The effects of structural changes on deindustralization: Evidence from high, medium and low income countries

In this study, the effect of structural changes on (premature) deindustrialization is investigated by panel data analysis methods. Using annual data for the period 1970-2017, models for four different country groups (high-income, upper-middle-income, lower-middle-income and low-income) are established. In these models, the share of manufacturing production in GDP is the dependent variable. Labor productivity, GDP per capita and the share of foreign trade volume in GDP are explanatory variables. Four hypotheses are determined to reveal and test the relationships between variables. Hypothesis 1: Factors affecting (premature) deindustrialization do not differ according to the development level of countries. Hypothesis 2: There is no linear-negative relationship between globalization and (premature) deindustrialization.Hypothesis 3: There is no linear-negative relationship between GDP per capita and the (premature) deindustrialization process. Hypothesis 4: There is no linear-positive relationship between labor productivity and (premature) deindustrialization. Panel data methods that take into account the cross-sectional dependence is used at every step of the method part of the study. As a result of the tests, a dependence is determined between the cross-sections in all models and tests is applied using this information in the later stages of the study. While investigating the stationarity of the series, Pesaran (2007), one of the second-generation unit root tests, is used instead of the first-generation tests. As a result of the stationarity test, it is determined that all variables (for each country group) become stationary when the first difference is taken.In order to prove the long-term relationship between the variables, Westerlund (2007)cointegration test, which takes into account the cross-sectional dependence, is applied. It is proved by this test that the variables move together in the long term for all models. Since the models have cross section dependence, Augmented Mean Group (AMG) estimator, which is one of the modern estimators, is used for coefficient estimation. The hypotheses are tested by applying this method. While an increase in foreign trade volume in the high-income country group has positive effects on the manufacturing; the effect of an increase in GDP per capita is negative. In terms of labor productivity variable, a consensus is not reached, and Hypothesis 4 is accepted. An increase in foreign trade volume in the upper-middle income country group also positively affects the manufacturing sector. In terms of the effect of GDP per capita, it is observed that the negative-positive effect differs according to the country specific coefficients. It is among the findings of the upper-middle income country group that the increase in labor productivity has positive effects on the development of the manufacturing. In lower-middle and low-income countries, the hypotheses is accepted due to both the data limitation and data inappropriateness, and the effects are not be proven from a holistic view for the country groups. The findings show that the factors affecting (premature) deindustrialization differ according to the income levels of the countries.