:
A Transition Analysis of Economic Growth: The Case of Taiwan
NSC 88-2415-H-006-004
whsieh@mail.ncku.edu.tw !" #$%
)J*!! ! !
&'!"() 1987 *+, (12.7%) -./0123 456789:;* 6%
<=>?@ 4* 5% ABCDEFG0H IJKLMNOP&'!"(*QRST
<!"UVWX*YZ[ \]&'!"D 67^_`abcC SEd#$efg hij!"kYlm<nopq*rst (robustness)Huvw logistic xyWXz{
(logistic smooth transition regression, LSTR)
|}kY~UVWX*t (discrete breaks) <xy a~pt*
H4OP&'!"UVWX*Dt (existence) <C * Solow
(kYa<!"UVWYZa
H4]!"UVWX*[
&'¡¢q<!"£¤a ¥¦§C
¨Z©ªkY*«¬U®¯¤ 96%
!"(aW¨~[E * Solow kY4 lm°±«j²d<³¨q´µ¶ GDP *
30·H¸q²¹<º»]¼ST½¶¾
* 40·CE~¿HÀÁÂÃfq²¹<
º»]¼aMLt½kYU®uÄÅÆ M*ÇÈCÉ Ê* ¥IËMÌq²¹ aÍÎ<º»ÏÐa H[\]Ñ'a!
"(C
logistic xy*WXÒ{ (logistic smooth transition regression)HUVt*W
(structural changes)H!"(Hq²¹<
º»]¼
Abstract
Prior to the turmoil of Asian financial crisis in 1997, Taiwan had shown signs of declined economic growth. There is an urgent need to reexamine the determinants of economic growth in Taiwan to revive the post-crisis economy. This manuscript first identifies the transition of Taiwan's economic growth by applying logistic smooth transition regression (LSTR) model. As suggested by the estimated results of the LSTR model, the analysis of Solow’s growth model can be proceeded without concerning the possibility of a discrete break. The estimated regressions, based on the out-of-steady-state dynamics, have a relatively high explanatory power.
More than 96% of the variations in output per worker can be explained by the models. The estimated results indicate that physical capital and labor, respectively, account for about 30%
of the GDP. The balance, roughly about 40%, is contributed by the factors of human capital and technical progress. Thus the omission of the importance of human capital and/or technical progress would cause biased estimates of regression parameters.
Government policies to facilitate the advance human capital and technological progress are essential for promoting Taiwan’s economic growth.
Keywords: logistic smooth transition
regression (LSTR), structural changes, economic growth, human capital, and technological progress
(II) Project Report
1. INTRODUCTION
Prior to the onset of the Asian Financial crisis in July 1997, economic activities in Taiwan had shown signs of slowdown after the rapid growth in the period of 1960s to 1980s. For instance, real per capita gross national product (GNP) grew at an average annual rate of 4.98% in the period of 1990-98 as compared with the long-run growth rate of 8.21% since the early 1950s. The decline in the growth rate is accompanied by the declining trends of the investment rates. The gross fixed capital formation rate, measured as the percentage of GNP, declined from its peak of 31% in 1975 and 1980, to 21.6% in 1998 (CEPD, 1999: pp.46 & 56).
Nevertheless, there are some positive developments in Taiwan that may contribute to continual economic growth. One of the positive developments in Taiwan is more emphasis on research and development. R&D expenditures as a percentage of gross domestic product (GDP) rose from 1.01% in 1986 to 1.92% in 1997. Still another strength is the improvement in human capital.
Educational expenditures as percentage of GNP increased from 5.1% in 1985 to 6.5% in 1998, with the maximal level at 7.1% in 1993 (CEPD, 1999: pp.11 & 121).
Does the declined Taiwan’s GDP growth rates from previous high at 12.7% in 1987 to about 6% in recent decade, and further dropped to about 5% after the Asian financial
crisis, indicate that the epoch of Taiwan’s rapid growth is concluded? This debatable question is not intended to be answered here.
However, to provide an initial attempt and make an inference to this question, both the transition and the determinants of Taiwan's economic growth need to be examined carefully. Therefore, the analysis of this manuscript is in two stages. First, there may exist structural changes in Taiwan's economic growth, thereby a transition or a discrete break needs to be identified to enhance the robustness of any related estimated results.
