Intertemporal Preferences during the Covid-19 Pandemic. Analysis of Preference Shocks

Table of Contents

1. Introduction

2. Theoretical Background and Hypotheses

3. Time Preference - Literature Review

4. Method
4.1. Survey design
4.2. Time preference measure
4.3. Risk preference measure
4.4. Measuring affectedness
4.4.1. Level of Affectedness
4.4.2. PANAS
4.5. Participants

5. Results
5.1. Sample Description
5.2. Time preference
5.3. Risk preference
5.4. Level of Affectedness
5.5. PANAS
5.6. Regression Analysis

6. Discussion and Conclusion

7. References

8. Appendix

Abstract

This thesis analyses time preferences in times of the Covid-19 pandemic. In accordance with literature analysing preference shocks such as natural disasters, the hypothesis that a great­er impact of Covid-19 and associated lockdown and after effects on an agent leads to higher impatience is formed. Data from an incentivized online survey with a sample consisting most­ly of students (n = 379) is used. In order to measure the impact of Covid-19 on respondents’ lives, several potential avenues of impact are assessed (including loss of earning opportuni­ties, affective states during severe lockdown restrictions etc.). Respondents indicated wheth­er they had tested positive for Covid-19, whether somebody close to them had to be admitted to hospital due to Covid-19 and indicated their agreement with statements concerning their fear of contracting Covid-19 when they left their homes as well as their uncertainty regarding the future due to a second wave of the virus. Time preferences are assessed using a com­mon incentivised method, called the Monetary-Choice Questionnaire (Kirby et al., 1999). Ad­ditionally, a measure of risk aversion was also collected (Holt & Laury, 2002). Using regres­sion analysis, a significant positive correlation between the affectedness by the virus and discount rates was found. People whose lives are more impacted by Covid-19 displayed higher discount rates. Results for a classification framework that segregates different levels of affectedness as well as a self-reported measure of affectedness point in the same direc­tion.

Keywords

pandemic • time preference • preference shocks • covid-19

1. Introduction

The Covid-19 pandemic has shaped the year 2020 unlike any other major event before. The number of reported Covid-19 cases increased exponentially leading to containment measures from countries all over the world. In Austria, the government imposed a lockdown from 16 March 2020 until step-wise relaxing these measures after Easter and opening up commercial activities and outdoor and indoor meeting places (e.g. restaurants, pubs, parks etc.). Since then, daily routines have started to slowly get back to normal. However, the pan­demic is not over yet, and it is possible that a second wave of the virus will arrive before vac­cinations are distributed on a large scale. In these unstable times, it is interesting to look at how the behaviour of people changes depending on the extent to which they have been af­fected by the pandemic.

When people are confronted with intertemporal decisions, they may be patient enough to wait for a possible higher reward in the future or prefer to choose a smaller reward that ar­rives sooner. Due to increased uncertainty about the future caused by the current pandemic, (see. Ahir, Bloom & Furceri, 2020 and Jain, 2020) agents may rationally choose to receive a smaller certain reward now rather than wait for a larger uncertain reward in the future. In the current situation, the general uncertainty about the future as induced by the pandemic might carry over into decisions even when the future reward is assured, resulting in greater ob­served impatience.

In order to measure time preferences as well as assess the impact of Covid-19 on the lives of subjects, an online survey was conducted. The idea is to understand whether individuals’ time preferences are correlated with the extent to which they were affected by the lockdown and the pandemic in general.

The next part of this thesis deals with the background of major events influencing different preferences and presents literature in which similar analyses was employed. In accordance with existing literature, the main hypothesis is formulated. In the next part I review the litera­ture about the theory of time preferences. Next, I present the method I used to gather rele­vant data to test the hypothesis, which I then test in the Results section. At the end, I discuss the limitations and conclusions of the current study.

2. Theoretical Background and Hypotheses

For a long time, economists have assumed that individual time preferences are a part of one’s character and do not change across contexts. On the contrary, more recent research findings support the idea that these preferences might be influenced by individual experienc­es. (Cassar, Healy, & Von Kessler, 2017)

The focus of standard economics is homo economicus, a perfectly rational agent who max­imises his expected utility, is emotionless and selfish. On the other hand, behavioural eco­nomics focusses on the activities of homo sapiens, who act in a boundedly rational manner, are emotional and do not always behave in a purely self-interested way. Moreover, the pref­erences of homo sapiens may be inconsistent and affected by psychological factors. (Hinvest, Fairchild, & Elkholy, 2018) In the process of decision-making, other people and their wellbeing could be considered as well. These premises allow for preferences that may be influenced by certain events and experiences.

