國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24
unbalanced panel database. There exist some missing values in the statistical data of each country, including both stock market return and independent variables. But we still keep all the main economy and do not delete any economy, which is too small and always not be taken into account. We hope this research will be a common model for most countries and economies of the world, and not confined to the developed countries or bigger economies that are always in the spotlight. In addition, we expect that we could analyze the long-term relationship between stock market return and national competitiveness. Nowadays many literatures find that the changing process of national competitiveness will have a profound influence, and which is why we have to take a longer research period for the relationship between national competitiveness and stock market return. This is also consistent with our goal to pursue the long-term stock market return through the national competitiveness.
4.2 Data analysis
Table1 Variable definition
Variables
Definitionstockreturn
The rate of return of major stock index of a countrycapital
Gross fixed capital formation, percentage of GDPGDPgrowth
Real GDP growth Percentage change, based on national currency in constant pricesgdp
Gross Domestic Product, US$ billionssaving
Gross domestic savings (%), Percentage of GDPeducation
Total public expenditure on education (%), Percentage of GDPproductivity
Overall productivity, GDP per person employed, US$change_lead
Percentage of competitiveness rank change, lead index‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25
Table2 Descriptive statistics of each variables
Variable Obs Mean Std. Dev. Min Max
stockreturn 700 0.162542 0.443496 -0.8919 5.121357
capital 682 -0.00846 0.090428 -0.55371 0.388858
GDPgrowth 700 3.280325 3.640773 -14.7 18.3
gdp 693 810.3585 1815.076 7.230608 14526.55
saving 680 24.83636 8.574466 -7.131525 55.18897
education 600 0.004286 0.111484 -1.0215 0.480785
productivity 668 0.049877 0.126579 -1.00539 0.408816
change_lead 614 0.039454 0.318213 -0.66667 3.5
Most data for variables in our sample is nearly complete, but the data for variable of education expenditure on GDP is not so complete. Some countries do not disclose the statistical data about their education expenditure until recent years, so we do not have enough data to do more researches about the influence of education on the national competitiveness. The education is also a subject we want to focus on in the beginning. Due to the constraint of data collection, we do not view it as a significant influence on the competitiveness or the stock market return.
When we start to use the random effect model and fixed effect model, we translate the gross fixed capital formation, GDP, saving rate, and education expenditure, and productivity variable into the logarithm function, and then place these logarithms into the formula. In this way we could avoid the autocorrelation problem.
‧ 國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
Table3 The correlation matrix for each variable
Note: the stars in the cell represent significant at the 5 percent level
stockr~n capital GDPgro~h gdp saving educat~n produc~y change_lead
stockreturn 1
capital 0.3628* 1
GDPgrowth 0.4824* 0.5597* 1
gdp -0.0476 0.0115 -0.0685 1
saving 0.0428 0.0460 0.0729 -0.0038 1
education -0.1398* -0.0828* -0.0917* 0.0209 -0.0089 1
productivity 0.2796* 0.3436* 0.5582* -0.0012 -0.014 0.0846* 1
change_lead -0.0949* -0.1183* -0.1615* 0.0529 -0.0199 0.0401 -0.0067 1
‧
We find that the real GDP growth is highly correlated with capital formation. It is not surprising, because according to the classical economic theory we know that an important source inducing the GDP growth is the input, that is, the capital formation.
These two variables are highly correlated with each other, and it will bring a good feedback to real GDP growth if we make more capital formation. According to Solow growth model, the accumulation of capital is an important factor to drive the real GDP growth, the model shows that investment is in direct proportion to economic growth.
There are many empirical facts consistent with the Solow model, such as (Barro and Sala-i-Martin, 1992, 1995; Levine and Renelt, 1992; Mankiw et al., 1992 or Nonneman and Vanhoudt,1996). These empirical researches prove that there exists strong relationship between the accumulation and real GDP growth in both long-term and short-term period. As a result, we avoid puting these two variables at the same time in order to avoid the collinearity problem. And in the model we set, we take them as substitutes for each other.
We expect that when a country puts more resources into education, it will cause the increasing of labor productivity. Moretti (2004) took direct approach to the estimation of human capital externalities and emphasized the productivity of manufacturing plants. He finds that plants located in cities with high level human capital can produce much more output with the same input than plants located in cities with lower level human capital. He defines the overall level of human capital in the city by calculating the fraction of college-educated workers among all the workers in the city outside the plant. After controlling for a plant's own human capital, he finds that the productivity of plants located in cities that experience increases in the overall level of human capital rises more than that of the plants located in cities where the overall level of human capital is constant”. According to his research, a one-percent increase in the city share of college-educated student related with 0.5-0.6 -percentage
‧
model at the same time. We take them as substitutes to avoid the collinearity problem.Since the World Competitiveness Center editing the national competitiveness rank takes the real GDP growth as a hard data criterion, and hard criteria represent a weight of 2/3 in the overall ranking, we know that the economy performance correlates with the national competitiveness rank significantly. As what we expected, there exists significant correlation between the real GDP growth and the change of national competitiveness. As a result, we prefer not to take these two variables into the model at the same time to avoid the insignificant problem of beta caused by collinearity.
Overall, there still exist some correlation between the variables, but because the correlation value is not as high as we mention previously, and we do not think they will cause the problem to this study if we do not take them into account.