Population ageing is putting pension systems under increasing fiscal pressure. As a result, it is increasingly important that pension systems minimise work disincentives. Such pension systems are less vulnerable to fiscal pressures arising from ageing. A central motivation behind many pension reforms over the past three decades has therefore been to strengthen the link between social security contributions paid during working life and pension benefits received in retirement.

Beginning in the 1980s, many countries (including Italy, Sweden, and Poland) moved away from traditional defined benefit (DB) systems toward notional defined contribution (NDC) systems. NDC schemes build in a close link between contributions during the entire working life and benefits in retirement. This means that these pension contributions resemble forced savings more than they resemble taxes. An aim of these reforms has been to encourage employment and extend working lives. Many multinational organisations such as the World Bank (e.g. World Bank 1994) and the IMF (Cangiano et al. 1998) have advocated tightening the link between current contributions and future benefits by switching from a DB to a NDC system.

For such reforms to potentially affect labour supply, that labour supply must respond to pension incentives that affect benefits many years in the future. If individuals are uninformed (Mitchell 1988, Bottazzi et al. 2006), inattentive, myopic, or liquidity constrained, changes in future pension benefits may have little impact on employment long before the age of retirement benefit receipt. Conversely, if individuals are attentive, informed, and forward-looking, then pension design can influence labour supply decades before benefits are received.

In French et al. (2026), we use a 1999 pension reform in Poland to study the extent of such forward-looking behaviour and show that employment responses to pension reforms are economically significant even many years before retirement. This implies that strengthening contribution–benefit links can have meaningful impacts on aggregate labour supply, especially if incentives are targeted towards those with elastic labour supply.

A reform that changed incentives but little else

Poland’s 1999 pension reform provides a uniquely clean setting for studying how pension incentives affect work decisions far from the retirement age. The reform replaced a traditional defined benefit system with a notional defined contribution system for individuals born after 31 December 1948, while those born on or before that date remained entirely under the old DB rules.

This reform created a sharp cohort discontinuity: two otherwise similar individuals born only days apart faced fundamentally different pension incentives from age 50 onward. Using administrative tax records covering the universe of Polish workers, this discontinuity allows credible estimation of how changes in future pension rewards affect employment well before retirement.

Crucially, many core features of the pension system were unchanged. Contribution rates, retirement ages, and financing according to the pay-as-you-go principle all remained the same. What changed was the relationship between current earnings and future pension benefits.

Under the previous DB system, pension benefits depended heavily on earnings during a limited number of an individual’s ‘highest earning’ years (usually ten years drawn from the last 20 years before the retirement date). Because earnings typically peak late in the career, employment around age 50 had a particularly large impact on future pension income. The system therefore generated strong work incentives concentrated in a narrow window of the lifecycle.

The NDC system instead linked benefits proportionally to contributions made in all working years. Contributions at any age increased future benefits in roughly the same way, spreading incentives more evenly across the lifecycle.

In regions of Poland with high wage growth, individuals were particularly likely to be in their best earnings years when in their early 50s. For these individuals, the DB pension scheme provided particularly strong work incentives due to high pension accrual in the DB scheme. In other areas with lower wage growth, individuals were less likely to be in their best earnings years in their 50s; thus, the DB scheme provided lower pension accrual and thus weaker work incentives. The move to the NDC scheme thus provided differences in pension accrual across locations. This variation underpins our research design: we compare employment responses in areas where the reform had larger effects on work incentives to areas where effects on incentives were smaller.

Although pension wealth declined somewhat under the new system, these wealth effects were similar across high and low wage growth regions. What differed sharply was the change in pension accrual. This distinction allows us to isolate how the change in incentives to work arising from pension accrual, rather than overall pension benefit levels, affects labour supply.

Workers respond long before retirement

We estimate employment responses using a regression discontinuity design comparing adjacent birth cohorts around the reform cut-off. The analysis focuses on men aged 51–54 — between 11 and 15 years before the normal retirement age.

