Eric Boccaccio and Jhaelle Payne

3/30/21

Abstract

           This study examines the correlation between Trade Openness and Entrepreneurship in 5 African nations: Angola, Egypt, Madagascar, Morocco, and Sudan. While previous research has shown a positive correlation between these factors in other regions (Bayar et al. 2018), this is the first study that exclusively focuses on low to medium-low income African countries. Due to limitations in data availability, the design of this study is cross-sectional, meaning that it only examines data for one time point, 2018.

           Using a linear regression model, we found that—holding poverty, average years of education, and law and order constant—a 10% increase in trade openness will lead to a 5.75% increase in entrepreneurship on average. Although we found a positive correlation, there is not enough data available to reject the null hypothesis that no correlation exists between these two factors in Africa.

Future study requires more abundant and frequent regional entrepreneurship data, as more data points are needed to judge whether or not the correlation found in this study is significant.

Keywords: Entrepreneurship, Africa, Trade, Trade Openness

 

Introduction

           This study analyzes the correlation between trade openness and entrepreneurship in 5 African nations: Angola, Egypt, Madagascar, Morocco, and Sudan.

In recent years, there has been an increase in cooperative trade deals between African and non-African countries. Often, these deals are made by African leaders with the intent to stimulate economic growth. To citizens, the commonly cited justification for trade openness is that a better economy creates a more upwardly mobile populace.

To test this theory, we’ve decided to focus on one method of attaining upward mobility, entrepreneurship. We aim to examine how trade openness correlates with entrepreneurship using data from 5 African nations. We control for education, strength of governance, and poverty levels.

 

Literature Review 

The Relationship Between Trade Openness and Entrepreneurship 

This study analyzes the correlation between trade openness and entrepreneurship in 5 African nations. While past scholarship has shown these two factors to be positively correlated, the African region has not been a focus of previous work. Bayar et al. (2018) studied 15 countries worldwide from 2001-2015 and found that “banking sector and capital market development, FDI inflows, and trade openness affect the total early-stage entrepreneurial activity positively.” 

Notably, of the 15 countries that Bayar et al. considered, only one (South Africa) was located in Africa. The study also was exclusively limited to upper-middle and high income countries, whereas all 5 countries in this study were classified as either lower-middle or low income at the time of data collection. 

Importance of Studying Entrepreneurship 

Entrepreneurship has long been thought to be a key method of attaining upward socioeconomic mobility. While there are few studies examining the relationship between entrepreneurship and socioeconomic development in Africa (Nafukho and Helen Muyia 2010), studies in Latin America have shown that “entrepreneurship is a channel of intergenerational mobility,” with most of this mobility being upward (Castellani and Lora 2014). 

Entrepreneurship has also been shown to drive overall economic growth in Africa. Adusei (2016) compared results from 12 African countries and found that “entrepreneurship positively explains the variations in the growth of the study countries.” He goes on to contend that “entrepreneurship in developing economies including Africa, even if replicative, is instrumental to economic growth.” 

Why focus on Africa? 

Africa is an important region to study in this context due to its recent rapid increase in entrepreneurship levels (Singh and Belwal 2008). This increase is especially intriguing given Africa’s colonial history. By the end of the European imperial era, the region had been stripped of much of its culture, resources, and governmental structures (Halkias, et al. 2011). 

After the imperial powers exited the region, Africans were forced to rebuild their societies with very little resources. Some African countries have thrived, while others have yet to fully recover. Therefore, it is important to understand what factors contribute to African economic and socioeconomic growth. In this study, we examine the relationship between two known predictors of economic growth: how trade openness correlates with entrepreneurship. 

Trade Openness 

In the past, trade openness has been shown to correlate with a variety of factors. Le Goff and Singh (2013) studied the relationship between trade openness and poverty in a panel of 30 African countries from 1981 to 2010. After controlling for factors such as education and strength of governance, they found that “trade openness tends to reduce poverty in [African] countries where financial sectors are deep, education levels high and governance strong.” This study and Le Goff and Singh’s work use similar experimental methodology and control factors. 

To test the impact of trade openness on economic growth in Africa, Brueckner and Lederman (2015) studied 41 African nations and found that “trade openness has a significant positive effect on economic growth.” Controls included GDP growth of trading partners, civil conflict, and even amount of rainfall. 

 

Methodology

This study is designed to determine whether a correlation exists between trade openness and entrepreneurship levels in 5 African nations: Angola, Egypt, Madagascar, Morocco, and Sudan. The study controls for poverty, education, and strength of governance. 

