Cross-country variation in time preference

Time preference is a measure of one’s future orientation. It has been linked with long-term individual outcomes (as was shown by Mischel’s marshmallows) and influences macroeconomic outcomes such as the level of saving and investment.

Time preference is not consistent across countries. This was shown in a working paper by Wang, Rieger and Hens, who examined variation in time preference across 45 countries. They found significant cross-country variation in time preference and further, that it could not be explained by differences in interest rates or inflation. Instead, evidence was found for the variation being based on cultural factors, risk preferences (such the degree of loss aversion) and macroeconomic variables. I found the raw results interesting, and a contribution in themselves, but as I note below, it would be good to further tease out the explanations for the variation.

The paper was based on surveys conducted on university students in 45 countries. The first question on time preference was whether the subject would prefer to receive $3,400 now or $3,800 one month into the future (a wait-or-not question). The responses to this question showed a large degree of variation, ranging from 8 per cent of Nigerians who were willing to wait for the $3,800 versus 89 per cent of Germans. Austria, Switzerland, Denmark and Norway also had well over 80 per cent of the survey subjects who choose to wait (amusingly, some students in Norway complained the question was ridiculous as everybody would choose to wait). When the groups were split by cultural groups, Germanic-Nordic and Anglo groups demonstrated the highest level of patience, while the African group was the lowest.

The students were also asked a one-year and ten-year matching question, where they state what sum they would need to receive in one or ten years to be indifferent between that payment and payment of $100 now. The results to these questions showed that most people discounted the near future more than the far future, indicating that the classical consistent approach to risk does not hold. As a result, the authors calculated rates of time preference from the survey questions using the implicit risk and hyperbolic discounting approaches. The implicit risk approach is based on the idea that risk and time are conceptually separate. There is an implicit risk that the delayed outcome will not occur, so people try to avoid delaying positive consequences. Hyperbolic discounting results in people discounting the pay-off  by a large amount for small delays, but with a decreasing level of discounting as the pay-off moves further into the future. The implicit risk approach is mathematically equivalent to a quasi-hyperbolic discounting model, which is a tractable approximation of hyperbolic discounting based on a large initial discount for any delay, and then a constant lower discount rate beyond that point. Despite that mathematical equivalence, the implicit risk approach is typically treated as a rational response, while quasi-hyperbolic discounting is usually based on concepts such as lack of self-control.

Across the sample, the authors calculated a mean value of β (the risk or self-control part of the discounting) of 0.60. This means that for a delay, the mean respondent would value the pay-off at 60 per cent of what they would if they received the pay-off immediately. The United States had a value of 0.77, Japan of 0.70 and Germany of 0.60. The countries with the lowest values were former Eastern-bloc countries such as Georgia, Estonia, Russia and Bosnia and Herzegovina. Splitting the countries into cultural groups, East Europe (β=0.38) and Africa (β=0.43) had the strongest present bias, Anglo cultures the least (0.76) while the others cluster around 0.60. All cultural groups had a roughly similar long-term discount factor of between 0.77 and 0.84 per year, so after the initial impatience or risk response, there is less difference between groups.

The questionnaire also included questions to calculate the students’ degree of risk aversion, largely through lottery based questions, and some questions that sought to measure cultural dimensions such as individualism, uncertainty avoidance and long-term orientation. The authors regressed the waiting tendency results from the first question and present bias and long-term discount rate from the second question against the cultural measures obtained from the survey, their risk preferences and some other factors including age, gender and whether they were an economics major. They found that cultural factors such as individualism and long-term orientation had a significant effect. Higher risk aversion and loss aversion were also found to be statistically significant factors. (Is this link between risk and time preference a causal link, or are the attitudes to risk and rate of time preference subject to some other underlying variable?)

But despite the findings of statistical significance, it is hard to argue that these measures are important. The regression showed that, at best, 10.9 per cent, 12.2 per cent and 4.6 per cent of the variation in waiting tendency, present bias and the long-term discount rate can be explained by these factors combined. Many variables may be statistically significant, but as a predictive measure, they are not important.

Further, for some of the statistically significant findings, I would like to see the potential underlying factors teased out. What is the causal relationship between economic growth and time preference? While the regressions included whether the subject was “native”, immigration could be useful in teasing out the link between time preference and cultural factors, the external environment and inherent characteristics. Even though there are some similarities between the countries included in the survey, there is no substitute for comparing different cultures in the same environment. Immigrants who have been in the country for several generations could be used to examine any inherent characteristics.


  1. It seems to me that the death rate in the respondent’s age group, either the actual local death rate or the rate perceived by the respondent, would strongly influence present bias.

Comments welcome

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s