Studies that explored the impact of major events like natural disasters or the financial crisis in 2008 have shown a connection between these exogenous shocks and a change in prefer­ences and behaviour of people. Being affected by such major events might raise awareness of potentially adverse events and life expectancy. Consequently, disasters may have a posi­tive effect on impatience as well as on risk aversion. (Cassar et al., 2017) Moreover, a loss of income might as well be a possible channel through which changes in preferences can be explained. Psychological effects caused by major events include the impact of strong emo­tions, for example fear or sadness, which could affect the cognitive process of decision mak­ing. Furthermore, as recovering from and facing the challenges of major events typically in­volves collaboration between people, the strengthened solidarity might be another channel through which individual’s preferences are influenced.

Future income is associated with some uncertainty by default, but in times of a global pan­demic people may perceive additional risk associated with future payments by virtue of the fact that they happen to take place in a future that is already fraught with uncertainty. The volatility (a term commonly used to indicate the level of risk exposure in financial markets (Hillier, Grinblatt, & Titman, 2011)) of the U.S. stock market has reached levels in 2020 that occurred before only on Black Monday in 1987 and at the start of the Great Depression in 1929. In comparison to other health-related global crises such as the Spanish Flu in the 1920s, Baker et al. (2020) found that news coverage related to Covid-19 played a large role in the daily stock market jumps in 2020. The impact of Covid-19 and its associated contain- ment measures on the economy can be felt instantly on a regional level and has enduring consequences on the worldwide economic situation. The restrictions on commercial activity such as business closures or the lockdown in general, as well as disrupted cross-border supply chains underline the influence of Covid-19 on the economic situation. Apart from the economic impact, health and safety concerns stemming from the severity of the virus, its mortality rate as well as its potential for rapid and uncontrolled spread add to existing levels of uncertainty, especially in the beginning of the pandemic when there was very little infor­mation about the nature of this virus.

The increased level of global uncertainty might impact people’s preferences - The focus of this thesis is its impact on time preferences. It is also well-known in the behavioural literature that individual risk preferences are closely linked to the time preferences (Andersen, Harrison, Lau, & Rutström, 2008), (Andreoni & Sprenger, 2012b). People with higher risk aversion might be more impatient, meaning that they would prefer a sooner, smaller sure payment over a later, but larger payment that might be more uncertain. Consequently, in or­der to understand the differences between the time preferences of people, it is necessary to control for individual risk preferences. (Kuralbayeva, Molnar, Rondinelli, & Wong, 2019)

Formally, the effect of increasing future uncertainty on discount rates has been analysed by Halevy (2008). He suggests that discount rates might be influenced by the certainty of the present in comparison to the uncertainty of a delayed consequence. Increasing uncertainty of the future would potentially increase discount rates.

Several studies have previously drawn a link between the macroeconomic situation and indi­vidual preferences. Krupka and Stephens Jr. (2013) used panel data from US households in the 1970s and found that higher measured discount rates correlate with an increased infla­tion rate. The variation in discount rates over time supports the idea that elicited discount rates from experiments reveal the market interest rate that the person faces, which, in turn, can be influenced by the current macroeconomic situation.

Hong and Hanna (2016) analysed the mean financial planning horizon as an indicator for households time preferences during a time-frame that included the Great Recession of 2008. They found a significant drop after 2007, meaning that the financial planning horizon re­duced, and households concentrated more on the more recent future.

Economic anxiety, which is related to risk factors such as those caused by a pandemic, shapes households economic decisions. Beliefs about the severity of the pandemic are linked to economic worries about the public as well as personal economic situations. By us­ing Google search intensity data for topics related to economic anxiety, Fetzer et al. (2020) found increased intensity both at the arrival of Covid-19 in a country and during the later spread of the virus. In their study, data about the topics "Recession” and "Stock Market Crash” was used for several countries around the world.

Natural disasters are also common sources of preference shocks. Cassar et al. (2017) ana­lysed preferences in rural Thailand after a tsunami in 2004 and found higher impatience for people that were more affected by the tsunami. In order to determine the level of affected­ness by the tsunami, data about the level of impact on the respondent’s village, financial damage for the household and health implications were used. In particular, subjects that suf­fered financial damage due to the tsunami were found to be more impatient.

Similarly, Kuralbayeva et al. (2019) found increased impatience in a sample of Italian house­holds that felt the shake of an earthquake in 2009. Moreover, these households showed in­creased consumption and decreased savings three years after the earthquake, which sup­ports the idea that such exogenous shocks impact the behaviour of people over a longer timeframe.

