Revisiting the Lewisian turning point in Taiwan

Abstract

The Lewisian concept of dualism was empirically relevant to Taiwan given its heavy population pressure on scarce land. Most of the literature, albeit based on little solid evidence, concluded that Taiwan reached the Lewisian turning point around the end of the 1960s. Others argued that the LTP was after 1979. This paper first follows three criteria proposed by Minami (1973). We observe the active reallocation of employment after the mid-1960s and the dramatic increases in real wages of the agricultural and unskilled workers at the end of the 1960s. Furthermore, we introduce a new empirical methodology on the agricultural production function by adding a year dummy variable to identify a potential LTP, which leads to the agricultural marginal product of labor being zero before the LTP and becoming positive thereafter. Our findings support the notion that the LTP in Taiwan occurred around the period 1969–1973, rather than after 1979.

Keywords:

• Lewisian turning point
• agricultural real wages
• labor surplus
• Taiwan

JEL Codes:

• O14
• J42

Notes

1 See the Appendix for statistical data for Taiwan for the period 1951–1981.

2 The following illustration is adapted from Islam and Yokota (2008).

3 For readers’ references, please refer to the first paragraph of section 5, our comment on Wu (1983), and footnote 20.

4 Kuo, Ranis, and Fei (1981) and Kuo (1977) also claimed that “by 1971, Taiwan’s economy achieved full employment.” Because LTP represents “the unification of one labor market,” rather than full employment, which represents labor market equilibrium, we would, therefore, skip any discussion on the timing of full employment, or other related variables, such as unemployment, and underemployment. For example, Fei, Ranis, and Kuo (1979) also claimed that “unemployment or underemployment in Taiwan was virtually eliminated by the end of the 1960s.” We would also skip the discussion on these related issues.

5 The authors are not convinced that their studies have provided enough evidence to identify the timing of Taiwan’s LTP. To be confident in this conclusion, we attempt to cover their papers as comprehensively and thoroughly as possible, which included Ranis (1973), Fei and Ranis (1975), Kuo (1976), Kuo (1977), Ranis (1978), Fei, Ranis, and Kuo (1979), Kuo, Ranis, and Fei (1981), Ranis (1995), and Fei and Ranis (1997). Ranis (1978), as a summary paper, referred his findings to Fei and Ranis (1975) and Fei, Ranis, and Kuo (1979), and we would skip the review of this paper. Note that all studies except Ranis (1973) concluded Taiwan’s LTP had been reached by the end of the 1960s.

6 Data analysis by Fei, Ranis, and Kuo (1979) referred a lot to Kuo (1976). Kuo, Ranis, and Fei (1981) cited Kuo’s (1977) analysis of “labor absorption” as their source. Although the focus of both Kuo (1976) and Kuo (1977) was not on the LTP, we also reviewed Kuo’s work. While Kuo (1976) was a subset of Kuo (1977), here we skipped the review on Kuo (1976). Kou (1977) added one more section of analysis on Taiwan’s input-output tables of 1961, 1966, and 1971.

7 Kuo’s (1977) labor absorption in either agricultural or nonagricultural sector was equivalent to the incremental increase in employment, which was identical to the difference between newcomers and change in unemployment. The number of newcomers was the difference between the incremental increase in population and those not in the labor force for those aged 15 and over.

8 Kuo (1977) also inferred other timing of the labor market transition that was not consistent with the argument of 1968 as the LTP. For example, she claimed “the agricultural sector has been releasing laborers since 1968. Labor migration out of the agricultural sector to the industrial sector was particularly rapid during the period of 1968–1974”

9 Fei, Ranis, and Kuo (1979) also depicted real wages of males and females in manufacturing, and transport and communication sectors in several figures. For a similar reason, none of these other wage trends served as a clear and supportive argument on the timing of the LTP.

10 In most of the literature, unskilled labor was proxied by agricultural workers, female workers, workers in textiles, or nonagricultural workers. As counterparts, skilled workers were those in manufacturing, male workers, workers in other manufacturing industries, or nonagricultural staff.

11 The ratio is derived as [1 − (170/365)].

12 Chen’s (1983) four estimates of output elasticity of labor were 0.25 for 1952–59, 0.07 for 1960–64, 0.26 for 1965–67, and 0.82 for 1968–78. The first three values were cited from other studies and Chen estimated the last one.

13 Chen (1983) adopted gender, age, industrial, occupational, and educational categories as his criteria differentiating unskilled and skilled workers. For details, refer to his Fig. 6 to Fig.13. Only two subcases shown in two figures indicated larger wage gaps, which were workers with primary school education compared with workers with bachelor’s degrees, and female compared with male workers aged 25–44. The paper also showed that the wage gaps of the unskilled relative to technicians and staffs was narrowed 10% to 11% from 1971 to 1976, which does not serve either as convincing evidence regarding 1966–68 as the LTP.

14 The main attempt of Hou and Hsu (1976) and Hsu (1982) was to argue that Taiwan economy as a whole had not encountered labor shortage in 1979.

15 The version of “Statistics on Agricultural Wages of Hired Workers in Taiwan” that we could find was published in 1985 and contained statistics only from 1961–84. The earlier series of 1952–60 data come from Wu (1984) which was based on the same statistical book with the version of 1983. Wu (1984) reported his estimates of annual real wages after multiplying by the actual annual working days, and we have converted his annual rates into daily rates.

16 The total working hours varied each year during 1973–81 and also varied across different manufacturing sectors, from 263 hours to 201 hours. The authors choose 30 days as the denominator and derived daily wage rates. Since most wage rates of 1973 from the second set were much smaller than those of 1972 from the first set, we suppose the two sets did not stem from the same origins.

17 Data source: DGBAS (1982).

18 Among several subgroups of manufacturing industries, the overtime wage of workers for motor production was about the highest. Wages of unskilled and skilled were also provided by DGBAS as two categories.

19 Wu (1984) did not explicitly address the timing of the LTP. His analysis was not only focused on the relationships between wages and labor productivities of agriculture but also on other industries, including manufacturing, mining, construction, and several services sectors.

20 Minami (1973) estimated MP_N^A by adopting pooled cross-regional data over time. However, adopting a smooth agricultural production function simply with “time series” data as Wu (1984) did, makes it impossible to pin down a kink or a particular turning point. We reprint his specification, estimate MP_N^A, and show the difficulty for applying CR1 and CR2 in identifying the LTP. Due to the space limits, the detailed illustration is upon request for readers who are interested in the details.