Digital Financial Inclusion and Trade Openness in Africa

The objective of this study was to test the impact of digital financial inclusion on trade openness using a panel of 16 African countries observed over a 17-year period from 2002 to 2018. T > N, this study favors a methodology based on static panel estimates using the generalized least squares (GLS) method. The results obtained revealed that only one variable (logGDP) out of the five retained has a statistically significant influence on trade openness at the 1% level (p > t = 0.06) with a coefficient opposite to the predicted sign of (−0.2371655). This coefficient shows that the decrease in national production in these countries by 0.23% leads to a decrease in the level of trade by 1%. In relation to the variable of interest (ATMs), it was found that it negatively and significantly influences trade openness at the 1% level. When the level of trade decreases by 0.6%, it means that the use of digital finance has decreased by 1%. Also, the study finds an R within of meaning 0.0527 that at 5.27% the fluctuations in trade in these economies are explained by the model variables.


Introduction
Having an account provides an entry point into the formal financial system. It makes it affordable and easy to pay and receive bills, send or receive funds. It is a place to save money, and opens the door to obtaining bank loans. Account ownership is an index of financial inclusion (Haoudi & Raibhi, 2018). For Global (Demirguc-Kunt et al., 2014), 60% of people worldwide reported having an ac- Liu, L. and Nath, H., (2013) focus on South Africa, Nigeria and Kenya. Most of these studies use primary data collected from a survey. For our part, the use of secondary data will be favored by using declarative data reported by financial institutions to regulatory agencies (Shankar et al., 2011). In addition, the above-mentioned studies use demand factors for digital financial services (purchasing power of agents, literacy rate, level of education, …) as well as socio-economic characteristics of households (age of the head of household, gender, household size, …) in the estimates. We, on the other hand, will focus on the supply factors of digital financial services (credits granted to households and the state by commercial banks, number of automated mobile money operators, size and density of the population, …) in the estimations. With this approach, our study aims to answer the following main question.

Does digital financial inclusion have an impact on trade openness for African countries?
The general objective of this study is to analyze the impact of digital financial inclusion on trade openness. Specifically, it aims to analyze the influence of digital financial services on trade openness (1), to determine the variables that accompany digital finance in enhancing trade openness (2). After this introduction, the second section is devoted to the literature review, the identification of the hypotheses and the presentation of the research model. The third section presents the methodology followed by the results and their discussion, economic and policy implications before the conclusion.

Review of the Literature
The aim of this section is to present the different theories on trade openness and digital finance. It is structured around two main paragraphs. The first paragraph presents the theories on international trade. The second paragraph presents theories on digital finance.

Theories on International Trade
Trade openness is defined as the gradual abandonment of protectionism on local industries against foreign competition (Kadid, 2015). It refers to the results of a process aimed at reducing barriers to economic exchange between nations. It refers to financial and trade liberalization. Financial liberalization is understood in the sense of free movement of capital. Trade liberalization is a set of internal and external policies aimed at removing barriers to trade in order to increase trade. They are based on the principle of physical gravitation which applies the attraction between bodies with masses. And by analogy, these models are used to explain bilateral trade, and are essentially dependent on the geographical factors and incomes of the economic partners. Apart from these theories, several authors (Smith, 1776), (Ricardo, 2007), (Heckscher, 1919) and (Moroney & Walker, 1966) have developed theories on international trade. These are the absolute advantage theory, the comparative advantage theory and the so-called (HOS) model. The HOS model explains international trade by differences in the factor endowments of each country. It is from the HOS theory that the technological gap thesis and the product life cycle thesis were born (Bourdin et al., 2009).
These theses were refuted by other authors while showing that this theory allows the economic life of a product to be rationalised by analyzing the period between its launch and its abandonment (Bonnieux & Rainelli, 2003). This last theory is much more applied in the context of marketing. It is in this sense that digital finance plays a crucial role in the business environment.