Callen (2015) studied the time preferences of people affected by the Indian Ocean Earth­quake in 2004. In contrast to the studies above, lower impatience was found for people that were more affected. The extent to which people were affected by the earthquake was deter­mined using data of the respondents place of residence and data of the high-water level fol­lowing the earthquake. This approach was necessary, as there was no pre-event data on time preferences available.

Based on the literature reviewed above, I test the following hypothesis:

H1: People that are more affected by the covid-19 pandemic display higher individ­ual discount rates than people that are less affected.

3. Time Preference - Literature Review

When facing intertemporal choices, individuals often have to choose between smaller out­comes that occur immediately or in the near future and higher outcomes that occur later. Ex­amples can be found in various fields. Not only economic decisions such as consuming to­day rather than saving for the retirement, but also individual decisions regarding education or public decisions concerning public investment and climate change can be regarded as inter­temporal choices.

The choices of individuals in such cases is determined by their utility functions. These utility functions are shaped by the individual preferences concerning goods, time and uncertainty. If there is no alternative good to choose from, the preference regarding goods can be neglect­ed which allows to focus on time- and uncertainty (risk) preferences. (Andersen et al., 2008)

According to Fisher (1930), whether an individual acts patiently or impatiently can further be explained by four characteristics of the income stream (not necessarily monetary income) associated with the respective decision. First, the size of the income stream. The preference for present over future income may be stronger the smaller the individual’s income at the moment. An income stream that is large in relation to the individual’s income now, may be valued differently than a proportionally small income stream. The second characteristic to be considered is the expected distribution of the stream of income across time, in other words the time shape. Valuing the income stream not only depends on the individual’s income at the moment, but also on its expected income at a later point in time. 500 € will be highly ap­preciated by a student, but the same amount might be valued differently by the same person at the height of his professional career. Accordingly, a choice between a small amount now or a larger amount later will be influenced by the expected appreciation of the larger amount in the future. Thirdly, the composition of the income stream, that is, the valuation of the dif­ferent elements that are part of the income stream. As mentioned above, if there is no varia­tion in the composition of the income stream, for example if the decision is between two monetary amounts, then this preference regarding goods or elements can be neglected.

The fourth characteristic is the influence of risk. Risk regarding the certainty of income streams can have different effects, depending on the distribution of risk in time. Usually, the distant future is more uncertain than the immediate future, as the possibility of unexpected circumstances happening, like accidents or diseases, is higher the more time passes. De­pending on preferences of the individual, this can have the effect that the uncertainty keeps the impatience down, as the need to provide for future risks might impact present behaviour.

However, if the uncertainty in the immediate future is high, perhaps even higher than in the distant future, the effect might be reversed. Higher uncertainty in the immediate future can be caused by wars, strikes or financial crisis, which are all believed to be temporary. Such cir­cumstances increase impatience.

In order to measure time preferences, researchers have come up with several measures that can be summarized into two categories: matching-based measures and choice-based measures. Matching-based measures ask the participants about an amount that they would set equal to a given amount at a different point in time. Such measures only require a single response of the participant in order to calculate a discount rate for the specified time period, but they are more cognitively challenging for the participants. Choice measures use a series of binary intertemporal choices to find a switching point, where the participant switches from receiving an earlier outcome to receiving a later outcome. These measures need more than one response to identify the discount rate, but are easier to understand for the participants. (Urminsky & Zauberman, 2015) The goal of these measures is to find a point at which two outcomes, one sooner outcome and one discounted later outcome, have approximately the same value for the individual. At this indifference point, the individual discount rate can be calculated. (Odum, 2011)

Moreover, other, more complex measures have also been developed to use in lab studies. As time preferences are not necessarily measured with a monetary perspective, Augenblick et al. (2015) proposed a design that focusses on allocating units of effort to different tasks across time. Here, time preference is measured over consumption (effort to be put into a task) instead of intertemporal monetary choices. Another method that uses monetary choices is the Convex Time Budget method (Andreoni & Sprenger, 2012a), that lets subjects allocate tokens out of their budget to sooner or later payments representing convex choice sets. However, as this thesis uses a one-shot online survey, a widely used choice-based measure was used to determine time preferences.

In an economic sense, modelling the individual time preference can largely be ascribed to the discounted utility (DU) model, developed by Samuelson (1937). Behind the DU model lies the assumption that individuals strive to maximize the sum of the future outcomes by comparing their discounted present values. Following this model, individuals discount future outcomes with a constant rate over time. This discount rate is meant to compress all factors that have an influence on intertemporal decisions.

