Knowledge, technological convergence and economic growth: a dynamic panel data analysis of Middle East and North Africa and Latin America

Abstract

This paper aims to investigate the role of various knowledge indicators, such as research and development, information and communication technologies and trade, in the economic growth and convergence performance of 17 Middle East and North Africa (MENA) countries and 17 Latin American countries (LACs) during the 1980–2014 period, by utilizing dynamic panel data techniques like the pooled mean group estimation method. Our results indicate that knowledge variables seem to play an important role in the economic growth performances of both regions and, overall, there is a significant convergence among the countries in MENA and LACs. Nevertheless, our main finding is that the impact of the knowledge indicator in Latin American countries is weaker compared to the MENA region.

Appendices

Appendix 1: Data and descriptive statistics

1.1 The definitions and the sources of data

The main variables that are used in the model are output, capital stock, human capital, R&D stock, trade, and ICTs. While the former three are obtained from PWT 8, the others are from the WDI database. The WDI provides various indicators, ranging from demographic to environmental topics and it contains more than 800 indicators for 214 countries for the years 1960–2016, compiled from officially recognized sources. Whereas the PWT provides 30 variables on purchasing power parity and national income accounts indicators for 167 countries for the 1950–2014 period.²⁴,²⁵

Output [Y] is the real gross domestic product (GDP) at constant 2005 national prices (in million US$) from the PWT 8 data set.

Capital stock [K] is the capital stock at constant 2005 national prices (in million US$) from the PWT 8 data set. In the PWT 8 data, capital stocks are “estimated based on cumulating and depreciation past investments using the perpetual inventory method” (Inklaar and Timmer 2013: 5).

Human capital per worker [h] is obtained by calculating the index of human capital per person based on years of schooling (Barro and Lee 2012) and returns to education (Psacharopoulos and Patrinos 1994) from the PWT 8 data set. Education is an important indicator of the capacity of the labor force to use the available information. Barro and Lee (2012) use a combination of data sources to infer the percentage of each country’s adult population (aged twenty-five and older) the particular level of education they obtained for each year. Census data provide direct measures of a country’s stock of education but, especially, in developing countries such data are only available for selected years. Barro and Lee (2012) use enrollment data and data on literacy rates to interpolate between census years to fill the missing data.

Human capital [H] is calculated as H = hL, where h is human capital per labor and L is total employment from the PWT 8 data set.

R&D [R] due to lack of data by utilizing total patent and total trademark we have calculated R&D index. Since these two variables are in different units and have different ranges (minimums and maximums), we used the Human Development Index (HDI) methodology to obtain a common range for them. That is, a minimum and a maximum bound is set to each of the four indicators and a number (index value) is obtained for each of these indicators between 0 and 1. After this process all of the raw variables turned into unit free indices, between 0 and 1, that can be added together. With this conversion the two variables become dimension indices which are labeled as IT and IP. The two dimension indices are calculated as follows:

$$IT_{t} = \frac{{T_{t} - Min(t)}}{Max(T) - Min(T)}$$

(8)

$$IP_{t} = \frac{{P_{t} - Min(P)}}{Max(P) - Min(P)}$$

(9)

where T_t and P_t, represents total trademark, total patent respectively. Min (X) is the minimum value and Max (X) is the maximum value of variable X during the time interval that is being investigated [The minimum and maximum values of each variable during the 1980–2014 period].

After normalizing the indicators and obtaining the dimension indices next we calculate the R&D Index (R) as a weighted average of the two sub-indices, as follows:

$${\text{R}} = {\text{w}}_{1} {\text{IT}} + {\text{w}}_{2} {\text{IP}}$$

(10)

where w_i’s denote weights of the respective dimension indices.

HDI used simple average methodology to determine the weights of each dimension index simply because all three dimensions were considered to be equally important. That is, the three dimension indices (Life expectancy index, Education index and GNI index) were considered to have equal weights (1/3 each). We also used this methodology since trademarks and patents are equally important (½ each). The data for all variables were obtained from the WDI.

Trade [T] is measured by dividing total trade (exports plus imports) to GDP from the WDI. It gives us information about the economic structure of the country, regarding the degree of integration to the world economy via foreign trade. That is, the share of trade (exports and imports) in GDP can viewed as an indicator of that countries level of globalization and competition in the global economy. Foreign trade is also a channel for knowledge spillovers across national borders. That is, trade is a mean to access foreign knowledge which is embedded in the traded goods. Sometimes the imitation of this acquired new knowledge may spur innovation that will enhance economic growth.