Theories on Digital Financial Inclusion
Digital financial inclusion is defined as the digital access and use of formal financial services by previously excluded or underserved populations. These services must be affordable and tailored to the needs of clients, offered in a responsible manner to ensure the sustainability of providers. Digital financial services are provided via cell phones, the internet, or cards to poor and low-income pop- The use of digital data is playing an increasingly important role in financial services and financial inclusion for high, middle and low income countries (Kumar & Muhota, 2012). Access to digital financial services by populations excluded from mainstream finance is summarized under the term "digital financial inclusion". In practice, they are provided by fintech companies and innovative financial service providers and public authorities (Gomber, Koch, & Siering, 2017). Currently, the relevance of digital finance and financial inclusion to poverty reduction and economic growth is attracting attention from policy makers and academics. This is largely due to the number of issues that persist and which, if resolved, will make digital finance work better for individuals, businesses, governments and economies (Barbesino, Camerani, & Gaudino, 2005 Figure 1 summarizes the benefits of digital finance for an economy (Cruces et al., 2020).
For an economy, digital finance reduces the costs of running the state and allows traceability in the use of finances for state sectors. Transactions are safe and fast, which reduces the administrative and financial burden on the sectors. It improves the financial system of countries by facilitating the redistribution of liquidity to all segments of the population, even those with very low incomes. It facilitates the repatriation of foreign currency by immigrants and expatriates.
These currencies help to offset the trade deficit in their current accounts and have a direct and immediate impact on the economy as a whole. However, at the same time, it has a direct and immediate impact on the economy as a whole.
Beyond these positive impacts, digital finance also has negative impacts on financial inclusion. Several actors interact in financial inclusion as well as in the use of digital financial services. Figure 2 illustrates the role of actors in digital finance and financial Inclusion (Ozili, 2018).
This diagram shows that for financial inclusion and poverty reduction to occur, the state must first invest in digital financial services. Establishing high-speed   They reduce barriers to accessing financial services such as lack of identity documents, lack of formal income and geographical remoteness. These technologies thereby reduce poverty and increase financial inclusion.

Study Variables, Hypothesis Development and Research Design
In this section we present the structure of the economies of the African countries under study. In the following, we return to the evolution of the variables under study.

Financial sector in 16 African countries
The depth and development of financial sectors in Africa generally remains weak even when per capita income levels are taken into account. They are generally small, underdeveloped and dominated by a banking sector that is highly concentrated in urban centers. With weak stock and bond markets, the intermediation role is mainly played by banks, which represent the main source of external capital for companies. Graph 1 shows the evolution of the financial sector structure of the African economies covered by this study.
The following figure shows that: Graph 1. Financial sector developments in 16 African countries. Source: designed by the author from the WDI database and processed with stata 13.0. Furthermore, this graph shows that Uganda's financial sector is underdeveloped and the most formally used inclusive financial services are provided through mobile money. This sector remains behind the regional level as Uganda has one of the lowest bank penetration rates in Sub-Saharan Africa. This is a particular constraint on access to business finance. This is why the World Bank indicated in its report that only 10% of Ugandan businesses had a bank loan or line of credit in 2013, which is less than half the average for low-income countries (22%) (Bank & Group, 2013). The economy of this country is dominated by small businesses because, the difficulty of access to credit is a major obstacle to the formation of the economy. The other countries considered have a weak, underdeveloped and small financial sector. This can be justified by the fact that in these countries, the banking sector is overrun by sectors with higher portfolio risks than in large companies. These are manufacturing, trade, real estate and construction. In addition, in these countries, the sector is overrun by cooperatives covering mostly microenterprises and micro-entrepreneurs, with SMEs being the missing link (Bank & Group, 2013). There is a low penetration of commercial banks in the country, whereas a financial sector with commercial banks that have wide geographical coverage and are able to provide credit to small and medium enterprises can have a greater impact on economic growth than a system concentrated in urban centers (Beck, Feyen, Ize, & Moizeszowic, 2008 In South Africa, production tailored to the exploitation of natural resources with the extraction of minerals is the basis for improving the terms of exchange. representing more than 30% of the export sector. In addition, there was local processing of raw materials and diversification of exports. During the same period, it recorded a deficit due to the fall in the world price of cocoa by almost 50%, the rise in oil prices and social movements and mutinies.
H2: The improvement in the terms of trade has a positive impact on trade openness.
Graph 2. Evolution of the terms of trade of the sixteen African countries. Source: designed by the author from the WDI database and processed with stata 13.0. Open Journal of Business and Management Foreign direct investment in the sixteen African countries FDI inflows to Africa amounted to US$54 billion in 2014, making it the third most attractive region among developing economies for FDI flows. Despite this ranking, Africa is nowadays the "hot spot" for FDI as it remains a high growth destination (Makoni, 2015). Between and 2002 the growth 2014, rate of FDI inflows is 267% in Africa. Africa is not only a preferred FDI destination, but also a supplier with outflows estimated at billions 13 of US dollars during the same period (Bruno, 2016). The evolution of FDI in the economies of the African countries concerned is presented in Graph 3.
From this graph it can be seen that: Among the countries considered in this study, Ghana, Kenya, Mauritius, Egypt, Equatorial Guinea, Namibia, RSA and Uganda are attracting foreign investors compared to other countries but to different degrees.
The strong attraction of FDI flows noted in Equatorial Guinea in 2010 is due to high oil and gas production, making it one of the largest recipients of foreign investment in Africa. For Egypt, the attraction of FDI is linked to the production, marketing and export of natural gas. The strong attraction in 2016 is due to the increase in natural gas discovered offshore the Nile Delta in the Mediterranean Sea.
In Ghana, the attractiveness of FDI is due to the political stability, economic and political management that has convinced most investors to move their capital to Ghana and Senegal. Ghana has thus taken the place of Nigeria and

