10. Capability, Social Welfare

Transcription

1 Capability, Social Welfare and Life Expectancy Capability, Social Welfare and Life Expectancy: Evidence from Taiwan Chu-Chia Lin and Chen-Yuan Deng * I. Introduction Traditionally, economists often use the availability of utilities or income to estimate the well being of an economy. However, utilitarianism was criticized by many scholars, such as Sen (1977, 1984, 1985a,b, 1987a,b) and Dreze and Sen(1989), because utilitarianism has information limitations in estimating well-being and we usually do not know a person's second and higher order desire exactly. Indeed, though income has an obvious influence on one's life, we cannot judge a person's quality of life only by his/her income since well-being does not mean only income. Sen's capability approach is a completely different concept of development. The capability approach emphasizes people's ability in living in a society. Presumably, as a society develops, its people should have a higher capability of doing things that he/she desires to do. In this sense, capability is a better proxy for indicating development of a society. * Professor and Ph.D. Candidate respectively, Department of Economics, National Chengchi University, Taipei, Taiwan. nccut001@nccu.edu.tw

2 192 Taiwan, East Asia and Copenhagen Commitment Therefore, Sen suggested that governments should pay more attention on social services and education that could improve people's capability, but may or may not directly relate to income. This paper applies a time-series data set of Taiwan from 1917 to 1997 to investigate the function of income, social welfare spending, and education in improving capability. Since capability itself is unobservable, we use life expectancy as a proxy variable because of high correlation. The structure of this study is: In Section Two, we discuss capability approach, and the problems of utilitarianism and income in estimating well-being. In Section Three we form a simple model and describe basic data in Taiwan from 1917 to Section Four presents regression results showing the relationship of life expectancy and some important variables, such as gross domestic product (GDP), education, and social welfare spending. Conclusions are made in Section Five. II. Capability Approach Before the emergence of the capability approach, utilitarianism played an important role in traditional welfare economy in estimating economic development. However, utilitarianism is criticized by Sen for several reasons. First, it is difficult to know a person's second or higher order desires. 1 Second, utilitarianism has information limitation in evaluating the quality of life 2. The capability approach views the standard of living as a combination of various doings and beings, and a person's capability is a combination of functions the person can achieve. Functions are various things a person manages to do or be. When one's realized functions increase, or the capability to achieve valuable The second order desire is a person's desire to desire. Utility, the only valuable thing in utilitarianism, is defined as desire fulfillment or preference satisfaction. But for a deprived person living in the lack or poor, he or she will be satisfied with small mercies. The process of training a hopeless person to be satisfied easily is called psychological adjustment to deprivation. Therefore it is inadequate for a deprived person if we evaluate his/her well-being by the utilities that he or she has. 2 One of the components of utilitarianism--welfarism requires that the well-being of a person is completely judged by the utilities he or she has, and the non-utility information, such as health, education, and liberties) does not count for it.

3 Capability, Social Welfare and Life Expectancy 193 functions increases, the capability approach contends the person is better than before and his/her well-being is improved. For instance, longevity and literacy are some valuable functions that are considered to be important by Sen in comparing the quality of life among people. 3 Sen believes that longevity, or life expectancy, is a very important function. A person has to be alive to be able to do other things. Therefore, it is not only a valuable function, but also a necessity for other functions to be realized. A person who lives in danger caused by illness or poverty cannot have much energy to improve his/her quality of life. In fact, life expectancy has an important role in deciding the quality of life and economic development of a nation. 4 Income has been a major indicator in evaluating the economic development of a nation or a person, and issues of income distribution have been important since 1970s. However, there is still some important non-income information that significantly affects quality of life. Many empirical studies using the capability approach as an indicator for well-being have shown that well-being does not always correspond with income, and thus income is not the most important factor which determines one s well-being. 5 Therefore, supporters of the capability approach suggest that one should pay more attention to education, public health, and social welfare, which may or may not directly contribute to personal income. Dreze and Sen (1989) argued that there are two different types of economic development, namely support-led and growth-mediated. The former does not promote high economic growth and the resulting longer life expectancy result mainly from increases in social services and education. On the other hand, the growth-mediated process is created by fast economic growth. Therefore, government spending on social welfare has less effect on development. The fast-growing Asian 3 Since we can avoid comparing the feelings among people when we compare these functioning's, the distortion caused by psychological adjustment to deprivation no longer exists. 4 Adelman and Morris(1973), Grant (1978), Morris(1979), Kakwani (1981), UNDP(1990, 1995), Fogel(1991) are some examples. 5 See, for instance, Bhalla and Glewwe (1986), Anand and Kanbur(1991), and Sen(1998). Anand and Ravallion (1993) have shown that public health spending has a larger effect than income in improving economic development. Laderchi (1997) even found that income does not influence capability significantly

