Dynamic Panel Data Approach

國

立 政 治 大 學

‧

N a

tio na

l C h engchi U ni ve rs it y

Chapter 4 Methodology Framework and Data Source

This chapter presents methodologies used in this study and lists data sources for measurement of variables. The first section introduces a

conditional convergence method for the investigation of the impact of between the telecommunications development and economic growth. The secondary section interprets variable measurement. Lastly is data source and research limitation of this study.

4.1 Dynamic Panel Data Approach

The objective of this paper is to verify the role of telecommunications development performed by explaining the different growth performances across 31 regions of China. The methodology conducted in this study adopted Ding’s (2005), Kingsley’s et al. (2006) and Liu’s (2008) general analysis of the impact of telecommunications development on regional economic growth. The common approach to studying the sources of economic growth is to use Barro-type analysis which questioned by Islam that all the countries or regions have identical aggregate production functions.

Islam (1995) investigated that the analysis of cross-country or cross-region regressions is weak in justifying the assumption of identical production function.

He revised the regression equation with a dynamic panel data model to proceed. This approach is beneficial to allow for differences in the aggregate production function across countries or regions. Ding (2006) mentioned that the method can also capture short-run autoregressive behavior by adding a lagged growth rate as an independent variable. Caselliet et al. (1996) also

‧

stated that the panel data approach is better than the cross-section regression approach due to its control over the omitted variable bias and endogeneity problem (Liu, 2008).

We follow Barro’s (1991) approaches which enable us to test the conditional convergence hypotheses. Liu (2008) mentioned that the

cross-section static approach would have the following specification on regions of China:

i: indexes the 31 regions, municipalities, and autonomous cities in China;

GRTH : represents the annual growth rate of real GDP per capita;

initial

GDP i

Ln( )_, : represents the initial level of real GDP per capita in

logarithm form for each region;

X: contains a set of variables accounting for production factors and other conditional variables at the beginning of the study period for each region;

A negative sign of the coefficient of initial real GDP can indicate the existence of conditional convergence while the set of economic condition variables are held constant. To avoid the statistically inherent small sample problem, the total number of regions must exceed some level.

The Islam panel-data Approach

Islam (1995) stated that a panel data approach is advocated and

implemented for studying growth convergence. The similar equation for testing

‧

convergence in this research is defined as a dynamic panel data model and different panel data variables are used to estimate it. The main usefulness of the panel approach lies in its ability to allow for differences in the aggregate production function across economies. This leads to results that are

significantly different from those obtained from single cross-country

regressions. Moreover, fixed effect rather than random is recommended to implement a panel data approach for the differences of the unobservable errors.

We conduct growth theory in this research to define variables including following explanatory variables: fixed investment, foreign direct investment, employment rate, human capital, population growth, urbanization level, industrial output share from state-owned enterprises, transportation

percentage, as well as telecommunications development, plus for the two level variables: the lagged GDP growth and lagged GDP per capita as the basic explanatory variable to test which has a significant impact on regional

economic growth according to the guidelines in the previous growth literature.

it

All the definitions and measurements of the variables are the same as those in the Barro cross-section static approach; _i, _t and _itrepresent the regional dummy, temporal dummy, and error terms, respectively. The variables in Equation 4.2 are measured as follows:

GRTH : annual growth rate of real GDP per capita which is the dependent variable;

‧

立 政 治 大 學

‧

N a

tio na

l C h engchi U ni ve rs it y

1

GRTHt : represents the one year lagged GRTH;

) 1

(GDP _t

Ln : represents the log value of the lagged real GDP per capita in 2000 value (RMB);

INV : percentage of total investment in fixed assets in GDP;

FDI: percentage of foreign direct investment divided by total investment in fixed assets;

POP : annual population growth rate;

EMP: percentage of employed persons at year-end to total population;

HC : human capital which is measured by the average years of schooling for the population aged 6 and above;

URBAN : the share of urban population to total population;

SOE : the share of state-owned enterprises in total industrial output;

TRANS : transportation percentage as measured by the length of rail, highway, and waterway networks per square kilometer;

TEL: the number of fixed line telephone users per 100 inhabitants;

MOB : the number of mobile phone users per 100 inhabitants;

NET : the number of Internet usres per 100 inhabitants;

Furthermore, we use additional variables with square terms to explore the character of returns to investment to check whether the relationship between telecommunications and economic growth is linear because some previous researchers found that investment in telecommunications

development will not significantly affect economic growth of a country until a critical mass of telecommunication development is achieved (Roller and Waverman, 2001; Savage, et al, 2003).

‧

立 政 治 大 學

‧

N a

tio na

l C h engchi U ni ve rs it y

Diagnosis of Causality

We study another important effect that the direction of causality between telecommunications development and economic growth can be tested using the Granger Test technique.²⁵ Ding (2007) researched that it is to avoid causality issues by identifying the determinative effects of initial per capita income and a series of other conditional variables on the conditional

convergence approach. Hence, we will use the Pairwise Granger Causality Test with the lagged values in the regression to detect the causality direction according to the approach implemented by Ding (2005) and Liu (2008).