GDP of the sixteen African countries
Economic growth in Africa is estimated at 3.4% for the year, 2019, roughly the same as in 2018. Although stable, this is below the ten-year average growth rate for the region (5%). The slower-than-expected growth is partly due to the moderate expansion of the continent's five largest countries (Algeria, Egypt, Morocco, Nigeria and South Africa), which together recorded an average growth rate of only 3.1%, compared to an average of 4% for the rest of the continent. Growth is expected to accelerate to 3.9 per cent in 2020 and 4.1 per cent in 2021 (Cuckler et al., 2018). Africa's estimated growth masks large variations between regions and countries. For the African countries concerned, the evolution of GDP can be seen in Graph 4.
The following can be seen from this graph: Four countries out of all the countries considered in this study show an in- Africa is the continent with the highest population growth (an annual growth rate of 2.5% compared to a rate of 1.12% worldwide) with a high proportion of young people. Fertility in Africa is between 6 and 7 children per woman (Gabon and CAR are exceptions with an index of 5.4). The generalized economic crisis in Africa multiplies the number of young people left behind street children and child soldiers (Tabutin & Schoumaker, 2005). Graph 5 shows the evolution of the rural population as a percentage of the total population of African economies.
Looking at this graph, the results show that: Algeria shows a growing rural population curve throughout the period under consideration. The fertility rate in rural areas is increasing significantly and regularly. This explosion in births can be explained by the improvement in living conditions, in particular better access to housing, more jobs and the improvement in the security situation following the end of the war (Breil, 2020). Botswana has a specific demography characterized by a mortality rate higher than the birth rate, i.e. 21.9% against 20.7%. It is only in the past that 2016 this country has experienced a 1.19% growth in its population. This under population of Botswana can be explained by the low life expectancy and the predominance of HIV (about 20% of the population affected by this disease) and the high rate of migration (4.5%). In Cameroon, the population curve also shows an increasing curve throughout the period in question with a variation of 1.06% in 2018. In the Central African Republic, this curve is downward. The decrease in the rural population in the Central African Republic is justified by the civil wars that have ravaged certain regions and by the permanent insecurity in others, pushing the rural population to seek refuge and work in the capital Bangui. As a result, the capital, the only city of any size in the country, is facing a large influx of Open Journal of Business and Management people 2 . Other countries such as Côte d'Ivoire, Egypt and Uganda show an increasing rural population trend over the period. Gabon and South Africa have a low percentage of rural population. In South Africa, the change in rural population over the period is less than 1. In 2018, the rural population change is 0.77% and in Gabon it is 0.44%. Libya, Kenya, Rwanda and Ghana show a curve with a constant evolution varying between 0 and 1% during the whole period concerned. Equatorial Guinea, Namibia and Mauritius show an upward curve reflecting the growth of the rural population.
H5: The rural population has a positive impact on trade openness.

