Fostering old-age saving under incomplete rationality

A TFI research project by Krzysztof Makarski, Artur Rutkowski, & Joanna Tyrowicz
Posted on November 29, 2019

Fostering old-age saving under incomplete rationality” is one of the short-term research projects supported by the Think Forward Initiative.

To live a care-free latter, we need to make choices early in life and save up money for our pension. Most developed countries do provide voluntary saving programs, based on a fully rational homo oeconomicus. But reality teaches us differently: people usually fail to stick to their pension planning. We do not put enough money aside and might risk poverty in the last decades of our lives. Could models introducing bounded or incomplete rationality help evaluate the current pension systems? This is the question that researchers Krzysztof Makarski, Joanna Tyrowich and Artur Rutkowski (FAME | GRAPE) investigated in their TFI research project. How far are we off “mark” when we retire, given our bounded rationality? And how could the current pension system, based on voluntary savings schemes be improved?

Download the research report below and find out!


Economic literature is populated by the infamous homo oeconomicus, a quite perfect model of our own species. This model is endowed with superhuman data-processing and decision-making abilities. No government help needed; homo oeconomicus will happily budget, spend and save smartly.

Unfortunately, a substantial body of empirical evidence shows that most real-world people differ greatly from this benchmark individual – especially when it comes to saving for old age. Our brain is simply not wired to think forward. Homo sapiens seems to find it very hard to plan for the near future, let alone for decades ahead.

When people make suboptimal decisions about their retirement savings, they risk accumulating insufficient assets to supplement their pension benefits. To tackle the issue of looming old-age poverty, policymakers across all OECD countries introduced government subsidized voluntary pension schemes, hoping to raise asset accumulation amongst households. They typically offer preferential tax treatment to encourage participation.

But are such schemes and treatments good enough? In our research, we study real individuals whose decision-making process deviates from that of homo oeconomicus, their perfect counterpart. We attempt to answer three questions:

  • To what extent do less perfect decision makers differ from homo oeconomicus?
  • Could a policy bridge the gap between fully rational and incompletely rational individuals in terms of their assets at various points in life?
  • Can policymakers improve people’s financial situation with the introduction of government subsidized voluntary pension schemes?

We conducted our study using a full-fledged macroeconomic model, allowing us to take into account that any subsidy granted entails an increase of taxes somewhere else in the country. We identified several characteristics of people’s suboptimal decision-making, observing how the lifetime wealth accumulation paths of people with such behavioural features differ from homo oeconomicus. We then investigated if these people’s welfare could be improved through subsidized voluntary pension schemes. The results are of interest to policymakers looking for the optimal design of these schemes.

Who is homo oeconomicus?

Homo oeconomicus stands for being fully rational, which implies three basic features of decision making:

  • Individuals have perfect foresight and their preferences are stable over time.
  • Individuals have unconstrained ability to process all available information and to absorb it in current and future choices.
  • Individuals have unconstrained ability to transfer assets between periods (i.e. accumulate for the old-age or store precautionary savings for the periods of adverse shocks to earned income).

These three premises guarantee that individuals construct optimal life-cycle profiles, execute their strategies to perfection, and use all resources efficiently. It does not mean that all individuals are wealthy or equal. It may still hold that one is endowed with lower productivity (earning less), higher leisure preference (working less) or lower patience (saving less).

Individuals who are fully rational make labour supply and savings choices in such a way that their lifetime consumption profile is smooth and possibly equal in present value terms. This means that individuals rationally adjust not only with respect to their time preference (such as patience, in this case being the willingness to postpone consumption to the future periods), but also to the probability that they will live long enough to experience this consumption (life expectancy). Since nobody likes sudden consumption drops, homo oeconomicus prefers smooth consumption over his entire lifetime.

Frequent - even if suboptimal - choices

Allowing for incomplete rationality, we offer novel insights by studying four types of individuals who do not follow the “gold standard” of homo oeconomicus:

