This study tells us something we perhaps feel inside, that giving us the freedom to get out of saving retirement doesn’t make retirement very good.
Here’s the abstract
Both insufficient saving and early withdrawal of retirement savings jeopardize financial well-being in old age, but the latter is rarely studied.
We examined which individual differences relate to both regular saving and early withdrawal decisions. These decisions are economically similar (accumulating retirement assets) but psychologically different (incurring regular small losses vs. abstaining from a one-off large gain).
We analyzed three representative surveys conducted in Estonia in 2020, 2021, and 2022, surrounding a policy reform that introduced an early withdrawal option to a previously mandatory pension saving scheme. Limited trust in financial institutions and non-Estonian nationality emerged as statistically and economically significant factors associated with the pair of decisions both not to save and to withdraw early.
Fostering institutional trust represents a key strategic goal for pension systems. Our findings also call for caution with pension freedom reforms, which may put those who already save too little in double jeopardy.
Here’s the introduction to the report that may lead to you the longer report that is linked below and here.
To achieve a financially secure retirement, people not only need to save adequately throughout their working lives, but they also need to avoid withdrawing their savings prematurely. Both behaviors can be influenced by macroeconomic trends (OECD, 2024; Fuentes et al., Reference Fuentes, Mitchell and Villatoro2025) as well as pension schemes and policies (e.g., auto-enrolment, tax incentives, options for hardship withdrawals, and withdrawal penalties (Benartzi and Thaler, Reference Benartzi and Thaler2007; Beshears et al., Reference Beshears, Choi, Hurwitz, Laibson and Madrian2015)).
Beyond such contextual factors, various characteristics of individuals – the focus of our study – also play a significant role. Whereas the relations of individual demographic and personality characteristics with saving behavior are relatively well researched (see Gerhard et al., Reference Gerhard, Gladstone and Hoffmann2018; Ye et al., Reference Ye, Post, Zou and Chen2025 for an overview), their associations with early withdrawal behavior remain understudied (see Loibl et al., Reference Loibl, Summers, McNair and Bruine de Bruin2019 for an exception). We address this gap by studying which characteristics are associated with not saving and with early withdrawal, and which of these overlap, constituting risk factors placing individuals into the ‘double jeopardy’ of not saving and withdrawing early.
The decisions to save for retirement and to withdraw savings early may be influenced by the same as well as by different factors. On the one hand, both saving and withdrawal are related to building up retirement assets, so one could expect withdrawal to be influenced by some of the same factors as savings. On the other hand, saving decisions involve trading small present losses for potential future gains, whereas withdrawal decisions involve trading large present gains for potential future losses, suggesting that each decision may also be sensitive to a unique set of factors. Indeed, prior literature has shown that reference points and aspirations influence risk-taking behavior over time (Hoffmann et al., Reference Hoffmann, Henry and Kalogeras2013), which is particularly relevant when individuals evaluate the trade-offs inherent in saving and withdrawal decisions.
Saving behavior is related to a variety of individual characteristics. For example, higher saving rates are associated with socioeconomic attributes such as higher income (Dynan et al., Reference Dynan, Skinner and Zeldes2004). In addition, higher savings are also related to personality traits, such as conscientiousness, suggesting that disciplined and organized individuals can withstand temptations for current consumption and generate persistent saving habits (Nyhus, Reference Nyhus2017). More broadly, personality traits such as conscientiousness help explain individuals’ simultaneous pursuit of a healthy lifestyle and financially responsible behavior (Hoffmann and Risse, Reference Hoffmann and Risse2020).
Much less is known about early withdrawals, potentially due to their episodic nature and lack of data. The limited existing research suggests that early withdrawals are used as a buffer against urgent financial needs (Amromin and Smith, Reference Amromin and Smith2003; Argento et al., Reference Argento, Bryant and Sabelhaus2015; Agarwal et al., Reference Agarwal, Pan and Qian2020; Wang-Ly and Newell, Reference Wang-Ly and Newell2022; Bateman et al., Reference Bateman, Dobrescu, Liu, Newell and Thorp2023) and are less common among people with higher general financial literacy (Tharayil and Walstad, Reference Tharayil and Walstad2022). We expect that just like regular saving, early withdrawal is not merely an economic decision, but psychological factors will play a role as well.
It is crucial to look for possible overlap of factors influencing both saving and early withdrawal decisions, as some groups of individuals could be in double jeopardy of risking building up too little retirement assets. Factors that concurrently intensify both behaviors will have a significantly greater impact. Thus, identifying these ‘double jeopardy’ factors is essential for developing targeted and effective interventions.
To this end, we explored a wide range of sociodemographic and psychological individual characteristics in relation to regular saving and early withdrawal decisions, utilizing three representative online household surveys from Estonia. The samples were collected around the 2021 reform that introduced the option to withdraw pension assets early (and completely), in an institutional setting where individual retirement savings adequacy is an economically relevant research question. According to the OECD, the pension savings of Estonians provide, on average, only 35 percent of their pre-retirement income, far behind the OECD average of 61 percent (OECD, 2023).