This can be facilitated by a simple logistic smooth transition regression (LSTR) model.
The LSTR model can be utilize to explore possible deterministic structural changes, either abruptly or gradually. By employing the LSTR model in time-series regression, without imposing discrete breaks, can mitigate the problem of misspecification, if any, in modeling. Moreover, Taiwan's policy reform or economic liberalization may actually initiate a transition, rather a discrete break. Having identified the transition in the growth series, then the next stage is to test whether Taiwan's economic growth can be characterized by Solow’s growth model augmented with human capital investment and technological progress. Some of recent studies on Taiwan's economic growth made substantial contributions to the understanding of the subject, but they did not explicitly test the augmented Solow’s growth model. Some of the previous studies began with the production function and employed capital stock as one of the explanatory variables, whereas this paper uses the share of inputs in
GDP to estimate regression parameters based on the steady-state equilibrium condition as well as the out-of-steady-state dynamics.
2. ESTIMATED RESULTS
Economic liberalization or policy change can be implemented in a “big-bang” or
“gradualism.” Even in a big-bang reform, the subsequent effects on economic growth will typically be gradual rather than dramatic. The special features of the LSTR model are twofold. First, there is no need to identify a structural change. Second, the standard structural model is embedded in the LSTR model.
Following Granger and Terasvirta (1993, chapter 7), a LSTR trend model of real GDP (denoted by Y) may be written as
(1) ln(Yt) = c0 + c1 St(,) + c2 t + c3 t St(,) + t t = 1, ..., T, St(,) = {1 + exp[-(t - )]}-1.
Where St is a curvilinear logistic function that maps trend variable t onto the interval (0,1) andt is a zero mean disturbance term. The location parameterdetermines the timing of the transition and parameter controls the velocity of transition. If > 0, the model transition occurs smoothly from the initial state to the final state. Meanwhile, If takes a large value then the transition is completed in a short period of time. As approaches to infinity, the model collapses to one with instantaneous structural break in intercept and trend at time t =. Hence, it is clear that the standard structural break
model is a special case of the LSTR trend model. If < 0, the interpretation of the parameters remains the same although the model’s initial and final states are reversed.
In contrast to standard structural break model, no a priori information is used to locate the date of a transition since the midpoint of the transition is determined endogenously via the parameter. Therefore, the nonlinear (in parameters) model (1) is estimated by nonlinear least squares using Taiwan’s annual data of real GDP (at 1991 prices) for the period of 1953-1997. However, the stationarity of the time-series data needs to be examined first. Hence, the unit root test is performed. Based on the augmented Dickey-Fuller (ADF) test, the null of unit root is rejected.
The estimated results, as indicated by the adjusted R-squared, more than 99.9% of variations in output can be explained by the LSTR trend model. The estimated coefficients have the expected signs and are significant at the 1% level, except c3. The insignificant magnitude of the estimated c3 indicates very limited fluctuations across different states of transition. Additionally, the small magnitude of the estimated = 0.2544 shows the transition from the initial to the final states is slow and smooth. Owing to the nature of policy reform or economic liberalization, the impacts on Taiwan’s economic performance may not show any discrete structural break.
After the smooth transition in Taiwan’s economic growth is confirmed, one can proceed with the construction of growth model. This paper is an extension on the works of Mankiw, Romer and Weil (1992),
and Nonneman and Vanhoudt (1996), who recently tested the augmented Solow model by incorporating, respectively, human capital investment and technical progress in cross-sectional data. This paper considers both human capital investment and technological progress in a time-series (rather than a cross-sectional) analysis, thereby the special characteristics of economic growth trend can be captured. Then, following Barro and Sala-i-Martin (1992), the out-of-steady -state dynamic situation is also considered, with the parameter λ indicates the speed of convergence toward the steady state (y*).
More than 96% of variations in output per worker can be explained by the model with appropriate value of Durbin - h statistic.
The input shares for physical capital, human capital, technology, and labor in real GDP are estimated to be 0.283, 0.237, 0.189, and 0.291, respectively. The importance of human capital and technical progress as additional inputs cannot be ignored as shown either in the steady state or in the dynamic process.
3. CONCLUSIONS
This paper first employs the LSTR trend model to examine the transition in Taiwan’s economic growth. There is no need to identify a structural change and the standard structural break can be embedded in the LSTR model.
