1. Introduction
People make decisions that involve risk and time preferences in their day-to-day life. The decisions made are across spectrum of activities and in most cases with some degree of uncertainty. Day-to-day decision-making processes also involve thinking on the margins (
Bangs 2009), when additional benefits are compared to additional costs. An emerging line of thinking in the economic preferences domain relates to the economic and social conditions under which people make economic decisions. The current study attempts to contribute to this emerging body of literature by exploring the association between financial status (a temporary measure of poverty) and economic preferences. Economic and social conditions influence the choices at stake and marginal sizes differently, and for different people.
Haushofer and Fehr (
2014), for instance, stated that economic and social conditions that poor people live in influence their discount rates and risk taking behaviours. This suggests that the differences in risk and time preferences between the poor and the non-poor is not intrinsic. Further, it suggests that preferences are not always permanent but are malleable (
Alan et al. 2020).
Ravallion (
2012) proposed that relevant data in a given context, inclusive of people’s preferences, must inform actions to fight social ills such as poverty.
In order to obtain better measurements of risk and time preferences, economists have turned to experiments, among other methods. The recent past has seen growth in experimental economics literature that focuses on the role of risk preferences (
Andreoni et al. 2020) on life outcomes such as labour outcomes (
Bertrand 2011); holding stocks, occupational choices and smoking (
Dohmen et al. 2011); investment outcomes (
Insler et al. 2016), and socio-economic status or interventions (
Sutter et al. 2019). A similar trend can be observed for time preferences and
Cohen et al. (
2020) provide a detailed survey categorising between studies that use financial flows ‘money earlier or later’ decisions and those that use time-dated consumption or efforts.
The quest to establish the association between financial status (an economic condition) and economic preferences is driven by the zeal to advance knowledge on the views to explain decision-making processes identified as rational choice, pathological, and bounded rationality (
Bertrand et al. 2004;
Mullainathan and Shafir 2013). While rational choices assume that individuals perfectly adapt to the prevailing economic environments and make optimal decisions that are commensurate with conditions at stake, the pathological view argues that there are psychological pathologies that characterise circumstances of scarcity (
Mullainathan and Shafir 2013). Due to the psychological pathologies, the culture of poverty may shape and dominate the preferences of those who are poor and even make them prone to making more mistakes and make decisions that do not maximise their utility. However, boundedness (i.e., bounded rationality, bounded willpower and bounded selfishness) characterises humans’ decision-making processes as humans and not machines (
Thorgeirsson and Kawachi 2013). Investigating associations between economic conditions such as being poor and economic preferences thus becomes crucial in as much as poverty and psychology could be related though not in a simple way (
Haushofer and Fehr 2014). In addition, there are claims of socio-economic status influencing cognition and decision-making (
Sheehy-Skeffington 2020), assertions contrary to the classical economists’ view of rational preference choices (
Bruine de Bruin et al. 2020).
Literature on relationships between economic conditions and economic preferences are scanty.
Lawrance (
1991) reported that poor US households’ discount rates were substantially higher than for rich households reflecting impatience among poverty-stricken families.
Pender (
1996), and
Kassie et al. (
2008) showed a trend where lower wealth predicted higher behaviourally measured discount rates among Ethiopian farm households and South Indians, respectively.
Dohmen et al. (
2011) and
Guiso and Paiella (
2008) reported that wealthier households displayed lower levels of risk aversion suggesting that richer individuals are more willing to assume risk. On the contrary,
Bosch-Domènech and Silvestre (
2006) concluded that non-wealthy students were more risk loving at higher stakes in an experimental study.
Spears (
2011) used smaller budget versus larger budget experiments and found that decision-making under difficult trade-offs is likely to consume scarce cognitive resources, such that smaller budget subjects are impaired in tasks that require willpower and self-control.
Shah et al. (
2015) also observed that decision-making under scarcity—whether the scarcity is temporal, financial or another type, leads to frequent irrational decision-making especially in poverty settings due to attentional capture by salient cues.