4 194 Taiwan, East Asia and Copenhagen Commitment economies are some examples of growth-mediated economic development. Generally speaking, capability is determined by the size of per capita income, education, government's spending on social welfare, and some other important factors. More specifically, according to the growth-mediated development hypothesis, capability is positively correlated with income since a rich person could afford to do more economic and non-economic activities. For the same reason, a person with higher education is more capable than the one with less education. Finally, government spending on social welfare has also a positive effect on people's capability. Furthermore, according to Dreze and Sen s (1989) support-led development hypothesis, the effect of government social welfare spending on capability is more significant when a country is poor since people have less income and need more help from the government. III. Data Description In this study we apply a time-series data set of Taiwan to test whether the capability hypothesis is supported by the development experiences of Taiwan. Though Taiwan has received significant international attention for its economic success during the past 50 years, this achievement is mainly in economic terms, such as a high growth rate of income, low inflation, and wide income distribution. 6 However, the study on economic development in terms of capability is rare. Therefore, it is interesting to see if the capability approach holds in an economy with a high growth rate, as in the case of Taiwan. Restricted by the available data, the time-series applied in this paper is from 1917 to Moreover, the data is divided into two time spans, one is from 1917 to 1943, and the other is from 1951 to Since it 6 For instance, the average growth rate of real per capita income was 6.30% from 1951 to 1997, while the GDP deflator grew 5.33% per year in the same period of time. 7 Taiwan had been colonized by Japan before 1945 for more than fifty years. After WWII, Taiwan was handed back to KMT's (Kuo-Ming-Tung, the ruling party) China in After losing in the civil war, KMT retreated from Mainland China to Taiwan at the year of Most provincial level statistics of Taiwan between 1945 to 1950 were missing owing to the civil war and government switch. Meanwhile, Barclay (1954) is a good reference on Taiwan's colonial development.

5 Capability, Social Welfare and Life Expectancy 195 may have a significant structural change for the relationship between capability and key variables in this study after 1951, we apply a dummy variable to capture the effect of structural change on our key variables. Definitions and sources of variables applied in this study are described below: D51: A dummy variable. D51: 1 if the time span is from 1951 to 1997; while D51 is 0 if it is between 1917 to EDU: Percentage of six-year olds enrolled in school. The enrollment rate from 1917 to 1943 is from The Statistic Abstract of Taiwan Province in the past 51 years, published by the Statistics Bureau of the Executive Office of Taiwan, 1946, p The enrollment rate from 1951 to 1997 is from The Statistic Yearbook of Taiwan, 1998, pp , published by the Department of Statistics, Taiwan Province. GDP: Real per capita GDP in NT dollars with 1991 as the base year. The per capita GDP and the GDP deflator between 1917 to 1943 is from Wu(1991), pp , Appendix 2. The income data and the GDP deflator between 1951 and 1997 is from National Income Statistical Yearbook, 1998, p.35, published by DGBAS, Executive Yuan, ROC. LIFE: Years of life expectancy. The data is from The Simple Life Table of Taiwan Province, 1999, pp , published by the Department of Statistics, Taiwan Province. SOC: Real government social welfare spending per capita in NT dollars with 1991 as the base year. The data between 1917 and 1943 is from The Statistic Abstract of Taiwan Province in the Past Fifty-one Years, published by the Statistics Bureau of The Executive Office of Taiwan, 1946, p.336 and p.348. The data between 1951 and 1997 is from The Fiscal Statistic Yearbook, 1998, p.25, published by Ministry of Finance, Executive Yuan, ROC. Life expectancy is the proxy variable for capability in this study since to live itself is a good signal of capability. Moreover, according to the capability approach, we expected GDP, EDU, and SOC all have a positive effect on life expectancy. The model could be written as follows: LIFE = f (GDP, EDU, SOC) Furthermore, during the time period of 1917 to 1943 when the