However, it should be noted that “[d]espite the overwhelming popularity of the simple trade ratio measure, researchers should be aware that this measure is a measure of country size and integration into international markets rather than trade policy orientation … [T]he five least open countries are (in order) Japan, Argentina, Brazil, the United States, and India… While it is clear that these countries have trade restrictions in varying degrees, it is difficult to believe that they are the most restrictive countries in the world in terms of trade policies” (David 2007: 9).

ICTs [C] telephone lines, internet hosts/active Internet Protocol (IP) addresses, mobile phones, personal computers are the variables used to capture the levels and the growth rates of ICT.

In this study, in order to extend the time dimension and to incorporate all aspects of ICTs, we used the average of mobile phone subscribers per 100 people, telephone lines per 100 people and internet hosts/active Internet Protocol (IP) addresses per 100 people from the WDI.

1.2 Descriptive statistics

Descriptive statistics and correlation for the variables used in the model are provided in Tables 2 and 3.

Table 2 Descriptive statistics of the variables

Table 3 Correlation matrix of MENA and Latin American countries

Appendix 2: Individual country PMG estimation

Individual country short-run PMG estimation for MENA and LACs regions are presented in Tables 4 and 5, respectively.

Table 4 PMG results for MENA countries 1980–2014 period (Δy)

Table 5 PMG results for Latin American countries 1980–2014 period (Δy)

As can be seen from Table 4, with the exception of Bahrain and Mauritania speed of adjustment estimates (error correction terms/ECTs) of PMG in MENA countries implies long run adjustment in terms of expected signs.²⁶ However, it is important to note that even though the error correction terms of Israel and Kuwait have the expected signs they are not statistically significant. The error correction coefficients for the rest of the countries are negative and range from − 0.83 to − 0.09. So we can say that with the exception of these four countries the convergence to equilibrium in MENA varies between 9% and 83%. The speed of adjustment has been relatively fast in Turkey, Iran and Iraq during the 1980–2014 period in the MENA region.

With the exception of United Arab Emirates all countries have the expected sign for capital per efficient worker in the short run. However, the capital per efficient worker is statistically significant only in Algeria, Iran, Israel, Jordan, Qatar, Syria, Turkey, Yemen and Mauritania. In terms of the knowledge indicators trade is statistically significant in more countries compared to the other indicators.

Not surprisingly, trade has positive and statistically significant impact on the output in Saudi Arabia. In Jordan, Syria, Turkey, Kuwait and Morocco trade has negative and statistically significant impact on the output. The impact of trade is the highest on Kuwait. A unit change in trade yields a 1.1% decrease in Kuwait’s output. According to our estimation results ICTs seem to have positive and statistically significant effect on the output growth of Algeria, Iran, Jordan, Syria, Turkey and Yemen. The highest impact of ICTs in the short run is on output growth of Syria (1.3%).

As expected in the short-run compared to the other knowledge indicators R&D has the virtually no or negative impact on the economic growth performances of the countries in the region with the exception of Israel (which is also minimal but it is statistically significant). As indicated before, it takes very long time for the economy to obtain the returns of R&D investments, if it is successful. So, in sum among the three knowledge indicators only ICTs seem to have positive effect on the output growth in the countries in the MENA region.

The speed of adjustment estimates (error correction terms) of PMG has the expected signs for all countries in Latin America. However, the error correction terms are not statistically significant for Belize, Colombia, Honduras, Paraguay and Uruguay (see Table 5).

The error correction coefficients for the rest of the countries are all statistically significant and range from − 0.78 to − 0.18. This confirms the relationship between output and knowledge indicators in LACs and it varies between 18% and 78%.

While Chile has demonstrated fastest speed of adjustment the lowest speed of adjustment is observed for Bolivia, in Latin America, during the 1980–2014 period.

The results for LACs demonstrate positive sign for short run capital per efficient worker as expected. However, the capital per efficient worker is not statistically significant for Honduras, Paraguay and Uruguay. In terms of the knowledge indicators in LACs ICTs is statistically significant in more countries compared to the other indicators. With the exception of Mexico ICTs impact on output is positive, indicating that this region has an advanced ICTs infrastructure that contributes to the economic growth performance of countries.

Trade has positive and significant impact on the output for Belize, Colombia, El Salvador, Honduras and Panama during the 1980–2014 period. Interestingly, compared to MENA region in the short-run R&D has higher impact on output.