The current study aims to establish the relationship between self-reported financial status and economic preferences using experimental data in a developing country setting. The financial situations for many households in developing countries are usually erratic suggesting that the study can shed more light on how individuals make choices under financially stressful or less stressful conditions. Self-reported financial status is a proxy of financial well-being. It is important to know how financial status associates with economic preferences to inform policy meant to address social and economic ills that characterize the poor. In the current study, being ‘very broke’ (and being broke in general) is positively associated with risk aversion and being ‘very broke’ is positively associated with impatience. Thus, the results suggest that in times of financial distress, individuals are averse to risks. ‘Very broke’ subjects exhibit a higher time discount rate on payoffs showing desire to have immediate gratification with a smaller-sooner monetary reward.
The structure of the paper is as follows. The next section provides a description of the experiment. The empirical model section follows. After the empirical model section, next is the results section. Then there is the conclusion and limitations of the study. The data description is provided under the
Appendix A.
3. The Empirical Model
Two models are estimated here using the extended random effects panel probit regression models to explore the relationships between (i) self-reported financial status (FS) variables and risk preferences, and (ii) self-reported financial status (FS) variables and time preferences. The models closely follow
Van Praag’s (
2015) work. The two self-reported financial status variables are coded as binary variables. The variables are coded from the question “how will you describe your financial status?” and the question is answered on a five-point Likert scale with the options “very broke; broke; neither; in good shape, and in very good shape”. The first variable compares those that reported to be very broke (an extreme situation) versus other options. Thus, the variable ‘financial status: very broke’ assumes the value 1 if very broke or 0 otherwise (Models 1 and 2 in
Table 3 and
Table 4). Likewise, the variable ‘financial status: broke’ assumes the value 1 if the options are very broke/broke, or 0 otherwise (Models 3 and 4 in
Table 3 and
Table 4). As this paper attempts to measure association between self-reported financial status and the economic preferences, the variables for financial status (dummies) are entered as dependent variables in our models because they do not vary by a given game of the economic task. In other words, for the four tables (games) of say risk preferences, the participant’s financial status remains constant.
The two main independent variables that are entered in separate models are the risk preference and the time preference. The measure for time preference i.e., (im)patience, is elicited from four task tables for time preference. When Lottery A is chosen in a row (smaller sooner), that is counted as an impatient choice and when Lottery B is chosen (larger later), that is counted as a patient choice. The measure ‘impatience’ is thus a discrete variable ranging from 0 (when no Lottery A is chosen) and 10 (when Lottery A is chosen for all rows in a given task). In other words, the more the number of Lottery A chosen, the more impatient the subject is. For each participant, impatience is measured across the four tasks such that refers to the level of impatience for subject ‘i’ in task table (game) ‘j’.
The measure for risk preference, i.e., risk aversion, the second covariate of direct interest, is obtained from the task tables (games) for risk preference. When Lottery A is chosen in a row in this set of tasks, that is counted as ‘risk averse’ choice and when Lottery B is chosen, that is counted as a risk loving choice. The measure ‘risk aversion’ is, therefore, a discrete variable that increases with the number of Lottery A choices made. So,
refers to the level of risk aversion for subject ‘i’ in task table ‘j’. The four task tables per subject per given economic preference (time or risk) allow for application of panel data analysis in this study and following
Van Praag’s (
2015) panel probit approach.
Van Praag (
2015) argues that contrary to the existing literature (
Cameron and Trivedi 2010) which states that the within estimator can only be determined by regressing on the differences, regressing on the original observation in the extended model, gives the same estimates. Hence, the extended model ensures that a probit model can be used to estimate qualitative panel equations, an appropriate approach in the current study. Following
Van Praag (
2015) closely, both deviations from the mean (Δ risk aversion/Δ impatience) and the averages over time (mean risk aversion/ mean impatience) were included in the models. Finally, financial literacy, amount held in bank account or as cash equivalent (in natural logarithms), gender (female), age, geographical location of subjects (urban) and game (with values 1 to 4) are included as controls in the models.