Abbildung in dieser Leseprobe nicht enthalten

Equation 1: Exponential Discounting (Original DU Model) (Green & Myerson, 1996, p. 497)

Here, V describes the present (discounted) value of an amount A that is available with a de­lay of D units of time. The parameter k is the discount rate, it describes the rate at which the value declines with a delay in time. The larger the parameter k, the steeper the discounting, meaning that the individual discounts future rewards more impatiently. On the other hand, a lower k corresponds to a more gradual discounting.

Over time, several studies have shown that this simplified model is not always accurate. Frederick et al. (2002) summarized potential anomalies. One pattern that was found to be an anomaly of the DU model is the so-called “hyperbolic discounting” or "present bias”. This means that the individual discount rate is declining over time, and not constant as the model suggests. Moreover, empirical evidence suggests that earnings are discounted more than losses. This is also called the “Sign Effect”. It was also found that small amounts get dis­counted more than large amounts (“Magnitude Effect”) and that choices over different se­quences of outcomes regularly lead to different results as the model suggests.

Subsequently, other models of discounting have been developed. Hyperbolic functions have been shown to describe individual’s discounting curves accurately. (Laibson et al., 1998)

Abbildung in dieser Leseprobe nicht enthalten

Equation 2: Hyperbolic Discounting (Mazur, 1987, p. 58)

In contrast to exponential discounting, the hyperbolic function accounts for declining discount rates as the amount A is delayed further in time (D increases). The function initially has a steeper decreasing course but flattens out as D gets larger. This property of the hyperbolic function is useful for factoring in the present bias anomaly.

Adding to the economic view, time preferences can also be analysed with focus on psycho­logical and neuroscientific perspectives. Not only do character traits and cognitive abilities play a role in situations where individuals face intertemporal choices, but also emotions. (Urminsky & Zauberman, 2015)

In addition to characteristics of the income stream, Fisher (1930) found personal characteris­tics that influence impatience decisions, which include foresight, self-control, expectation of life and concern for the lives of other people. Foresight means the ability to evaluate the importance of future needs. A lack of foresight can have two forms: Either the person under­states or overstates the future needs. Understating the future while concentrating on imme­diate needs corresponds to high impatience, while overstating the future can be associated with low impatience. Adding to foresight, the concern for the lives of other people, for exam­ple the own children that will one day inherit the income, also contributes to the individual time preference. Low self-control is associated with impulsive behaviour, which is an indica­tor for high impatience. Another characteristic is the individual life expectation. Intuitively, older people might be more impatient, since their expected remaining life is shorter, and the chance of death is a rational factor that affects impatience. Controlling for age to separate age effects on time preferences from other effects is therefore necessary, although this age effect might not be that present in a sample of (mostly) students.

Impulsive, impatient behaviour has been analysed in the field of psychology and found to be displayed by people while affected by visceral factors (Loewenstein, 1996). Visceral factors include emotions such as anger or fear, which positively impact affinity to impatient behav­iour. As described above, such strong emotions are possibly evoked by exogenous shocks.

Another possible factor that influences time preferences is happiness, which is thought to decrease impatience. A possible explanatory mechanism is that positive affect increases cognitive flexibility and therefore raises the attention that the individual gives to the evalua­tion of all options. Subsequently, more thought might be put into comparing net benefits of the options related to the intertemporal choice, rather than intuitively deciding that the earlier option is more beneficial. (Ifcher & Zarghamee, 2011) However, other research suggests the opposite effect, meaning that positive affect might lead to not elaborating the tasks in a com­prehensive way as a result of avoiding cognitive effort which might interfere with their will to maintain their positive emotional state. (Bless, Bohner, Schwarz, & Strack, 1990) and (Isen, 1987) For the regression analysis, self-reported happiness at the time that the survey was taken will therefore be included to control for possible effects of the happy emotional state of the participant on the observed impatience.

4. Method

4.1. Survey design

In order to determine individual time- and risk preferences, a short experiment as a part of an online survey was used to observe the behaviour of people when confronted with decisions that involve intertemporal choices (time preferences) and lotteries (risk preferences).

In addition, a questionnaire for measuring positive and negative affect has been added to the survey. The participants also answered several questions regarding their experiences with Covid-19 and with the associated containment measures in order to understand to which extent they had been affected by the virus. These questions included, for example, whether the participant or somebody close to them had been tested positive for Covid-19, whether they experienced a loss of or gain in income and how much they agreed with the statement that they were afraid of contracting Covid-19 when they left their homes. The whole survey can be found in the Appendix (A.6). At the end, I also asked them the usual demographic questions (age, sex, income, level of education, employment status).

