Financial scarcity could be both an objective assessment and a subjective state (Cook and Sadeghein, 2018; Mullainathan and Shafir, 2013). Experiencing financial scarcity can have severe negative consequences for wellbeing (Martin and Hill, 2015; Netemeyer et al., 2018). Studies have linked financial scarcity to adverse effects in general and mental health, such as anxiety and depression (Dijkstra-Kersten et al., 2015; Sommet et al., 2018; Sommet and Spini, 2022). Financial scarcity also often leads to taking on more debt (Cook and Sadeghein, 2018; Shah et al., 2019), which has been causally linked to depression (Gathergood, 2012; Richardson et al., 2013).

Furthermore, executive functions (behavior and emotion regulation, metacognition) are also impacted by experiencing financial scarcity (O'Neill et al., 2021), as is prudent healthy behavior such as regular exercise and eating healthily (Venn and Strazdins, 2017). While detrimental effects can certainly be moderated by objective factors of lacking enough money for necessities, a significant part is subjective, how much stress and worry somebody feels about their financial situation (de Bruijn and Antonides, 2020; Haushofer and Fehr, 2014; Simonse et al., 2022; van Dijk et al., 2022).

Financial scarcity also affects cognitive functions by focusing or tunneling attention to finances (Mullainathan and Shafir, 2013; Shah et al., 2012; Tomm et al., 2023). This means that individuals experiencing financial scarcity are likely to (over)focus on financial matters at hand, which can impose cognitive strain, leading to a decline in mental performance (Mani et al., 2013). This effect has seen more recent corroboration, as a scarcity mindset could lower available attention to allocate to other tasks (Kalil et al., 2022; Lichand and Mani, 2020). Cognitive effects have also been documented as an increased cognitive load closer to the paydays, even if the changes to their cashflow were trivial (Mani et al., 2020). Perhaps financially constrained individuals even perceive the world quite differently to non-constrained individuals, placing an economics dimension to everyday life tasks (Shah et al., 2018).

However, the financial scarcity and its effect on cognition stream of research has seen mixed results (Carvalho et al., 2016; de Bruijn and Antonides, 2022). In particular, the effects of financial scarcity and cognitive fatigue originally posed by Shah et al. (2012) do not replicate very well (Camerer et al., 2018; Shah et al., 2019). Furthermore, the results by Mani et al. (2013) on diminished mental performance do not replicate either, when income levels are not analyzed as a dichotomous variable (Wicherts and Scholten, 2013). Therefore, it is unclear if subsequent cognitive tasks suffer in performance when an individual is in a financial scarcity mindset, but the evidence for attention tunneling remains strong.

The literature shows the connection between a financial scarcity mindset and several issues ranging from wellbeing to lifestyle choices but remains unclear on cognitive performance. However, the question of whether experiencing financial scarcity increases prudent financial behavior or leads into an objective “poverty trap” remains unanswered.

The effect of financial scarcity of prudent financial decision-making has seen mixed results in the literature (Hamilton et al., 2019). It could be argued that having limited means focuses individuals experiencing financial scarcity toward being more efficient with money, as they stretch every bit of money to their maximum efficiency (Shah et al., 2012). At the same time, there is evidence of financial scarcity being connected to a more present biased financial decision-making (Carvalho et al., 2016) leading to a poverty trap (Hilbert et al., 2022).

Financial constraints could influence individuals to make more prudent financial decisions (Shah et al., 2012), mainly through focusing attention to the financial task at hand (Shah et al., 2019), better understanding of correct item price valuation (Shah et al., 2015), and theoretical observations of better trade-off analysis of opportunity costs (Bertrand et al., 2006; Frederick et al., 2009). As such, opportunity costs are more likely to be considered when individuals perceive (financial) constraints (Spiller, 2011). However, recent research has produced results contrary to this idea, with financially constrained and non-constrained participants considering trade-off costs in a similar fashion, as both groups neglect opportunity costs (Plantinga et al., 2018).