Studies such as
Bellemare and Shearer (
2010),
Drichoutis and Lusk (
2016),
Eckel and Wilson (
2004) used the number of safe choices while
Bellemare and Shearer (
2010) and
Drichoutis and Lusk (
2016) used the number of impatient choices to measure risk aversion and impatience respectively. Our current study follows same approaches in terms of measurement of economic preferences.
Table 3 outlines the specifications of the estimated random effect panel probit regression models. A Hausman test and a Breusch-Pagan-Lagrangian multiplier test for random effects favours the use of the random effect panel model ahead of the fixed effect panel model.
4. Results
The descriptive results for the study sample are presented in (
Appendix A) of our paper. The
Appendix A section gives a summary of the data in form of some measures of central tendencies and measures of dispersion. The section also provides histograms and correlations for selected key variables. Here, we present and discuss regression results (
Table 4). Models 1 and 2 in
Table 4 show that both impatience (β = 0.075;
p = 0.017) and risk aversion (β = 0.069;
p < 0.05) associates positively with a financial situation of being ‘very broke’. The more the average number of impatient choices that are selected by the participant, the more likely the participant is in a distressed financial status ‘very broke’. Similarly, the more the average number of safe choices that are selected by the participant the more likely the participant will report a distressed financial status. The fact that ‘very broke’ situation statistically significantly associates with both impatience and risk aversion could be revealing the financial constraints faced by the financially distressed participants. These financial constraints necessitate opting for safe and low paying choices and settling for smaller-sooner and immediate gratification in order to mitigate the stressful financial condition faced by the subjects. These results are similar to
Mani et al. (
2013) who reported that people with low income or who are in poverty have a high discount rate to payoffs and are risk averse.
Jachimowicz et al. (
2017) reported that the immediate financial needs could be the reason for impatience and risk aversion in economic experiments.
Model 4 in
Table 4 shows that a financial situation of being ‘broke’ associates positively with risk aversion (β = 0.023;
p = 0.005). This positive association is similar to one reported for the extreme situation of being very broke although the ‘broke’ analysis has a smaller coefficient. The results show that financial situation induces subjects to be more cautious about their risk choices. Our results are similar to
Dohmen et al. (
2011) and
Guiso and Paiella (
2008) who concluded that individuals with low income are more risk averse but contradict results by
Bosch-Domènech and Silvestre (
2006) who found out that low income students reveal risk loving behaviour.
However, there is no significant association between being broke and impatience (model 3). Thus, the extent of financial distress matters when intertemporal decisions are made. In the current study, comparing those on the negative extreme side of the financial scale as measured by self-financial status reporting, shows that impatience matter. However, relaxing this stringent categorisation such that those that are broke (mild negative financial status) are grouped together with the very broke, show that impatience is insignificant. Nevertheless, risk aversion is statistically significant even as we relax the stringent categorisation. There is some consistence with the risk aversion results.
There are also some results for control variables worth discussing here. The study shows in both Models 1 and 2 that the lower the level of financial literacy, the more likely the participant is to report that they are ‘very broke’. However, when applying relaxed categorisation, when financial literacy improves, the likelihood of being broke increases too as shown by Models 3 and 4. As one would expect, participants that reported higher values of money held in the account or as cash were less likely to report that they were in a ‘very broke’ or ‘broke’ financial status. In addition, female participants and urban dwellers were associated with being less likely to report that they are ‘very broke or broke’.
Risk preference and time preference choices play an important role in every person’s life. These preferences influence economic agents’ various activities such as saving, consumption and investment decisions (
Angrisani et al. 2020). Association of these preferences with situations that economic agents find themselves in, is equally crucial.
Angrisani et al. (
2020) emphasised the importance of assessing whether risk preferences are stable individual characteristics that are not affected by economic and social conditions among other factors. The notion that individuals are always rational when they make decisions has gradually lost its truthfulness. In reality, psychological biases, bounded rationality, bounded willpower and bounded selfishness play a vital role in shaping preference choices of individuals. Exploring the relationship between financial status and the time and risk preference choices can provide information on whether biasedness in choices exist or not, under different financial circumstances. Financial status could be related to personality traits and emotions such as fear, loss aversion, self-control, confidence, anger, hope and sadness among other factors (
Aren and Nayman Hamamci 2020). Evidence also show that overconfidence bias and self-control bias are positively related with risk preference while loss aversion bias and regret aversion bias are negatively related with risk preferences (
Ritika and Kishor 2020).