The survey was designed using the online software Qualtrics (Qualtrics, 2020). Participants were informed that after the survey was closed, one of them would be randomly chosen and paid according to his choices in the time- and risk preference tasks. Thus, once selected, one of the participant’s choices that were part of the time preference task and one that was part of the risk preference task would be randomly selected. In this way, I was able to ensure that the survey and specifically the measures of individual preferences were appropriately incentivised. The budget for incentivizing the survey was provided by the Institute for Markets and Strategy (IMS) of the Vienna University of Economics and Business (WU). Participants were asked whether they wanted to participate in the random draw, and if so, to provide con­tact details. 399 out of 401 participants provided their contact details and entered the random draw.

4.2. Time preference measure

To measure the individual time preference, an incentivised measure commonly known as Monetary-Choice Questionnaire was used. Subjects faced nine intertemporal choices in the form of "Would you prefer to receive 54 Euro in 7 days or 55 Euro in 124 days?”. Monetary amounts, as well as the delay to the later payment varied between the nine choices. The amounts as well as the order of presentation were adapted from Kirby (1999). A front-end delay of 7 days was added in order to ensure that the participant that was randomly selected to be paid would receive the payoff according to her/his choices in the survey. The original measure consists of three delayed reward categories. For this study, only the medium re­ward category has been chosen.

The discount rate associated with the choices was calculated using hyperbolic discounting. Rearranging Equation 2 leads to:

Abbildung in dieser Leseprobe nicht enthalten

Equation 3: Discount Rate for Hyperbolic Discounting

Where A corresponds to the Later Reward, V to the Sooner Reward and D to the Delay in days. The discount rate k indicates the rate which would make an individual indifferent be­tween the sooner and the later reward, meaning the discounted later reward would be equal to the sooner reward.

Abbildung in dieser Leseprobe nicht enthalten

Table 1: Monetary-Choice Questionnaire (Kirby et al., 1999, p. 81) Reward values and the delay to the later re­ward of the 9 choices. The associated discount rate (k) corresponds to the discount rate that would set the sooner and the later reward equal (see Equation 2), Note: The position of the sooner / later reward answer options were randomised in order to avoid position bias

The decisions of the measure are designed in a way that participants would prefer the soon­er reward up to a certain point, after which the (discounted) later reward is more preferred. The point at which they switch from preferring the sooner to the later reward, their "Switching Point”, can be used to calculate their individual discount rate. This individual discount rate indicates the rate at which the subject discounts future rewards. If the later reward, discount­ed with the individual discount rate, equals the sooner reward, the participant would be indif­ferent between choosing the sooner or later option. For example, if the participant prefers the sooner option for the first four decisions, but the later option for following five options, his switching point would be the fifth decision. His switching point indicates that his individual discount rate lies in the interval between the discount rate associated with the last decision 10

where he chose the sooner reward and the first decision where he chose the later reward. His individual discount rate is then determined by calculating the geometric mean of these two discount rates. The geometric mean is used in order to prevent underweighting the lower bound rate. (Kirby et al., 1999)

Participants that do not switch between the sooner and the later option reveal that they are either very patient (always choosing the later reward) or very impatient (always choosing the sooner reward). Their individual discount rates are assigned by taking either the highest (for very impatient subjects) or lowest discount rate associated with the decisions. This is due to the fact that the lower bound for the discount rate of a very impatient subject would be great­er than or equal to the highest possible discount rate associated with the decisions. The par­ticipants real individual discount rate would be higher than the one offered, so the best ap­proximation available is taking the maximum discount rate presented in the decisions. Ac­cordingly, a participant that chose the sooner reward in every decision would be assigned a discount rate of 0.25. The same principle but reversed is also used for very patient subjects. They would be assigned a discount rate of 0.00016.

Some participants would display not one but more switching points, which indicates their in­difference across a certain interval. For them, the bounds that depict the interval which con­tains the real individual discount rate are broadened. These bounds are defined by the first row at which the participant switched, and the last row at which he switched. (Andersen, Harrison, Lau, & Rutström, 2006)

4.3. Risk preference measure

The risk preference measure was the same as that used in Holt and Laury (2002). Subjects faced ten decisions between two lotteries, Option A and Option B. The lottery in Option A provided a payoff of 2 Euro or 1.60 Euro with varying probabilities, while the lottery in Option B provided a payoff of 3.85 Euro or 0.10 Euro with the same probabilities as Option A. These probabilities changed from row to row, starting with 0.10/0.90.

[...]