Imports and exports of the sixteen African countries
Export diversification can accelerate economic growth in a sustainable way.
The majority of Africa's exports are unprocessed goods. More diversified export baskets are associated with higher growth rates. The introduction of new products into export markets is strongly correlated with long-term cumulative growth in GDP per capita (Klinger & Lederman, 2004), (Rieländer & Traoré, 2016). Graph 6 shows the evolution of imports added to exports relative to GDP in the economies of the African countries concerned.
The graph shows that: Most of the African countries considered are not very open to the rest of the world. This is the case of Rwanda, Mauritius, Ghana, Kenya, Libya, the Central African Republic, Cameroon, etc. In Rwanda, for example, beyond its dependence on international aid, the problem of being landlocked hinders its commercial Graph 5. Evolution of the rural population of the sixteen African countries. Source: designed by the author from the WDI database and processed with stata 13.0.
Graph 6. Level of trade openness of the sixteen African countries. Source: designed by the author from the WDI database and processed with stata 13.0. Open Journal of Business and Management openness. This country has no access to the sea and therefore no port. Equatorial Guinea has a better position in terms of trade with the rest of the world, thanks in particular to exports of oil, methanol and some forestry products (exotic wood) and agricultural products (notably cocoa). These exports are facilitated by the ports of Malabo and Bata.
Egypt trades mainly with France and other countries along the Mediterranean Sea, but to a small extent. This participation in trade is largely due to the diversification of its economy invaded by the manufacturing sector (16%), the real estate and construction sector (15%), the wholesale and retail trade (13%), the extractive sector (12%), the agricultural sector, the forestry and fishing sector (11%).
However, despite this potential, Egypt has a structural deficit that reflects a defi- Namibia exports fish to Spain, which accounts for 30% of its export earnings.
It is the fourth largest mineral exporter in Africa and the fourth largest producer of uranium (9% of GDP in that higher prices lead to lower demand and hence supply. Therefore, the persistence of high volatility in inflation and economic activity leads to uncertainty.
The monetary structure of the African countries covered by this study can be seen in Graph 7.
The results of this graph show that: The

Methodology
The study sample consists of 16 African countries. These are mainly eight up-

Specification of Model
To test our hypotheses, we used the model developed by Sarma and Pais (2011).
This model considers the financial inclusion index as a dependent variable. Consequently, in the empirical model, the study specifies a dynamic log-linear equation of the financial inclusion index, which includes a dependent variable lagged by one period (Sarma & Pais, 2011). This model is presented as follows:  from the export and import price index, (Ifl), represents inflation captured from the price movements of goods and services as well as that of the currency. However, the population will be considered as an error term that contains the country effect (1), time fixed effects (2) and random error (3). = i t it u u + ε + which is assumed to be independently and identically distributed with zero mean and Σv it variance 2; while u i and v it are as discussed above as in the reference model. The operationalization of these variables and the expected signs are presented in Table 1.

Presentation of the Panel Methodology
Panel data consist of a set of temporal observations on several statistical units (an individual, a company, a country, etc.). Since the number of observation periods for our study T [18] The estimator defined on this model is the Double Within.
2) The random effect model The random effect model is written as follows: This model is without specific effects and without random effects. The estimator defined is ordinary least squares (OLS).

Specification Tests on Panel Data
When considering panel data, the very first thing to check is the avec homogeneous or heterogeneous specification of the data generating process. From an econometric point of view, it amounts to testing the equality of the coefficients of the model studied in the individual dimension. From an economic point of view, the specification tests consist in determining whether we are entitled to assume that the theoretical model studied is perfectly identical for all individuals or, on the contrary, whether there are specificities specific to each individual. To test this specification, several tests can be carried out:

A. M. Mulungula, F. Nimubona Open Journal of Business and Management
The Fisher test which allows to choose between the fixed effects model and the model without effects. The Breusch and Paga (1980) test is a Lagrange multiplier test; it tests the random effects hypothesis. It is based on errors obtained by OLS. The Hausman (1978) is a general test that can be applied to many specification problems in econometrics. Its most common specification is the individual effects panel specification. It also allows to choose between fixed and random effects.

Stationarity with Panel Data
Unit root and co-integration tests on time series panel data are indeed more powerful than their analogues on individual time series in small samples. The use of panel data makes it possible to work on smaller samples (in the time dimension) by increasing the number of available data (in the individual dimension), thus reducing the probability of facing structural breaks and alleviating the problem of the low power of small sample tests. As Baltagi & Kao, (2001) note, non-stationary panel data econometrics aims to combine the "best of both worlds": the treatment of non-stationary series using time series methods is the increase in data and test power with the use of the individual dimension.