6 196 Taiwan, East Asia and Copenhagen Commitment average income in Taiwan was low, the effect of government spending on social welfare (SOC) should have a larger effect on life expectancy (LIFE) according to the support-led development. On the other hand, during the time period of 1951 to 1997 when the average income in Taiwan was much higher, the effect of SOC on LIFE should be smaller according to the growth-mediated development hypothesis. To test the effect of structural change before 1943 and after 1951, we apply a dummy variable to test both the change of intercept on the regression and the change of slopes for all independent variables in the regression model. IV. The Evidence of Taiwan The male life expectancy in Taiwan in 1917 was only years and it grew to years by 1997, see Table 1. The annual growth rate of life expectancy was 0.87 percent, while the growth rate was higher for the first period of 1917 to 1943 at an annual rate of 1.23%, while it was only 0.67% for the second period from 1951 to The female life expectancy in Taiwan is longer than male. For instance, the female life expectancy in 1917 was years, and it was years in Moreover, the annual growth rate for female life expectancy was also higher (1.02 percent) than that of males (0.87 percent). However, the discrepancy of the growth rate between the life expectancy for males and females was greater during the first time period (1.23 percent vs. 1.63percent), while the difference was much smaller during the second period ( percent vs percent). Taiwan's real per capita income in 1917 was NT$30,374, while it grew to NT$31,392 in 1943 at a very slow rate, 0.30 percent per year, see Table 1. Per capita income in Taiwan had dropped from NT$30,374 to NT$19,807 from 1943 to 1951 caused by the civil war. Since 1951 when Taiwan experienced a remarkable economic performance, the real growth rate is 6.30 percent per year from 1951 to In 1997, the real per capita income in Taiwan was NT$322,700. The economic performances for these two periods of time were quite different, though the growth rate of government spending on social welfare followed a similar pattern. For example, the annual growth rate of

7 Capability, Social Welfare and Life Expectancy 197 government spending on social welfare between 1917 and 1943 was percent, while the rate was percent between 1951 and Table 1 Basic Statistics Year LIFE, male(year) LIFE, female(year) GDP per capita(nt$) 30,374 31,392 19, ,700 SOC(NT$) ,585 EDU, male(%) EDU, female(%) Annual Growth Rate LIFE, male(%) LIFE, female(%) GDP per capita(%) SOC(%) EDU, male(%) EDU, female(%) Notes: (a)the definitions of variables are as follow: LIFE: Life expectancy; GDP: Real GDP per capita, 1991 is the base year; SOC: Real government social welfare spending per capita, 1991 is the base year; EDU: School enrollment rate for six-year-old children. Finally, the school enrollment rates for six-year-old male and female children were only percent and 3.71 percent in 1917 and quickly grew to percent in Apparently, even under Japanese rule, Taiwanese people still paid a lot of attention to their children s education. During the second time period, the school enrollment rate quickly reached its peak in 1980s. 8 In order to run regression for LIFE on GDP, EDU, and SOC, we have made a few arrangements: First, all variables are taken in logarithmic values to increase its fitness on the data set. 9 Furthermore, following Anand and Ravallion (1993) we use -log 8 In fact, school enrollment rate for six-year-old kids in Taiwan had reached 99.99% in the year In fact, we have also tried some regressions without taking log values. We found that the estimated results are similar in terms of signs and significance of estimated coefficients, while the adjusted R 2 are a little lower for equations without taking log values.

8 198 Taiwan, East Asia and Copenhagen Commitment (80-LIFE) as the dependent variable instead of log(life). 10 Finally, to avoid the autoregression problem, which exists in certain time-series data set, we apply an atuoregression adjustment to eliminate possible bias. 11 Applying the ordinary least-squares method in our regressions with the data set from Taiwan, the estimated results with all samples are shown in Table 2. In Model (A) of Table 2, where only income (GDP) and the dummy variable (D51) are used as independent variables, the coefficient of GDP is significantly different from zero. Most importantly, the coefficient of GDP was 0.48 and similar to Anand and Ravallion s (1993) result. 12 Moreover, the adjusted R2s were as high as even when we have used only two explanatory variables, GDP and D51. The above results have provided us strong confidence in our data set. In Model (B) of Table 2, we put all relevant independent variables, including GDP, SOC, EDU, and D51 into the regression. Again, we found that all coefficients have significant and correct signs as expected. In other words, GDP, social welfare spending, and education all have positive effect on capability and thus on life expectancy. However, the size of effect was found to be different among variables. For instance, GDP has the largest elasticity with while EDU also has a large elasticity at Among the three independent variables, SOC has the smallest coefficient at In the meantime, we find that the data shows the existence of autoregression since the coefficient of AR (1) is significantly different from zero. 13 Furthermore, we wanted to see whether the structural change before 1943 and after 1951 affects only intercept or also affect the slope, i.e. the coefficients. 14 In Model (C) of Table 2, we found that both the estimated coefficients of GDP and EDU 10 There are two reasons for us to take Anand and Ravallion (1993)'s format. One is that, according to Anand and Ravallion, this format could get the best fitness. Secondly, we could compare our results with Anand and Ravallion's directly. 11 According to our estimation, our data do show the existence of autocorrelation with degree one. Therefore we apply a first degree of autoregression, i.e. AR (1), to eliminate autocorrelation bias. 12 Anand and Ravallion (1993) applied a cross-country data to estimate how GNP affects life expectancy. Using an exact similar formula as we have here, they found that the elasticity of -(80-LIFE) with respect to GNP per capita is We did check the second order of autocorrelation and found that the coefficient of correlation is not significantly different from zero. 14 The change of coefficients before and after 1951 may imply that the influence of GDP, SOC, and EDU on life expectancy has been changed, too.