Financial scarcity could also have a detrimental effect on financial decision-making. Financial scarcity is linked to financial avoidance, such that an individual experiencing financial scarcity is likely to avoid assessing their current financial situation, learning financial information and making a financial decision that suits their long-term goals (Hertwig and Engel, 2016; Hilbert et al., 2022). Not wanting to know about ones' current financial situation could be a result of an utility calculation – if the financial situation of the bank account (once checked) would turn out worse than expected, it might trigger sadness. Financial scarcity can also disrupt important financial skills such as calculating loan costs and reborrowing to cover previous loans (Cook and Sadeghein, 2018). Likewise, there is evidence for financial constraints and financial scarcity related to overborrowing (de Bruijn and Antonides, 2022; Shah et al., 2019). Therefore, individuals in a financial scarcity mindset could be present biased to the extent that they seek relief to their financial worries in loans, disregarding the high costs associated.

Finally, financial scarcity, is found to be connected to higher risk aversion and increased temporal discounting (Dohmen et al., 2011; Green et al., 1996; Haushofer and Fehr, 2014). This means that the individuals experiencing financial scarcity are more likely to focus on present financial issues (or be present biased, e.g., O'Donoghue and Rabin, 1999), disregarding future financial gains, which in turn lowers lifetime wealth (Finke and Huston, 2013). Sequences of overly discounting future yields could create a feedback loop which reinforces financial scarcity. In other words, the financial scarcity mindset is associated with lower retirement savings due to temporal discounting (e.g., and attention tunneling to present financial worries. This leads us to our hypothesis:

As the experiment took place in Estonia, we need to address key institutional details about the Estonian pension system. The Estonian pension system relies on three pillars: the first pillar is a tax financed pay-as-you-go pension, with the individual accumulating rights to receive a state pension after 15 years of working in Estonia. The second pillar is a mandatory funded tax-favored pension scheme with the option to opt out, while the third is a tax-favored and funded voluntary pillar (Estonian Social Insurance Board, 2025; Piirits and Võrk, 2019). The second pillar was established in 2002 and is mandatory for individuals born from 1983 and onwards, others can apply to join. Individuals can choose a mutual fund to invest in both the second and the third pension pillar (Piirits and Võrk, 2019). Individuals are recommended to supplement their retirement income by having savings independent of the pension system, an unofficial fourth pension pillar.

Inadequate retirement preparation is a critical issue in Estonia. According to Eurostat (2021), Estonia has the largest share of employed elderly (age 65-69) in Europe (32.5%). That is more than double the proportion of 65- to 69-year-olds in the European Union average (13.2%). Compared to the two other Baltic countries, the Estonian elderly employment figures are similar (Latvia at 29%, Lithuania 26.8%). Estonia is also one of the countries with the lowest pension income replacement ratios in the OECD, where an average pension replaces 34% of pre-retirement income, according to 2022 data (the OECD average is 61%) (OECD Net Pension Replacement Rates, 2021). With such a low replacement ratio, the relative poverty rate of pensioners in Estonia is one of the highest among OECD countries. In 2022, 37.6% of individuals aged 65+ in Estonia are in relative poverty, while the average for OECD countries is 13.1% (OECD Relative Poverty Rates, 2021). There are again parallels with the two other Baltic countries, even though Latvia has a higher pension replacement rate of 53%, Latvian elderly relative poverty rate is 33%, while Lithuanian' figures are 29 and 25%, respectively.

Based on the year-end data from the Estonian Ministry of Finance Database (2022), 43% of the population (or 574,137) were enrolled in the 2^nd pension pillar with a fund size of 4 billion €, while 147,800 were enrolled in the 3^rd pension pillar with a fund size of 0.67 billion € (and a further 38,700 people in an insurance fund). Therefore, the current average accumulated pension assets per capita in these two pillars is small, 6,745€ for the average 2^nd pension pillar and 3,602€ in the 3^rd pillar. The median accumulated savings are 2,566 in the 2^nd pillar and 1,015 in the 3^rd pillar, indicating high variance between accounts. Furthermore, in 2022, about 447,000 people made contributions to their 2^nd pillar and 108,000 made contributions to their 3^rd pillar, indicating that many do not regularly contribute to their 3^rd pillar funds. The average age of a 3^rd pillar contributor is 44.3, their average income is 2,645€, which is considerably higher than the average salary of 1,685€ (based on 2022 data).