Im, Pesaran and Shi test
The tests proposed by these authors respectively in (1997, 2002 and 2003) make it possible to respond to the criticisms levelled at Levin and Lin's test concerning the independence of the error terms in the individual dimension. Indeed, these authors were the first to develop a test allowing, under the alternative hypothesis, not only heterogeneity in the autoregressive root, but also heterogeneity in the presence of a unit root in the panel. These tests consider a model with individual effects and no deterministic trend. In the absence of autocorrelation of the residuals, this model is written: IPS model: The IPS test is a joint test of the null hypothesis of unit root (  ) against an alternative unit root hypothesis. To test the performance of his test, Hadri, (2000) conducted Monte Carlo simulations. The overall result is that the accuracy of the test is higher when T and N are sufficiently large. More specifically, the size of the test Z is close to the theoretical size of 5% for T > 25. As for the power, it appears that it increases with the value of λ for all T and N.

Results
The results of the descriptive statistics are presented in Table 2,  Furthermore, credit to the private sector averages about 26% of GDP in the different countries. This shows that access to financial services is still low on average in the different countries. However, some countries have a well-developed

Results of Specification Tests on Panel Data and Stationarity
When considering panel data, the very first thing to check is the homogeneous or heterogeneous specification of the data generating process, i.e. whether the coefficients of the model studied are equal in the individual dimension.

1) Results of specification tests on panel data
The Fisher test makes it possible to choose between the model with no effects (OLS estimator) and the model with fixed effects (Within estimator). The results are obtained by performing automatically estimate the parameters of the fixed effects model and are given in Table 3.
For the sixteen African countries considered in our study, the hypothesis of absence of effects is rejected at the 5% threshold because the probability associated with the Fisher statistic is equal to 0.0000%. This test therefore suggests the use of the within estimator (fixed-effects model) because it is more efficient  The results are given in Table 4.
Through these results, we reject the hypothesis of absence of effects at the 5% threshold. This is because the probability associated with the Breusch-Pagan statistic is equal to 0.0000%. This test therefore suggests the use of the between estimator (random effects model) because it is more efficient than the OLS estimator (no effects model). The Hausman statistic, under the null hypothesis of the presence of random effects, asymptotically follows (N tends to infinity) a Chi-square distribution with k degrees of freedom. The null hypothesis of the presence of random effects is not rejected if the Hausman statistic is lower than the critical value read from the Chi-square table. Table 5 gives the results of this test.
The calculated Hausman value is 116.81 and an associated probability below the conventional 5% significance level. The null hypothesis of no correlation between the individual effects and the explanatory variable is therefore rejected.
The fixed-effects model should therefore be preferred and the within estimator should be retained. and Levin-Lin-Chu (LCC) tests. To do this, we remove these two variables from our study model. Indeed, as the six variables retained are integrated of order (0), it is not necessary to proceed to any co-integration test.   Table 7.

3) Estimation of the specified models
Through these results, we accept the hypothesis of the nullity of time effects.
Only the fixed effects are sufficient. The probability associated with the Fisher statistic (0.0000%) shows that the model is globally good. In addition, the R 2 within, which gives the share of intra-individual variability of the dependent va-  and statistically significant at the 10% level. However, when analyzing the coefficient generated by the log (GDP), it turns out that this variable has a negative influence on trade openness. This means that when production decreases by 0.23%, the trade openness rate decreases by 1%, all other things being equal. Similarly, the use of digital finance also has a significantly negative influence on trade openness at the 1% threshold. This means that when the use of digital financial services by actors in the economy of the countries under consideration decreases Open Journal of Business and Management by one point 1, the level of trade decreases by 0.6%. This proves that the use of digital financial services has a positive influence on trade between countries. The exchange rate variable has a significant positive influence on trade openness. Indeed, when the terms of exchange improve to 0.17%, the openness rate increases by 1%. Similarly, when foreign direct investment flows increase by 2.5%, trade openness improves by 1%. Finally, when the general price index improves, i.e.
when these economies are deflated to 2.7%, there is an improvement in imports and exports relative to GDP of 1%.