9 Capability, Social Welfare and Life Expectancy 199 have changed drastically (0.416 vs for GDP) and the coefficients of D51*log (GDP) and D51*log(EDU) all have positive and significant signs (0.250 and 1.361). This result shows that the elasticity of GDP on life expectancy is smaller for the first period of time, i.e , while the elasticity is larger for the second period, i.e to The same strategy applies to EDU (school enrollment rate). On the other hand, D51*log (SOC) has a negative sign, though insignificantly different from zero. It may imply that SOC has a smaller effect for the second period of time. Table 2 Capability and Life Expectancy: All Samples, Dependent Variable:-log (80-LIFE) (A) (B) (C) Constant ** ** ** ( ) ( ) (-9.711) D51(b) 0.903** 0.743** ** (27.923) (17.860) (-6.728) Log(GDP) 0.485** 0.416** 0.216** (25.044) (10.623) (2.893) Log(SOC) * 0.052** (1.948) (3.617) Log(EDU) ** 0.160** (3.434) (5.678) D51* Log(GDP) ** (2.582) D51* Log(SOC) (-1.513) D51* Log(EDU) ** (6.185) AR(1) 0.587** 0.427** (6.860) (4.457) (-0.221) Adjusted R F-value Number of Obs Notes: (a)the coefficients with ** or * are significantly different from zero at 95 percent or 90 percent significance levels, respectively. (b)d51 is a dummy variable, where D51=1 if the observation is between 1951 and 1997; otherwise D51=0 if it is between Meanwhile, the coefficient of AR (1) is no longer significantly different from zero here. It implies that the autocorrelation of life expectancy could be explained by those independent variables and dummy variables.

10 200 Taiwan, East Asia and Copenhagen Commitment 1. Gender Difference Since life expectancies are quite different for males and females in Taiwan, and in other countries, too, it will be interesting to compare how these crucial variables influence life expectancies for males and females. Table 3 and Table 4 show the relationship of capability and life expectancies in Taiwan for males and females, respectively. Basically, the two regressions have similar results in terms of signs of estimated coefficients and adjusted R 2. However, there are some differences worth mentioning. First, GDP has a larger effect on female life expectancy than male s. 15 Secondly, EDU has a greater effect for male than that for female. 16 Apparently, education has a significant effect on males since the enrollment rate for male was much higher in the early time. Constant D51(b) Log(GDP) Table 3 Capability and Life Expectancy: Male, Dependent Variable: -log (80-LIFE) (A) (B) (C) ** ( ) ** ( ) 0.854** (26.027) 0.346** (17.526) 0.718** (18.093) 0.292** (7.511) Log(SOC) (1.605) Log(EDU) ** (3.177) D51* Log(GDP) - - D51* Log(SOC) - - D51* Log(EDU) ** ( ) ** (-5.450) 0.231** (2.990) 0.042** (2.687) 0.173** (4.152) (1.632) ** (-2.155) 1.363** (5.622) (0.617) AR(1) 0.630** 0.496** (8.371) (5.600) Adjusted R F-value Number of Obs Notes: (a)the coefficients with ** or * are significantly different from zero at 95 percent or 90 percent significance levels, respectively. (b)d51 is a dummy variable, where D51=1 if the observation is between 1951 and 1997; otherwise D51=0 if it is between For instance, the extimated coefficient for GDP for female in Model (B), Table 4, is and it is only for male in Model (B), Table For example, the coefficient in Model (B), Table 4, is and it is in Model (B), Table 3.

