regime

Source: Beetsma et al. (2017).

A country may have implemented more than one reform measures in a year, so we define three possible policy regimes for each (country, year) combination. A country has an 'Expanding only' regime if it only implements expanding reforms in that year, with 'Contracting only' defined similarly. We see also 'Expanding and contracting' regimes, in which a contractionary measure becomes more politically acceptable when combined with an expansionary measure. From Table 1, we can see that about three-quarters of the 'Contracting only' and ' Expanding and contracting' regimes occurred after 1992.

The business cycle drove pension reform

We use panel logistic regressions to relate the occurrence of each regime to potential driving factors. For example, these might have been a change in the current and future old-age dependency ratios, or the change in business cycle indicators such as GDP growth, the public deficit or unemployment. We focus on changes in the underlying variables, because only a change would explain why a reform occurred at that moment. Tests suggested a structural break for the 'Expanding only' regimes in 1988, for the 'Contracting only' regimes in 1992, and for the 'Expanding and contracting' regimes in 1997. We make estimates for the full sample and the two sub-samples.

Remarkably, changes in the current and projected old-age dependency ratios are never significant. Business cycle indicators, on the other hand, are highly significant for each of the three regimes in the period between 1992 and 2013. In particular, expanding reform measures occur more frequently when the economy is good, while contractionary reform measures occur more frequently when the economy is bad.

The findings are robust when we add additional economic and political controls. Contrary to conventional wisdom, political variables play an almost negligible role in pension reform.

How can we explain this?

New information about future demographics and the state of the economy arrives continuously, but our results show that governments have waited to implement reform measures until the state of the economy was particularly good or bad. They have equivocated even when they knew that their pension systems were unsustainable in the long run.

One explanation: a policy reform measure has immediate political costs. Demographic shocks typically involve news about the distant future. These developments may render the system unsustainable, but gains from reform measures also occur in the future. Business cycle shocks, however, change the current optimal payout to the elderly, for example due to optimal risk-sharing. People who have retired can only benefit from an economic windfall gain, or share in the costs of a recession, through their savings and pensions. In many countries private savings are small. This means pension benefits are the primary policy instrument that involves retirees.

In this situation, the government may ignore demographic shocks, as long as the current benefit is reasonable given current economic conditions. The government would benefit because it could postpone paying the fixed cost, it would not need to set the benefit to a level that would be sub-optimal at that time, and by postponing the decision it could reset the benefit to an optimal level given the circumstances in the future.

Governments may also ignore forecasts of demographic changes if the current state of the economy is such that the immediate gains of a pension reform measure are large enough to implement it at that time. Again, it would be optimal for the government to respond to shocks that change the current optimal payout. Information about future demographic change would be irrelevant in the decision to implement a reform measure at that moment.

Boom and bust

Population ageing calls for enhancing the financial sustainability of pension arrangements. However, our empirical findings suggest that the timing of reform measures depends on the state of the economy: governments tend to expand pension arrangements during booms, while they tend to contract these arrangements during downturns.


Abiad, A and A Mody (2005), "Financial Reforms: What Shakes it? What Shapes it?", American Economic Review 95(1): 66-88.

Beetsma, R, Romp, W and R van Maurik (2017), "What Drives Pension Reform Measures in the OECD? Evidence based on a New Comprehensive Dataset and Theory", CEPR Discussion Paper No 12313.

Rodrik, D (1996), "Understanding Economic Policy Reform", Journal of Economic Literature 34: 9-41.

3,045 Reads