4) Results of diagnostic tests on residues
These are the autocorrelation test and the heteroscedasticity test. The results of these tests are given in the following. Autocorrelation is a problem that is only relevant in the case of time series. The ρ-test is the simplest test to perform for the presence of autocorrelation. The procedure is as follows (Ouellet et al., 2005): recover the residuals of the regression to be tested; regress ˆt u on 1t u − at ˆt n u − and X and test the joint significance of the coefficients of this regression by an F-test. The results of this test are given in Table 8.

The autocorrelation test
The results in the table above show that the Fisher statistic is greater than the probability attached to it. Therefore, there is autocorrelation of the errors; it is therefore necessary that an autocorrelation correction is made for this model. Table 9 gives the results of the correction.
Using the "xtpcse" command, we find that the Wald statistic is higher than the probability which is below the 5% threshold. As a result, we reject the hypothesis that the errors are correlated. The good thing is that when we analyze the results of the test that corrects for error autocorrelation, we find that only the logarithm of GDP per capita positively and statistically explains trade openness at the 1% level for those African countries considered in this study. However, all variables retained their predictive signs except for the logarithm of automated mobile money operators per 1000 inhabitants (−0.00644*) meaning that when the use of digital financial services decreases by 1%, trade openness decreases by 0.64%; and the logarithm of foreign direct investment flows (−0.0048547*) meaning that when FDI flows decrease by 1%, trade openness also decreases by 0.48%. Previously, we hypothesized that these two variables should positively influence trade openness.
In the context of a heteroskedasticity test, the null hypothesis is that all coefficients of the squared residuals regression are zero, in short, there is homoscedasticity. The results summarized in Table 10 shows that, for the model under study, the null hypothesis of homoscedasticity is rejected because the probability associated with the Chi2 statistic is less than 5%. It can therefore be concluded that heteroscedasticity is present, i.e. that the variance of the error, with respect to each equation, is not constant over time. It then becomes necessary to perform another test to try to obtain more information on the form of the heteroscedasticity. Modified Wald is a heteroskedasticity test designed to test the specific hypothesis of inter-individual homoscedasticity. It is essentially an F-test which, under the null hypothesis, assumes that the variance of the errors is the same for all individuals. The statistic follows a χ 2 distribution with degree of freedom N. Table 11 gives the results of the modified Wald test.
Since the probability associated with the Wald statistic is less than 5%, the null hypothesis that the error variance is not constant is rejected. By accepting the alternative hypothesis, it is appropriate to conclude that the error variance is the same for all individuals. Since we had already concluded that heteroscedasticity is present in some form with the previous test, this rejection of the null hypothesis does not allow us to specify the structure of the heteroscedasticity further. We remain with the conclusion that there is heteroscedasticity for all i, t.

5) Policy implications of the results
Based on the analysis, our study concludes that digital financial inclusion has a negative and statistically significant impact on trade openness. These results are consistent with those of (Macmillan et al., 2016) who confirmed that the number of payment instruments in an economy creates inflation. These results corroborate with the reality of the evolution of imports and exports in the African countries presented in the stylized facts. Despite the high penetration of mobile phones and internet in these African countries, most of them have less diversified economies and are mostly landlocked and without access to the sea and ocean. Furthermore, the populations in these countries are numerically included but not financially for the most part. That is, they use the internet but not digital financial services. This significantly reduces exports and imports from these countries. This leads us to make the following recommendations: For a more open economy, government strategies need to put technology investment issues at the heart of strategies and therefore make them a priority. They should diversify the tools in the use of digital finance to facilitate financial inclusion. They must work towards the creation of public-private partnerships between the banking sector, technological finance operators and the public authorities with a view to moving towards exports that value local raw materials. This would enable them to rehabilitate and create new industrial zones. The digital transfer of capital can radically change the structures of economies and thus accelerate growth.
Similarly, these countries need to build the technological capacity of local financial institutions (banks and other financial institutions) to enable them to carry out capital transfer operations using digital finance. This allows countries to compete with the rest of the world by increasing the flow of trade and thereby financial inclusion.
As inflation hinders or paralyses trade between countries, a deflationary policy must be implemented to restrict the volume of money held in virtual wallets, with the aim of restoring or maintaining the value of money. This is because, as the theory has shown previously, every time an electronic wallet is created, there is an automatic risk of inflation. It is also necessary to fight against imported inflation due to cost increases resulting from the rise in prices of imported goods, whether raw materials, semi-finished goods or finished products. This must be done by the institutions in charge of money management in collusion with the technology finance companies that regulate the price in relation to the use of digital finance. To reduce barriers to entry, governments need to deregulate the goods market by increasing competition. This should reduce prices and increase innovation and thus productivity. Indeed, by boosting factor productivity in the long run, this policy would reduce competitiveness gaps between countries as it would have the same effect on unit labor costs as a wage cut.
These countries need to address the problem of exchange rate depreciation. Indeed, when the exchange rate depreciates, the impact on the price of imported products, expressed in local currency, may be less than the change in the exchange rate. This is the case if part of the price increase is absorbed, for example, by a fall in the margins of intermediate (importing) companies. The depreciation of the exchange rate would reduce the use of digital finance. In this perspective, these countries need to apply the trade facilitation policy. This policy simplifies and lowers the costs of trade transactions. This makes trade activities more efficient, more predictable, and based on internationally accepted norms, standards and best practices. These practices should take into account the use of digital finance as a favorable means of transferring capital internationally to facilitate trade.