11 Capability, Social Welfare and Life Expectancy 201 Table 4 Capability and Life Expectancy: female, Dependent Variable: -log (80-LIFE) (A) (B) (C) Constant ** ** ** ( ) ( ) (-3.037) D51(b) 0.941** 0.785** (18.440) (9.355) (-0.679) Log(GDP) 0.747** 0.631** (25.959) (10.050) (0.570) Log(SOC) ** 0.077** (1.984) (2.007) Log(EDU) ** (1.622) (3.103) D51* Log(GDP) ** - (2.436) D51* Log(SOC) (-0.198) D51* Log(EDU) (-0.320) AR(1) 0.643** 0.601** (6.733) (5.901) (0.617) Adjusted R F-value Number of Obs Notes: (a)the coefficients with ** or * are significantly different from zero at 95 percent or 90 percent significance level, respectively. (b)d51 is a dummy variable, where D51=1 if the observation is between 1951 and 1997; otherwise D51=0 if it is between The dummy variables have different effects on male and on female life expectancy. For instance, D51*log (GDP) has a significant sign for female (0.550) in Table 4, but it is insignificant for male (0.162) in Table 3. D51*log (SOC) has a negative and significant sign for male (-0.087) in Table 3, while it is insignificant for female (-0.015) in Table 4. Finally, D51*log (EDU) has a positive and significant sign for male (1.363) in Table 3, while it is negative and insignificant for female (-0.345) in Table 4. The results showed that drastic structural change between 1943 and Structural Change after 1951 In order to check whether the relationship between capability and life expectancy has been changed after 1951, we reran the regressions using two separated time periods. The results are shown in Table 5 and Table 6. In general, the regression results are quite satisfactory since almost all of the estimated coefficients have correct signs as we expected. Moreover, the regressions fit the data very well in both periods as the adjusted R 2 are

12 202 Taiwan, East Asia and Copenhagen Commitment quite high. Though most of the estimated coefficients are significantly different from zero, their magnitudes vary between male and female and between the two periods. The estimated coefficients of GDP are much smaller in the first period than that of the second period for both male or female. For example, the coefficient of GDP for all samples in Model (B) of Table 6 is and it is only in Model (B) of Table 5. This result implies that the marginal effect of GDP on life expectancy as increased after Table 5 Capability and Life Expectancy: Dependent Variable:-log (80-LIFE) All Sample Male Female (A) (B) (C) (D) (E) (F) Constant ** (-5.706) ** ( ) ** (-6.726) ** ( ) ** (-4.849) ** ( ) Log(GDP) 0.583** (3.561) ** (4.375) 0.528** (4.009) 0.257** (4.341) 0.653** (3.174) 0.203** (3.129) Log(SOC) ** (4.415) ** (3.116) ** (4.734) Log(EDU) ** (6.790) ** (5.031) ** (7.673) AR(1) 0.461** (2.259) ** (-2.272) 0.419** (2.206) (-1.572) 0.478** (2.108) ** (-2.757) Adjusted R F-value Number of Obs Notes: (a)the coefficients with ** or * are significantly different from zero at 95 percent or 90 percent significance levels, respectively.

13 Capability, Social Welfare and Life Expectancy 203 Table 6 Capability and Life Expectancy: Dependent Variable: -log (80-LIFE) All Sample Male Female Constant Log(GDP) Log(SOC) Log(EDU) (A) ** ( ) 0.480** (29.772) (B) ** ( ) 0.469** (8.242) (-0.312) ** (6.784) (C) ** ( ) 0.339** (19.394) (D) ** (-9.949) 0.378** (6.120) (-0.929) ** (4.293) (E) ** ( ) 0.744** (21.987) (F) (0.814) 0.697** (6.680) (0.521) ** (-3.366) AR(1) 0.635** ** ** 0.858** (7.982) (0.372) (9.540) (1.542) (8.079) (17.037) Adjusted R F-value Number of Obs Notes: (a)the coefficients with ** or * are significantly different from zero at 95 percent or 90 percent significance levels, respectively Moreover, the estimated coefficients of EDU after 1951 are greater than that prior to For instance, the coefficient of EDU for all samples in Model (B) of Table 6 is 1.257, while it is only in Table 5. Again, this result shows that education has a larger effect on life expectancy after The changes in the estimated coefficients of SOC are even more severe before and after In Table 5, as we expected, all coefficients of SOC are significantly different from zero, for both males and females. However, the coefficients of SOC in Table 6 are all-insignificant and even have reverse signs. The result showed that government spending on social welfare had a large effect before 1943 in Taiwan when people were much poorer. After 1953, when the economy had grown fast in Taiwan, the effects of social welfare spending on life expectancy become trivial. 17 However, the estimated coefficient of EDU for female in Table 6 is negative as It implies that a female with higher education will have a shorter life expectancy, which is against our expectation. And I think we have to put more effort to explain this puzzle.