It is unlikely that individuals are preparing for retirement with independent investing, as the share financial assets in household net worth in 2021 is 15%, which is lower than the European average of 1/5 of net assets (Estonian Central Bank, 2021). Finally, as a recent study on Estonian pension literacy showed that only 10% of the population is sufficiently on track with retirement preparation, while many future retirees appear to depend on the state pension instead (Pulk et al., 2024), which was recently indexed to an average of 700€ per month for 2024 (85% of the national minimum wage).

Joining the 2^nd pension pillar scheme is mandatory for all Estonian tax residents when they turn 18 (Estonian Social Insurance Board, 2025). Even though contributing is not possible without being employed, each individual can still set up a pension fund for when they eventually start to work. If a fund is not chosen, a passively managed index fund is randomly assigned when receiving the first paycheck. However, it is possible to opt out of the 2^nd pension pillar contributions entirely, in which case there is a 10-year ban from joining the scheme again. The 3^rd pension pillar is voluntary, and savers can freely deposit and withdraw from the fund at any time.

For the 2^nd and the 3^rd pension pillars, different tax incentives are in place to encourage saving. Contributing to the second and third pension pillars is income tax deductible (20%, increases to 22% in 2025), although there is a maximum contribution cap of 15% of yearly income (up to a maximum of 6,000 €) for the 3^rd pension pillar. In addition, there is an important policy change that is implemented during this experiment regarding pension pillar contributions. Namely, starting in 2024, Estonians can select to increase their contribution into the 2^nd pension pillar from 2% of their monthly gross salary to either 4 or 6%. This gives a suitable avenue to examine the intention to change the current contribution rate into the second pillar which we use in our experimental materials.

The sample consists of 134 bachelor and master level economics students in Estonia (see Table 1 in the design section for details). The experiment was conducted in October 2024. Participants did not receive an incentive to complete the study, apart from a learning experience. We deliberately decided not to include an incentive as it might distract participants from the research question, as there is no normatively objective correct benchmark available to incentivize for a pension savings rate.

Our experiment² features two treatment groups, in which participants receive a similar but slightly modified vignette that describes their friend Keit (Keit is a common unisex name in Estonia). We construct our vignette based on The Psychological Inventory of Financial Scarcity, or PIFS (van Dijk et al., 2022) scale (see Appendix A for the PIFS survey questions). The PIFS is a self-assessed subjective appraisal of one's affective response toward their financial situation, combining the psychological stress framework (Cohen et al., 1997; Cundiff et al., 2020) and scarcity theory (Mullainathan and Shafir, 2013). Therefore, the PIFS highlights situations where financial concerns exceed available resources, which in turn narrows attention and focus to these concerns. The PIFS consists of twelve items in four main components – scarcity of money, lack of control over personal finances, financial worries and rumination, and short-term focus. It has shown to be a robust measure (Hilbert et al., 2022; Simonse et al., 2022; van Dijk et al., 2022).

We combine the four main components of PIFS into a vignette of a person. We opt for a vignette design as their precision is usually as good (or sometimes even better) as personal recommendations (Stantcheva, 2023) and avoids the ethical concerns of inducing financial scarcity in the participants. For the vignette, one treatment group receives a description of a person that highlights subjective financial scarcity components (see Appendix A for more details), while the other group receives a description that refers to the opposite. We call these treatment groups “stress” and “no stress” respectively. The text the participants receive is the following (no-stress group modifications in parenthesis):

We include details about Keits' living situation, which is common for both experiment groups, informing participants that Keit has full-time employment, earns the national average wage, lives in a (to the sample audience) known and common neighborhood, is not in debt, and a few lifestyle elements. We chose these lifestyle elements (dining out and going to the cinema) as examples of common social activities among younger Estonians living in urban areas. While affordability varies, we wanted to avoid having too costly elements, while still showing that Keit has discretionary income to spend. The function of this text is to reduce ambiguity in assessing living conditions and show that it is purely a subjective perception of financial scarcity.