More specifically, Estonia’s pension system comprises three pillars. The first pillar is the pay-as-you-go state pension, where the current pensions are paid from current social tax revenues. First pillar pension income is a function of employment income and years of employment and provides only a very basic level of income (old-age poverty rates in Estonia are among the highest in the OECD; OECD, 2023). The second and third pillars consist of tax-favored, funded, individual accounts, similar to US 401(k) plans. Both pillars have tax benefits during the accumulation and decumulation phases. Individuals do not have to pay income tax on savings made (a universally applicable flat rate of 20% at the time of the study). At the time of data collection, contributions to the second pillar were 2 percent of gross salary (plus 4% added by the state). The tax benefit for the third pillar is capped at 15 percent of yearly income or €6,000 (whichever is lower). Savings in those pillars can be flexibly invested into a broad range of mutual funds differing in style (e.g., passive vs. active) and focus (e.g., index vs. specific economic sectors). Pension payouts at retirement age from those pillars can flexibly combine lump-sums, payout plans, and annuities. Tax rates on lump-sums are 10 percent, while income from payout plans and annuities is tax-free. Additionally, pension savings are inheritable upon death. Taken together, the tax incentives and investment flexibility make the second and third pillars an attractive scheme for retirement savings.
In this paper, we focus on a reform that affected the second pillar. Prior to the reform, participation in this pillar was obligatory; that is, contributions were made based on all taxable income for individuals born after 1984 (participation in the third pillar is optional). With the reform introduced in 2021, participation in the second pillar became voluntary, and for the first time, early withdrawals were permitted. At this stage, approximately 700,000 individuals had accumulated a total of €4 billion in the second pillar. Over 25 percent of account holders chose to withdraw funds in the first year after the reform, and by 2025, the proportion of withdrawers had reached 36 percent (Pension Registry, Nasdaq, 2025). Withdrawing savings comes with significant disincentives, including a 20 percent income tax, a 10-year waiting period to rejoin the second pillar, and inflexibility, as only complete balances can be withdrawn. Importantly, the reform combined two changes: making second-pillar participation voluntary and introducing the option of early (full) withdrawal. Our setting does not contain within-reform variation that would allow us to separate the effects of these two arms.
Our analysis of the three household surveys shows that, first, there is a set of ‘double jeopardy’ factors associated with both saving and early withdrawal decisions. Notably, lack of trust in financial institutions was the strongest predictor of both not saving for retirement and early withdrawal of funds after the reform, extending earlier findings on savings (Ricci and Caratelli, Reference Ricci and Caratelli2017; Goedkoop et al., Reference Goedkoop, Mangan, Mastrogiacomo and Hochguertel2023) to withdrawal decisions. Moreover, non-Estonian nationality (typically Russian) was consistently associated with a lower likelihood of saving and a higher likelihood of early withdrawal.
Second, we find that withdrawal decisions cannot be explained solely by economic motivations such as liquidity constraints, which were the main explanation highlighted in earlier research using data that did not include psychological factors (Argento et al., Reference Argento, Bryant and Sabelhaus2015; Agarwal et al., Reference Agarwal, Pan and Qian2020; Wang-Ly and Newell, Reference Wang-Ly and Newell2022; Bateman et al., Reference Bateman, Dobrescu, Liu, Newell and Thorp2023; Fuentes et al., Reference Fuentes, Mitchell and Villatoro2025). Third, in line with previous findings (e.g., Donnelly et al., Reference Donnelly, Iyer and Howell2012), conscientiousness was a significant predictor of saving but not of early withdrawal decisions, suggesting psychological differences between these decisions.
A deeper understanding of individual predictors of saving and early withdrawal decisions is important not only from an academic perspective but also for public policy (Newall and Peacey, Reference Newall and Peacey2021). For example, recent findings from Chile indicate that a quarter of national retirement assets were withdrawn early during the COVID pandemic, resulting in over 4 million individuals depleting their entire accounts (Fuentes et al., Reference Fuentes, Mitchell and Villatoro2025). Evidence from Australia, which allowed early withdrawals during COVID as well, indicates that total future pension payments will be reduced by €51 billion, amounting to €56,000 per person that withdrew (Chong, Reference Chong2024) or 2 percent of GDP (Hamilton et al., Reference Hamilton, Liu, Miranda-pinto and Sainsbury2024). In the United Kingdom, a total of more than £83.6 billion has been withdrawn flexibly from pensions since the pension freedom reform in 2015 (HM Revenue & Customs, 2024). Recent analysis also reveals that early withdrawal from 401(k) accounts in the United States has reduced the previously reported success of auto-enrollment efforts, with an estimated 42 percent of 401(k) balances withdrawn upon job separation (Choi et al., Reference Choi, Laibson, Cammarota, Lombardo and Beshears2024). In the long-term, therefore, such withdrawals can considerably decrease financial retirement security.