14 204 Taiwan, East Asia and Copenhagen Commitment V. Conclusions In order to investigate the relationship of capability and life expectancy in Taiwan, we applied a time series data set from 1917 to 1997 to estimate the influence of GDP, education, and government spending on social welfare on life expectancy since the above three variables are taken as proxy variables for capability. In general, our regression results are quite satisfactory in several aspects, such as high level of valid and significant signs. Our main findings are summarized here: First, GDP, education, and government spending on social welfare all have a positive and significant effect on life expectancy as we expected. Secondly, the relationship between capability and life expectancy is different between males and females. For instance, GDP and education have a stronger effect on the life expectancy of males than females. However, social welfare has a larger effect on female life expectancy. Finally, drastic structural changes before 1943 and after 1951 in Taiwan were observed. In the first period, social welfare had a larger effect on life expectancy when people in Taiwan were poorer. It is a typical support-led type of development as developed by Dreze and Sen (1989). However, as income quickly grew in the second period, after 1951, the magnitude of social welfare on life expectancy became trivial in Taiwan. Meanwhile, income effect on life expectancy turned crucial to the second period of time. This result showed that the support-led type of development has shifted to the growth-mediated type of development after 1951 in Taiwan. References Adelman, I., and C.T. Morris (1973), Economic Growth and Social Equity in Developing Countries, Stanford University Press. Anand, S., and M. Ravallion (1993), Human Development in Poor Countries: On the Role of Private Incomes and Public Services, Journal of Economics Perspectives, Vol.7, No.1, pp Anand, S., and S.M. Ravi Kanbur (1991), Public Policy and Basic

15 Capability, Social Welfare and Life Expectancy 205 Needs Provision: Intervention and Achievement in Sri Lanka, in J.P. Dreze and A.K. Sen (eds.), The Political Economy of Hunger, Vol.3, pp.59-92, Clarendon Press, Oxford. Barclay, G.W. (1954), Colonial Development and Population in Taiwan, Princeton, University Press. Bhalla, S.S., and P. Glewwe (1986), Growth and Equity in Developing Countries: A Reinterpretation of the Sri Lanka Experience, World Bank Economic Review, Vol.1, No.1, pp Dreze, J.P., and A.K. Sen (1989), Hunger and Public Action, Claredon Press, Oxford. Fogel, R.W.(1991), The Conquest of High Mortality and Hunger in Europe and America: Timing and Mechanism, in Favorites of Fortune, Harvard University Press, Cambridge. Grant, J.P. (1978), Disparity Reduction Rates in Social Indicators, Overseas Development Council Press, Washington D.C. Kakwani, N.C.(1981), Welfare Measures: An International Comparison, Journal of Development Economics, No.8. Laderchi, R.C. (1997), Poverty and Its Many Dimensions: The Role of Income as an Indicator, Oxford Development Studies, Vol.25, No.3, pp Morris, M.D. (1979), Measuring the Conditions of the World's Poor: The Physical Quality of Life Index, Pergamon Press, Oxford. Sen, A.K.(1977), On Weights and Measures: Informational Constraints in Social Welfare Analysis, Econometrics, 45, Sen, A.K.(1984), The Living Standard, Oxford Economic Papers, No.36, Supplement, pp Sen, A.K. (1985a), Commodities and Capabilities, North-Holland, Amsterdam. Sen, A.K.(1985b), Well-Being, Agency and Freedom: The Dewey Lectures 1984, Journal of Philosophy, Vol.82, No.4, pp Sen, A.K. (1987a), The Standard of Living: The Tanner Lectures, Cambridge University Press.

16 206 Taiwan, East Asia and Copenhagen Commitment Sen, A.K. (1987b), On Ethics and Economics, Blackwell, Oxford. Sen, A.K. (1998), Welfare Economics and the Quality of Life, Life Expectancy and Economic Evaluation, The Institute of Economics, Academia Sinica, Taipei. UNDP (1990), Human Development Report, 1990, Oxford University Press. UNDP (1995), Human Development Report, 1995, Oxford University Press. Wu, T.M.(1991), The Estimation of GDP of Taiwan, , Taiwan Economic Review, Vol.19, No.2, pp