After the participants read through the randomly assigned description, they proceeded to answer three questions regarding saving for retirement and a few control questions, as detailed in the following measures section (see Appendix B for the full survey).

The study includes three dependent variables, (1) second pension pillar contribution rate, (2) third pension pillar contribution rate, (3) personal portfolio contribution rate. The second pension pillar contribution rate is measured as a categorical variable with 0, 2, 4, and 6% as possible options. Both the third pension pillar contribution rate and the personal portfolio rate are continuous variables ranging from 0 to 15%. The cutoff point was set at 15% as this is the maximum tax-incentivized contribution rate into the 3^rd pension pillar and to limit unrealistic total savings rates, which we compute as a sum over these three dependent measures.

The control variables we included were gender, work situation (working full-time, part-time, doing gig work, not working, or being an entrepreneur/CEO of own company), and how much the participant personally contributes to their second and third pension pillar funds³.

As financial scarcity is associated with risk aversion and increased temporal discounting (Dohmen et al., 2011; Green et al., 1996; Haushofer and Fehr, 2014), we included measures for personal financial scarcity (Sergeyev et al., 2023), risk attitude and time preference (Dohmen et al., 2011), confidence in making pension decisions (Tokar Asaad, 2015), and trust in pension providers (Hansen, 2012). These measures are all elicited on 10-point scales with 10 indicating highest agreement (see Appendix B for details).

We first analyze the sample and randomization between groups with chi-squared tests and one-sided ANOVA where appropriate. The main effects of the experiment are assessed with one-sided ANOVA. Finally, we examine the influence of the various attitude measures and background variables by estimating a two-limit tobit model. As explained in Section 1, two-limit tobit is used instead of OLS because the decision variable exhibits both lower and upper censoring as a consequence of the limits set in the experimental design.

We start with descriptive statistics (Table 1) and randomization checks through chi-square tests. Female representation is slightly higher in the no stress group (65.2%) compared to the stress group (49.2%), but this difference is not statistically significant (p = 0.14). Employment patterns are similar between groups, with full-time work being almost identical (29.0 vs. 29.2%), while part-time work and gig work have modest differences (21.7 vs. 16.9% and 4.3 vs. 12.3%, respectively) that are not statistically significant. Personal contribution rates to the second and third pillars are comparable, participants on average contribute 2% to the 2^nd pillar and 2% of their salary to the 3^rd pillar. The average age of the participants is 23 in the no stress group and 24 in the stress group. Psychological variables, including personally perceived financial scarcity, risk attitude, time preference, confidence in pension decisions, and trust in pension providers, suggest similar trends in attitudes and perceptions across the two groups. Therefore, randomization between groups was successful.

As the experiment design includes dependent variables that have lower and upper limits by design, we should assess possible accumulation of observations at the limits. For this reason, we examine histograms (see Figure 1). The histograms show rather marked patterns between the two groups, with the contribution rates in the stress group being clustered more to the left.

For all dependent variables, the effect direction was consistent – participants recommended the person experiencing financial scarcity to save less money for retirement (see Table 2). The participants who were shown the financially stressed depiction of Keit, suggested an average of 4.92 percentage points lower total contribution rate compared to their counterparts who saw a description of a financially non-stressed Keit. We saw a similar effect across all measured variables – contribution into the 2^nd and 3^rd pension pillars, and contribution rate to personal portfolio.

Next, we assess the effects of other variables on the contribution rates within the experiment. We do this by conducting a two-limit tobit regression for each of the four dependent variables (see Table 3).

The results of the two-limit tobit regression model indicate that the treatment group variable (i.e., whether the person was randomized into the scarcity or no-scarcity condition) remains strongly statistically significant for predicting contribution rates, even when taking control variables into account. We observe only minimal other statistically significant indicators, such as personal 2^nd pension pillar's contribution rate predicting contribution recommendation to the 2^nd pension pillar and likewise for the 3^rd pension pillar, gender for the 2^nd pillar and time preference for the personal portfolio contribution rate. The only control variable that remains weakly relevant for the total contribution rate is personal 3^rd pillar contribution rate, which can signify a more financially savvy investor⁴. We conclude that the treatment effect was the main relevant contributor to retirement savings recommendations.