Data 

 

 

 

Table 1 — Variable Definition and Sources 

Variables	Description	Data Sources
Trade Openness  Independent Variable	Sum of imports and exports as a share of GPD	World Bank Financial Statistics
Entrepreneurship  Dependent Variable	Proportion of the population actively engaged in starting or running a new business	Global Entrepreneurship Monitor (GEM)
Poverty	Percentage of the population living below the $1.90/day poverty line	Euromonitor International Passport: Economies and Consumers
Law and Order (Strength of Governance)	Strength and impartiality of the legal system, and popular observance of the law. Values range from 0 to 6, with a higher figure indicating a higher quality and more enforced legal system	International Country Risk Guide (ICRG)
Education	Population-weighted average years of formal education	Global Health Data Exchange (GHDE)

 

Data Sources: 

All data was collected from established datasets. Data on trade openness was collected from the World Bank. Entrepreneurship data comes from the Global Entrepreneurship Monitor (GEM). Data on education was collected from the Global Health Data Exchange (GHDE). To measure quality of governance, we used a law-and-order indicator from the International Country Risk Guide (ICRG). Poverty rates were sourced from Euromonitor International’s Economies and Consumers dataset. Raw data can be found in the appendix of this report.

Measurement Procedures: 

Trade Openness (independent variable/input): 

In economics, there are two accepted ways to measure trade openness. The first method is policy based; a measure of the openness of a particular country’s trade policy. The second method is outcome-based, which boils down to the amount of trade taking place across a given country’s borders (Spilimbergo et al. 1999). For the purpose of this study, we found an outcome-based measurement to be the most practical option. Not only is outcome data easier to obtain, it also provides a more complete view of each country’s total level of trade in the market. 

The most common outcome-based measurement of trade openness is the sum of imports and exports divided by a country’s GDP, and this is the method used here. Data is provided by the World Bank. Because the latest available entrepreneurship data for the selected countries is from 2018, trade openness data will also be taken from that year. 

Entrepreneurship (dependent variable/output): 

The Global Entrepreneurship Monitor (GEM) measures entrepreneurship levels through use of survey data. 

The GEM bases results on two surveys, the Adult Population Survey (APS) and the National Expert Survey (NES). The APS has a sample size of at least 2,000 working age adults in each country and asks about “entrepreneurial activities, attitudes, motivations and ambitions” (Bosma et al. 2020). The NES samples responses from at least 36 national experts in each country who provide insight on government policy, infrastructure, availability of entrepreneurial education, and more. 

Based on the survey responses, a Total Early-Stage Entrepreneurial Activity (TEA) index is created. Each country is assigned a TEA value for a given year. The TEA is defined as “the proportion of the working-age adult population actively engaged in starting or running a new business.” In this case, any business younger than 42 months is counted as new. 

Because the 2019 GEM report includes fewer African countries, data is sourced from the 2018 report. 

Control Variables: 

To attempt to isolate the relationship between our two main variables, we have chosen to control for 3 additional factors. 

Poverty is measured using the poverty headcount index, which is the percentage of a country’s population that lives below a set poverty line. The current International Poverty Line is $1.90 USD per day accounting for differences in purchasing power between currencies. 

Educational attainment is measured by average years of formal education among the country’s populace and is population weighted. 

Strength of governance takes into account the strength and impartiality of the legal system, and popular observance of the law. The ICRG uses quantitative analysis, forecasting, and data series to measure political risk by country. 

These 5 data sources pay particular attention to their methods for obtaining data because developing countries lack statistical infrastructure, which can lead to inaccurate results. 

Statistical Method 

To estimate the relationship between entrepreneurship level and trade openness, all other factors in the compiled dataset were controlled. To achieve this, linear regression was used to measure the exact effect a 10% increase in trade openness has on entrepreneurship levels, given all other variables are held constant. Then, a scatter plot displayed the residual values compared to the fitted values, to understand the accuracy of the model. 

 

Results

Figure 1

Summary Statistics of Numerical Variables

Figure 1 displays the mean, standard deviation, minimum, maximum, median and range of the numerical variables. The high standard deviation in Trade Openness, Entrepreneurship Levels, and Poverty indicates that the data points are more spread out. Also, in Trade Openness, Entrepreneurship, and Poverty, the range values are particularly high which highlights the vast political, economic, and social differences between the observed countries.

 

Figure 2

The Correlation Between Entrepreneurship Levels and Trade Openness

Figure 2 displays a visual representation of the correlation between Trade Openness and Entrepreneurship by Country. Given that there is not a linear trend between the two variables, when performing linear regression, a logarithmic transformation will have to be used on the data values.