Subjects that are in extreme financial stress (very broke), are substantially risk averse and they are impatient. Their behaviour shows aspects of bounded rationality and bounded willpower. The fact that the option ‘very broke’ was chosen by those subjects who also selected more safe choices reveals that such subjects’ circumstances induced loss aversion bias, a trait of bounded rationality. On the other hand, being present-biased and settling for higher impatient choices resonate well with bounded willpower where subjects value present gratification (smaller-sooner payoffs) at the expense larger-later payoffs. A US study survey on households before and after payday concluded that being liquidity constrained induced present biasedness in the households (
Carvalho et al. 2016). In addition, poor people tend to invest in low return investments and their business ventures are usually small with no economies of scale (
Haushofer and Fehr 2014). For the current experimental subjects, participation in these experiments where they stand to win some money is an opportunity that cannot be spilled. They opt for safe choices in that it guarantees an outcome although it is intermediate in magnitude (not very high or very low: ZAR60 vs ZAR50) as opposed to amounts that are widely apart (e.g., ZAR100 vs. ZAR25). Similarly, for the delayed choices, their financial stress situation associates with choosing immediate and lower payoffs.
Relaxing strict categorisation such that the ‘very broke’ and ‘broke’ subjects are put in one category watered down the impatience attitude. However, the grouped subjects still exhibited a significant risk aversion attitude. The current study’s results also reveal the importance of the relationship between financial status and the economic preferences in support of
Amir et al. (
2020) findings that risk and time preferences are environmentally flexible.
5. Conclusions
This study investigated the association between self-reported financial status and economic preferences in a developing country setting and using data from an incentivised experiment and a survey. Data collection process also included the use of financial literacy assessment task. The study contributes to literature on interlinkages between temporal poverty and decision-making processes under risky and intertemporal situations. The current study has some limitations worth highlighting. First, when subjects self-select into an experiment sample, the sample may characterise subjects with the same characteristics. Subjects that find it necessary to commit their time to that experiment for a certain and common reason. As a result, the sample may have biasedness with respect to that characteristic. In this case, since the experiment involves winning money, there is a possibility that those that are financially in need may show up. Of course, this study sent out an open invitation to all Bachelor of Commerce degree students with the hope to raise the response rate. A proper random sampling strategy may be necessary to minimise this problem. Second, intertemporal choices involving money depend on inflation rate especially as the period gets longer. That would need a good grasp of how interest rates work and the prevailing inflation rates in that context for the participants to establish how worthily it is to wait for a larger payoff. The financial literacy level among our subjects is quite low and so their level of understanding of inflation is questionable. However, the increase of interest rates up to 100% in the last row of the time preference tasks completed in an economy with an average inflation between 3–6% ensured that inflation problem was taken care of. Last, the question “how will you describe your financial situation today?” is quite narrow and so the measurement of self-reported financial status is not strong. Scripts that measure financial status over a reasonable period such as a month and not just for a specific day can be more useful.
Nevertheless, the study contributes to an understanding of how people make decisions which involve some degree of uncertainty and sometimes requiring thinking at the margins and that is particularly important in so far as decisions on investments, education, personal health, and other welfare-enhancing activities have such features. Socio-economic statuses play a critical role in decision-making processes that have long lasting life effects too. In the current study, self-reported financial status acts as a proxy for temporal poverty. The study established a positive association between bad financial status (both very broke and broke) and risk aversion behaviour on one hand, and a positive association between extreme financial stresses (very broke) and impatience. Such a finding illustrates the importance of psychology of poverty on economic preferences and in decision-making in general, even as poverty is temporary as represented by self-reported financial status.