**3. Results**

Figure 9 shows the estimation results of Equation (1), which includes all samples, Subsample 1 (Attachment dummy=1), and Subsample 2 (Attachment dummy=0). In these figures, monthly material consumption expenditure (unit: US\$) is shown on the horizontal axis, while life satisfaction, which was standardized from 0 to 1, is shown on the vertical axis. The solid line is the estimated SWB function curve, while the dotted lines represent 95% confidence intervals. The "0" on the vertical axis denotes the samples' average life satisfaction. As shown in Table 7, all estimated nonparametric functions are statistically significant. We found a monotonous decreasing trend for the pooled overall sample in the upper panel of Figure 9. For a robustness check, we applied additional estimation using Subsamples A (men) and B (women). The estimation results for Subsamples were almost the same as our main result using the overall sample in Figure 9. For Subsample 1, whose attachment dummy equals 1, although the confidence interval is relatively wide due to the relatively small sample, meaning the estimation result must be interpreted with caution, we found an increasing trend from US\$138 to US\$468. For Subsample 1, the upper and lower confidential intervals were both more than 0 from US\$138 to US\$468 with regard to the vertical axis, which corresponds to the situation where the predicted contribution to life satisfaction is larger than its sample mean. The predicted contribution to life satisfaction of Subsample 1 was 0.101 for the samples' average monthly household material consumption (US\$179). Although the confidence interval is too wide to interpret, we found a decreasing trend over US\$468. For Subsample 2, whose attachment dummy equals 0, we found a decreasing trend.

Table 8 reports the parameter estimates for control variables. We found the expected signs for most variables, with the exception of age, gender, and marriage. Regarding age, we found an inverted-U relationship between age and life satisfaction for Subsample 2, which is not in line with the literature. The exceptions may indicate unique characteristics for rural developing countries. The negative coefficients of marriage and the number of family members can be interpreted as an increase in economic burdens in budget constraints. Furthermore, the absolute value of the negative coefficient of the number of family members for Subsample 1 is larger than that for Subsample 2. This might imply that those who have an attachment to material goods tend to have a stronger scarcity consciousness of income. There is a possibility that those who have an attachment to material goods tend also to have an attachment to family or importance of family and, thus, have higher standards of expenses for each family member, which increases the sense of income scarcity. Another possibility is that Subsample 1 is attached to material goods, so much so that the expenditures on a marginal increase of family members would be more painful than Subsample 2. The positive coefficients of residual consumption imply that non-material consumption is positively correlated with life satisfaction, which is in line with [50].

Subsample 1 (Attachment dummy = 1, N = 892); Subsample 2 (Attachment dummy = 0, N = 2689)

**Figure 9.** Material consumption (US\$) and life satisfaction in Vietnam.



Note: \*\*\* and \* denote statistical significance at the 1% and 10% levels, respectively.


**Table 8.** Parametric estimation results for control variables.

Note: Standardized coefficients are shown. Standard errors are in parenthesis. \*\*\*, \*\*, and \* denote statistical significance at the 1%, 5%, and 10% levels, respectively.