 

Figure 3

The Predicted Percent Change in Entrepreneurship Level for A 10% Increase in Each Variable, Holding the Other Variables Constant

In Figure 3, our final model is as follows:

log(Entrepreneurship) = .575*log(Trade Openness) + -.01*Poverty + -.22*Average Years of Education + -1.56*Law and Order.

Holding poverty, average years of education, and law and order constant, a 10% increase in trade openness will lead to a 5.75% increase in entrepreneurship on average. This exhibits a positive correlation between trade openness and entrepreneurship levels within Angola, Egypt, Madagascar, Morocco, and Sudan. Logarithmic transformation was used because the data values are nonlinear. A logarithmic transformation can transform a skewed dataset into a normalized one, where the shape of the values become closer to a normally shaped bell curve. Only the log of the values shift, not the raw data. This transformation also allows the model to produce the smallest error possible when predicting values while preventing overfitting.

 

Figure 4

The Residual Values Compared to the Fitted Values

In Figure 4, the scatter plot demonstrates the accuracy of the predicted values. In which, the model was able to predict the values with high accuracy. This is shown through the lack of outliers and its residual value of 0, meaning there is not a numerical difference between the fitted and predictor values.

 

Discussion and Limitations 

Variables 

Variables were chosen based on Le Goff and Singh’s (2013) work on the correlation between trade openness and poverty. Le Goff and Singh found that “trade openness tends to reduce poverty in [African] countries where financial sectors are deep, education levels high and governance strong.” Therefore, this study controls for poverty levels, education levels, and strength of governance (law/order). Because our instrument only allows for three control factors, we were unable to include financial sector strength as a control. 

Data 

There are a few limitations stemming from the data used in this study, mostly having to do with age and availability. 

Total Early-stage Entrepreneurial Activity 

To create a valid cross-section, we needed to make sure that datapoints for each country were taken from the same year. While the Global Entrepreneurship Monitor does contain entrepreneurial activity data for 20 African countries, not every country is measured annually. In the past 5 years, the year containing the highest concentration of data from African countries in a single timepoint was 2018; with data from Angola, Egypt, Madagascar, Morocco, and Sudan. Therefore, our cross-section was limited to a maximum of 5 countries. 

The lack of consistent annual African entrepreneurship data is also the reason why a time-based model, such as the one used by Bayar et al. (2018), was unfeasible in this study. 

Average Years of Education 

Although every other factor considered in this study uses data taken in 2018, the Global Health Data Exchange only has relevant education data available up to 2015. Therefore, the education data used in our model is about three years older than the rest of the data. While we cannot ensure that this data remains accurate in 2018, we can test the degree to which education data varies in the time leading up to 2015. In the three-year period from 2012-2015, mean years of education for our 5 countries deviates by about 0.13 years on average. 

If this trend continued for the following three years, we can assume that the mean years of education in 2018 should be roughly similar to 2015 plus or minus a few months. This relatively small degree of uncertainty is unlikely to have changed our results significantly. 

Percent of Population Living Below International Poverty Line ($1.90 a Day) 

The poverty data used in this study is 2018 estimates from Euromonitor International’s Passport. Euromonitor estimates this data using previous poverty estimates released by the World Bank. The methodology that Euromonitor uses to calculate these yearly estimates, however, is unclear. This being said, Euromonitor International is a trusted, well established market research corporation. 

Model/Research Design 

This study is also limited in the sense that it provides a cross section of variables at only one time point, 2018. Because we were unable to model these variables over time, it is possible that the correlation we found between entrepreneurship and trade openness in 2018 doesn’t exist at other time points. This being said, we estimate the probability of 2018 being an outlier to be small, especially given that Bayar et al. (2018) found a similar relationship between these two variables worldwide in a study spanning 14 years. Given the differences between the countries studied here and the ones included in the work of Bayar et al., however, it is still worth noting the limitations of having used a cross-sectional method. 

Also, through performing a backwards stepwise regression, we found that our Law and Order variable would have been a better indicator of Entrepreneurship Level. A backwards stepwise regression starts with a model containing every variable, then at each step it removes a variable to find a regression model with the least number of variables needed to explain the data.

Figure 5

Training the Model 

Figure 5 demonstrates the backwards stepwise regression testing a certain number of variables at a time in the model. The column “nvmax” displays that number of variables. The Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) demonstrates that one variable in the model will best explain the data. Specifically, the lower RMSE and MAE value is, the lower the bias. This will ensure that our model is as accurate as possible because lower bias is an indication of higher accuracy.

Figure 6

The Best Set of Variables for the Model 

After training on the data, figure 6 demonstrates which variables will best explain the data. In which only Law and Order was selected by placing an asterisk under the variable.