Figure 1 shows that in these regions, employment fell by roughly 1-1.5 percentage points following the reform, corresponding to a decline of about 2–3% relative to baseline employment rates. In regions where incentives changed little, employment remained essentially unchanged.

Figure 1 Effect of the pension reform on employment

Combining these employment responses with estimated changes in the net return to work implies an employment elasticity of approximately 0.5 with respect to pension-related work incentives. This magnitude lies squarely within the range typically estimated for responses to contemporaneous taxes and transfers.

These behavioural responses occur many years before retirement benefits are received. Additional evidence from a later reform affecting contemporaneous unemployment benefits yields an elasticity of about 0.68 – only modestly larger than the response to future pension incentives. Taken together, the results suggest that workers respond almost as strongly to discounted future pension rewards as to immediate financial incentives.

Lifecycle effects of pension reform

The notional defined contribution system provides stronger work incentives at younger ages, whereas the DB system provides stronger work incentives for those in their best earnings years, which is often in their 50s. Thus it is not ex ante obvious which system provides stronger work incentives over the entire life cycle.

To assess the broader implications of a reform like the one we study, we embed the estimated behavioural elasticities in a lifecycle labour supply model which matches observed employment patterns, including the estimated labour supply responses to the reform. The model allows comparison of DB and NDC systems, holding government revenue constant.

Workers in their early 50s often face important margins of adjustment: job exit, disability pathways, or transitions toward early retirement. As a result, employment decisions at these ages are relatively sensitive to financial incentives. In contrast, labour supply among younger workers tends to be much less responsive. The DB system happened to target incentives toward ages when labour supply was most elastic – around peak earnings years. By contrast, the NDC reform redistributed incentives toward earlier ages when employment decisions are less responsive.

Figure 2 below shows key model predictions from the reform. The NDC system modestly strengthens incentives earlier in life and thus increases labour supply for those in their 30s. In contrast, the NDC system weakens incentives at ages when labour supply is most responsive. Overall, switching from the DB system to a hypothetical NDC system that is equally costly to the government would reduce total lifetime labour supply by roughly two months per worker.

Figure 2 Effect of switching to a NDC system on labour supply over the lifecycle

These declines in lifecycle labour supply arising due to a reform aimed at more tightly linking contribution to benefits are perhaps surprising. The reason is straightforward. Improvements in incentives at younger ages generate only small increases in employment because labour supply is relatively inelastic at those ages. By contrast, weakened incentives near age 50 lead to larger reductions in employment. Losses later in the lifecycle outweigh gains earlier on.

Implications for pension design

These findings carry broader lessons for pension policy.

First, workers appear to internalise how employment decisions impact future pension benefits when making these employment decisions, even 15 years before the official retirement age.

Second, the effectiveness of pension incentives depends critically on when they operate. Reforms that strengthen contribution–benefit links over the lifecycle taken as a whole may unintentionally reduce labour supply if they shift incentives away from ages at which workers are most responsive.

Finally, pension policy has important indirect fiscal consequences through behavioural responses. Policymakers often focus on the direct budgetary impact of reforms, but changes in employment throughout the lifecycle can substantially affect contribution revenues and economic activity.

As countries continue to reform pension systems in response to demographic pressures, understanding how incentives interact with lifecycle labour supply decisions will become increasingly important.

References

Bottazzi, R, T Jappelli and M Padula (2006), “Retirement expectations, pension reforms, and their impact on private wealth accumulation”, Journal of Public Economics 90(12): 2187-2212.

Cangiano, M, C Cottarelli and L Cubeddu (1998), “Pension Developments and Reforms in Transition Economies”, IMF Working Paper.

French, E, A Lindner, C O’Dea and T Zawisza (2026), “Labor Supply and the Pension Contribution-Benefit Link”, Review of Economic Studies, forthcoming.

Mitchell, O S (1988), “Worker Knowledge of Pension Provisions”, Journal of Labor Economics 6(1): 21-39.

World Bank (1994), Averting the Old Age Crisis, Oxford University Press.