By employing time-series data of Taiwan’s real GDP, the estimated magnitude of in the LSTR model is very small. In turn, it indicates Taiwan’s economic transition from the initial to the final model states is slow and smooth. Owing to the nature of policy reform or economic liberalization, the impacts on
Taiwan’s economic performance may not show any discrete structural break. In fact, the economic liberalization in Taiwan initiates a transition, rather a discrete break. The estimated results from LSTR model confirm that observation.
Then, the paper attempts to test the augmented Solow model in that human capital investment and technological progress are considered. More than 96% of the variations in output per worker can be explained by the model. The coefficients have correct signs and most of them are significant at the 1% level.
The estimated results suggest that physical capital and labor, respectively, account for about 30% of the income share. The balance, roughly about 40%, is contributed by the factors of human capital and technological progress. Thus, the omission of human capital and/or technological progress would cause biased estimates of regression parameters.
Moreover, the importance of human capital and technical progress as key elements for Taiwan’s economic performance cannot be over emphasized. Government policies to facilitate the accumulation of physical and human capital, the increase of labor participation rate, and the speedup of technical progress are essential for promoting Taiwan’s economic growth.
(III). Self Evaluations
Owing to the valuable comments made by the anonymous referee(s), the manuscript made significant improvements out of the original proposal. The most significant improvement is made pertinent to the growth model, in which the Solow’s augmented
growth model replaced the originally proposed IS-LM model. However, the paper has some limitations. A more flexible specification, other than the curvilinear logistic function St, should be considered to allow for non-monotonic and non-symmetric transition. Nevertheless, due to the limited number of observations available, the less complex LSTR trend model could be a better choice to avoid the problems of degrees of freedom. Furthermore, depreciation rates may differ among capital inputs in the augmented Solow’s growth model.
An earlier version of this manuscript was presented at the 1999 annual meetings of the Allied Social Sciences Association in New York, USA. The paper presentation is financed by the Ministry of Education. Currently, the project is under going a revision for paper submission to a reputable refereed journal.
(IV) REFERENCES (PARTIAL)
Barro, Robert J. (1991), “Economic Growth in a Cross-Section of Countries,”
Quarterly Journal of Economics 106(2):
407-443.
________ and X. Sala-i-Martin (1992),
“Convergence,” Journal of Political Economy 100(2): 223-251.
CEPD (1999), Taiwan Statistical Data Book 1999. Taipei, Taiwan: Council for Economic Planning and Development, Executive Yuan, R.O.C.
DGBAS (1999), Abstract of National Income Statistics Monthly September 1999. Taipei, Taiwan: Directorate General of Budget Accounting and Statistics, Executive Yuan, R.O.C. (available at web site
http://www.dgbasey.gov.tw/).
Granger, C.W.J., and T. Terasvirta (1993), Modelling Nonlinear Economic Relationships. Oxford: Oxford University Press.
Greenaway, David and David Sapsford (1994),
“What Does Liberalization Do for Exports and Growth?” Weltwirtschaftliches Archiv 130(1): 152-174.
Lee, Maw-Lin, Ben-Chich Liu, and Ping Wang (1994), “Growth and Equity with Endogenous Human Capital: Taiwan's Economic Miracle Revisited,” Southern Economic Journal 61(2): 435-444.
Levine, R. and D. Renelt (1992), “A Sensitivity Analysis of Cross-Country Growth Regressions,” American Economic Review 82(4): 942-963.
Lin, C.J., and T. Terasvirta (1994), “Testing the Constancy of Regression Parameters Against Continuous Structural Change,”
Journal of Econometrics 39: 211-228.
Mankiw, N. Gregory, David Romer, and David N. Weil (1992), “A Contribution to the Empirics of Economic Growth,”
Quarterly Journal of Economics, 107(2):
407-437.
Nonneman, Walter and Patrick Vanhoudt (1996), “A Further Augmentation of the Solow Model and the Empirics of Economic Growth for OECD Countries,”
Quarterly Journal of Economics 111(4):
943-953.
Tallman, Ellis W. and Ping Wang (1994),
“Human Capital and Endogenous Growth:
Evidence from Taiwan,” Journal of Monetary Economic 34(1): 101-124.