 

Significance, Future Study, Policy Recommendations 

Because of our limited number of data points, it’s impossible to reject the null hypothesis that trade openness has no effect on entrepreneurship in Africa. Therefore, the initial results found here need to be validated by further study. This being said, our results do show trade openness and entrepreneurship to be positively correlated. This is the case even when accounting for our control factors of education, poverty, and law/order. 

In the future, yearly entrepreneurial activity estimates will be needed for more countries in Africa. This way, a longitudinal study that includes more datapoints can be conducted. A study with more complete data would be able to determine with more certainty if a significant correlation exists between trade openness and entrepreneurial activity.

Also, given our findings that Law and Order best explains entrepreneurial activity in this dataset, it will certainly be worth studying the relationship between these two factors in future as well.

Until more complete data is collected and analyzed, no political or policy recommendations can be made to the leaders of Africa regarding trade openness and entrepreneurship. This being said, trade openness and entrepreneurship are both known indicators of economic growth as discussed in detail in the literature review portion of this report. Therefore, it would likely be beneficial for African leaders to continue the push toward a more globalized economic structure.

 

Appendix — Merged Data

 

Country	Trade Openness (%)	Entrepreneurship (%)	Poverty (%)	Education (Avg. Years)	Law and Order (0-6)
Angola	66.37801194	40.84	10.7	4.94	2.5
Egypt	48.27827106	9.84	1.2	7.45	3
Madagascar	62.50180122	20.74	61.5	5.54	2.5
Morocco	87.97578268	6.65	1.1	3.73	4
Sudan	17.92676158	22.17	3.9	4.61	2.5

 

 

References

Adusei, M. (2016), Does Entrepreneurship Promote Economic Growth in Africa?. African Development Review, 28: 201-214. https://doi.org/10.1111/1467-8268.12190 

Bayar, Y., Gavriletea, M., & Ucar, Z. (2018). Financial Sector Development, Openness, and Entrepreneurship: Panel Regression Analysis. Sustainability, 10(10), 3493. MDPI AG. Retrieved from http://dx.doi.org/10.3390/su10103493 

Bosma et al. (2020). Global Entrepreneurship Monitor 2019/2020 Global Report. Global Entrepreneurship Research Association. 

Brueckner, M. and Lederman, D. (2015), Trade Openness and Economic Growth: Panel Data Evidence from Sub‐Saharan Africa. Economica, 82: 1302-1323. https://doi.org/10.1111/ecca.12160 

Castellani, Francesca, & Lora, Eduardo. (2014). Is Entrepreneurship a Channel of Social Mobility in Latin America?. Latin American Journal of Economics, 51(2), 179-194. https://dx.doi.org/10.7764/LAJE.51.2.179 

Halkias, D., Nwajiuba, C., Harkiolakis, N. and Caracatsanis, S.M. (2011), "Challenges facing women entrepreneurs in Nigeria", Management Research Review, Vol. 34 No. 2, pp. 221-235. https://doi.org/10.1108/01409171111102821 

Le Goff, M. & Singh, R.J. (2013). Does Trade Reduce Poverty? A View from Africa. Policy Research Working Paper 6327, The World Bank, Poverty Reduction and Economic Management Unit. https://doi.org/10.1596/1813-9450-6327 

Nafukho, F.M. and Helen Muyia, M.A. (2010), "Entrepreneurship and socioeconomic development in Africa: a reality or myth?", Journal of European Industrial Training, Vol. 34 No. 2, pp. 96-109. https://doi.org/10.1108/03090591011023961  

Singh, G. and Belwal, R. (2008), "Entrepreneurship and SMEs in Ethiopia: Evaluating the role, prospects and problems faced by women in this emergent sector", Gender in Management, Vol. 23 No. 2, pp. 120-136.  

Spilimbergo, A., Londoño J.L., Székely, M. (1999), Income distribution, factor endowments, and trade openness, Journal of Development Economics, Volume 59, Issue 1, pp. 77-101, ISSN 0304-3878. https://doi.org/10.1016/S0304-3878(99)00006-1. 

 

Appendices

 

Jhaelle and Eric split the final report. Jhaelle wrote the Results, Methods and Keywords section. Eric wrote the Abstract, Introduction, Literature Review, Discussion and Limitations, and Significance, Future Study, Policy Recommendations sections. Jhaelle also helped with a portion of the Literature Review and Discussion and Limitations sections. Jhaelle created Figures 1-6, and Eric created Table 1 and the Appendix. Eric sourced and compiled the data, and Jhaelle created the linear regression model.