Limitations and Future Prospects
Although the results of our study are satisfactory, they are not without their limitations. The first limitation of our study relates to the nature of the data used to make different estimates. Indeed, by using secondary data, it was difficult to capture the opinions of users of digital financial services on the real impact of digital financial inclusion on trade openness. To do so, we would have had to either use primary data to allow users to express their views, or use mixed data that includes both primary and secondary data. The second limitation is related to the size of the individuals (i) observed in our study. Our study considered 16 African countries, which does not represent half of the countries on the continent. Therefore, given the small size of the countries observed, we could not generalize our results to the entire African continent. In addition, the choice of this country did not follow any objective criteria, but was dictated by the availability of data for each country. Therefore, the use of primary data would be advantageous because it does not exclude some countries to the advantage of others; all countries will have the chance to be selected. Moreover, in the analyses, our study did not take into account all the explanatory variables of trade openness such as the languages spoken between the countries considered, the distances separating them, the mineral wealth, the religions.
However, these limitations do not diminish the scientific value of this study and, to this end, our results remain scientifically valid. Therefore, to address these limitations, we intend to conduct a future study on the impact of digital financial inclusion on trade openness in Africa by expanding the number of countries (all of Africa) and increasing the period of analysis. Also, in this study, the estimates will combine qualitative and quantitative data (primary and secondary) while making use of dynamic panel estimates. Similarly, it is possible to analyze the adoption of digital financial services among small, medium, and large firms in Africa with the aim of understanding how digital financial services make business easier for firms. Ultimately, we plan to conduct an impact study of the digital financial inclusion of banks on the behavior of their customers in Africa. This study will provide a clear understanding of the relationships between banks and their customers once they are financially and digitally included.

Conclusion
One of the objectives of this study, which covered a set of 16 African countries (8 low-income and 8 upper-middle-income), was to measure the effects of digital A. M. Mulungula, F. Nimubona Open Journal of Business and Management financial inclusion on trade openness in these countries. Using the panel data specification, the Fisher test allowed us to retain fixed effects instead of the no-effects model and the random effects model. The within estimator (GCM) yielded a R 2 within of 0.527, which means that 5.27% of trade fluctuations in these African countries are explained by the explanatory variables of the model.
On the other hand, the results of the estimation of fixed effects on all the variables of the model revealed only one variable that positively explains and statistically significant trade in these countries at the 10% level; this is the log of national output (logGDP/capita). Other variables, logATMs, log terms of trade (logTe), log foreign direct investment flows (logIDE) and log inflation (logIfl) influence significantly but not statistically the trade movements at the 1% threshold, for logATMs with a negative sign contrary to the predictive sign, logTe with the same positive sign predicted beforehand and at the 5% threshold for lo-gIDE and log Inflation which has kept the same positive sign predicted beforehand. However, we do not believe that we have captured all the contours of this theme, particularly by considering all the necessary variables, given the unavailability of data either for certain countries or for all the countries under study.
Despite this, this does not in any way diminish the quality of the results obtained. We therefore propose to return to this theme by using not only secondary data but also and above all primary data in order to understand the opinions of users of digital financial services. This will also allow us to integrate other elements and aspects of analysis not taken into account, especially by focusing on the financial inclusion index.