6. Experiments on Risk Attitude: The Case of Chinese Students

Another response to this criticism is to use the results of TV shows that pay huge prizes. Fullenkamp et al. (2003) and Beetsma and Schotman (2001) reported that people are risk averse, while Metrick (1995) does not reject the proposition that they are risk neutral.

This figure is based on responses to our questionnaire from subjects of the experiment done in Shanghai and Tokyo. We asked about the cost of living per month.

Kachelmeier and Shehata (1992) is a notable exception. They paid Chinese students monetary rewards three times their monthly revenue.

Knetsch and Sinden (1984) is a notable exception. Their TEST3 consisted of selling and buying experiments which ask similar questions to ours. However, they are different from ours, in that different subjects are used for the selling and buying experiments.

One subject felt unwell and left after the selling experiment was completed, so the number of the subjects for the buying experiment was 29.

According to teachers and students at Fudan University, the students there study until around 10 pm every day, so the evening experiment was not a burden to them. We conducted the experiment in the evening, because it was difficult to recruit students during the daytime, as they were expected to attend classes.

This is necessary because in the selling experiment, subjects were given 20 lottery tickets with the expected payoff of 10,000 points. Furthermore, subjects would have been too embarrassed to buy a lottery ticket if they had no points at the outset.

In addition to payoffs, subjects received a 120 yuan (US$14) participation fee.

Note that Kachelmeier and Shehata (1992) show the certainty-equivalent ratio, which is equivalent to 1-TP.

They do not show the confidence interval, so we cannot evaluate the significance of their results.

According to a teacher at Fudan University, most Chinese students’ living expenses per month should be under or around 1,000 yuan (US$120). If this is true, our prize is tantamount to more than half their monthly living expenses.

These results are the same as those from Japanese experiments. See Tsutsui et al. (2005).

We would appreciate a comment by a referee on this point.

The results by OLS are almost the same as those in Table 6.5.

The sample size was 260, so the correlation coefficients were significant (5 % critical value is 0.12). Hirata et al. (2006) report that the correlation coefficient between time discount rates of parents and their children is around 0.2, while the correlation coefficient between random pairs who have no relationship is zero.

It cannot be denied that the questions on the subjects’ parents, RISK, ASSETS, and FINASSETS may not be answered correctly, so they may suffer from a measurement error. Nonetheless, the coefficient of AVARA is immune from such a measurement error, so its coefficient is not biased. As for the coefficients of ASSETS and FINASSETS, they may be biased if the measurement errors of RISK and ASSETS (FINASSETS) are correlated. However, because RISK is defined as the ratio of risky asset to FINASSETS, and because measurement errors of risky asset and FINASSETS are thought to be correlated, the measurement error of RISK is probably independent of that of FINASSETS. In addition, if we look at correlations between subjects’ own wealth (income, food consumption, expenditure per month, etc.) and their households’ wealth (annual income, real estate, financial assets, etc.) in the responses to some questions of the questionnaire, nine correlations out of 21 are significantly positive. This fact suggests that subjects may have adequate information about their parents’ wealth and those variables are somewhat reliable.

Friend and Blume (1975) themselves reported that risky asset share positively correlates to financial assets. However, when wealth is defined as the total of financial and real assets, they found a negative correlation.

Therefore, the low correlation mentioned below is not due to the difference in methods (experiment and questionnaire).

Only one case is significantly positive out of 21 cases between items in the questionnaire, which is the correlation between the most similar questions (Q13 and Q15).

Whether the prizes would be really paid or not for each round was written in the instruction and was announced carefully. The prizes were announced to pay for eight out of 12 rounds.

This addendum has been newly written for this book chapter.

For the details of the experimental results, see Ohtake and Tsutsui (2012).

The similar experiments conducted in Waseda University are reported in Hiruma and Tsutsui (2005).