  • Adaptive learners. We relax the assumption that gathering and processing the information is costless. This renders costly updating expectations (and changing one’s mind about what is optimal). Economists call this phenomenon the rational inattention. Individuals who find adjusting their expectations costly, update their lifetime plans only when the already materialized changes in the environment are large enough. We call such individuals adaptive learners, assuming that they notice the world has changed in the past, but believe that it will remain unchanged for the rest of their lifetime. Not realizing soon enough that today’s working age cohorts will live longer than the contemporaneously observed elderly will leave the adaptive learners unprepared for retirement.
  • Present biased individuals. Some people may be able to come up with a perfect plan, but are unable to implement it due to a variety of self-control issues. Individuals have present bias, which means that they exhibit time preference similar to homo oeconomicus. However, they have an additional preference for the present, and as a result they are time inconsistent, not sticking to their own plans. An illustrative example is to plan to start healthy lifestyle “tomorrow”. Tomorrow comes, and we still plan to start healthy lifestyle tomorrow.
  • Hand-to-mouth individuals. We added individuals who consume their entire disposable income. We call those people hand-to-mouth. It is well established in empirical literature that a fraction of each birth cohort holds no assets: be it financial or illiquid (e.g. real estate). Consuming the entirety of instantaneous income every period does not have to imply low income levels – many individuals with relatively high earnings are characterized by leniency towards reckless, compulsive consumption behaviours. By depriving fraction of the model population of possibility to store any wealth we hope to include this prevalent empirical pattern.
  • Individuals with low financially literacy. Based on the existing literature documenting low financial literacy around the world we model individuals who, due to their insufficient understanding of financial markets, cannot gain interest on income withheld from consumption. These individuals can put money aside, but this stored wealth does not bring them any income in the form of interest earned.

Types of incompletely rational behaviours:


These individuals believe that the economy remains unchanged in the future.

Present bias

Myopic individuals who procrastinate. They make plans, but do not follow through.


Consumption every period equals instantaneous disposable income.

Low financial literacy

Individuals unable to earn any significant interest on their assets.

Main results

1. Incomplete rationality matters for the level of old-age savings.

If the universal pension system is not generous, old-age poverty may be a real threat for many people. Moreover, incomplete rationality is a heterogeneous phenomenon: in some cases insufficient assets can be diagnosed already at the moment of retirement, but for some types of individuals the problem will be related not only to the low amount of wealth at retirement, but also to the excessive rate of de-accumulation. Hence, although each of the types of incomplete rationality results in lower than optimal old-age savings, the challenges for the policy makers are diverse.

2. The type of incomplete rationality matters for policy design.

The four types of incomplete rationality yield two groups of consumers. For the first group, populated by adaptive learners and individuals with low financial literacy, the government intervention requires “fixing” the world, i.e. complementing the existing solutions and instruments with better information and/or better regulation. Indeed, fiscal incentives are redundant. For the second group, fiscal incentives may be indispensable for the instrument to reach the policy objectives. Present biased individuals require additional incentive to save during the accumulation phase, hence fiscal incentives, thought costly, may prove of some merit.

3. Government-subsidized voluntary savings programs should not be evaluated by enrolment.

Often, government programs are judged successful based on the enrolment rates. We show why this is not a useful measure. In fact, homo oeconomicus participates even though such programs reduce welfare for this type of individuals. Other types of individuals – depending on the form of incomplete rationality – typically benefit from the programs and eventually enrol, but for many of them, enrolment occurs only later in their working life. If we measure the fraction of people who participate at any given point, we substantially underestimate the share of people who will benefit from enrolment.

4. The fiscal costs of government-subsidized old-age voluntary savings programs are large.

At first sight, fiscal costs in the form of tax exemptions seem cheap. However, this view does not take into account that individuals who actually have assets and save for their old age will shift their savings from voluntary savings to the instruments provided by the government, which are subject to tax exemption. They may even reduce other savings. The overall crowd-out effect makes fiscal costs of tax exemptions high: the tax base is substantially reduced. A rigorous model, such as ours, provides a useful way to quantify these fiscal costs.

Policy implications

Pension choices are difficult not only because many variables are unknown and finding optimal outcomes requires some serious calculations. The trickiest part is the fact that we make pension choices every single day – typically without fully understanding the long-term consequences. Introducing bounded or incomplete rationality to macroeconomic models of pension systems allows us to evaluate how far off from the fully rational homo oeconomicus we will find ourselves when we retire.

Raising awareness of economic changes helps individuals adjust their choices. Raising financial literacy and providing individuals with easier access to financial markets instruments – such as exchange-traded funds or similar low fee market-based instruments – helps to raise the rate of return to voluntary savings. These improvements are mainly the responsibility of financial institutions, even if they may require regulators to encourage this much needed assistance in consumers’ retirement choices. Tax incentives are necessary only if behaviour involves present bias. Individuals need to be actively encouraged to undertake responsible consumption and savings decisions.

So which policy treatment is most efficient? That diagnosis requires mainstreaming many well-established behavioural economics tools. It is also imperative to develop wealth panel data to be able to observe how individuals and households realize long-run savings plans. Without proper data, any attempt to design government-subsidized old-age voluntary saving schemes is just another policymakers’ blind shot.