1. Introduction

The last two decades have seen a significant policy shift in sub-Saharan Africa (SSA), from inward-looking import substitution to outward-looking, market determined strategies. The reasons for this shift have to do mainly with the inefficiencies of import substitution and the success of exporting in promoting economic growth (Greenaway et al., 2005). In fact, outward orientation has been held up as a potential source of growth booms for SSA economies. At a macroeconomic level, export is a means to generate foreign currency inflows required to finance imports as well as social development projects. At a microeconomic level, many arguments in favour of the export-market participation have been put forward. The participation in foreign markets is thought to improve firms’ efficiency through three major channels. First, the larger international market allows the exploitation of economies of scale that raises productivity. Second, international contacts foster a learning process through technology and knowledge spillovers (Clerides et al., 1998; Baldwin and Gu, 2003). Third, and as argued by Bernard and Wagner (1997), another link running from exporting to success stems from the more nebulous notion of international competition. Here, increasing export intensity leads to increased management efficiency due to competition abroad.

Several studies have examined the relationship between exporting and enterprises performance (see e.g. Aw and Hwang, 1995; Bernard and Jensen, 1999; Girma et al., 2004; Hansson and Lundin, 2004; Greenaway and Yu, 2004). Regardless of the data examined and methodology used, these studies find that exporters are more productive compared with non-exporters. There are two competing hypotheses regarding this result. First, enterprises self-select to become exporters, i.e., they are more productive before they enter export markets. Hence, exporting involves sunk costs and greater productivity is required of enterprises that become exporters.¹ The self-selection hypothesis is addressed empirically by looking at performance characteristics in the period prior to exporting using an export premium measure. The second hypothesis is that better performance of exporting enterprises may arise from the exporting process itself, through a type of ‘learning by exporting’ experience. The case-study evidence (e.g. Crespi et al., 2008) suggests that firms learn about new techniques and methods from the experience of exporting. The learning-by-doing hypothesis is addressed empirically by looking at performance characteristics of exporters compared with non-exporters in the period following their entry into export markets, again using an export premium measure.

Empirical investigations of the links between exporting and enterprise performance in SSA have emerged (see e.g. Zeufack, 2002; Söderborn and Teal, 2003; Bigsten et al., 2004; Van Biesebroeck, 2005; Rankin et al., 2006). However, and to the best of our knowledge, almost nothing is known about the relationship between export-market destination and the performance characteristics of SSA manufacturing firms. Yet, export-market destination matters. As noted by Ruane and Sutherland (2005) and Eaton et al. (2011), exporting ‘globally’ to unfamiliar markets where social, economic, and legal structure are different from those normally faced may really be what exposes the enterprise to competitive pressures and greater learning opportunities. Furthermore, export destination in relation to performance characteristics of enterprises is an important empirical question for informing SSA policies that seek to promote growth through exporting and diversification of trade partners, but without a coherent and coordinated strategy.

In sum, research on exports and enterprise performance has mainly been confined to the developed countries, Latin American and Asian countries but much less in the case of SSA countries. Yet, outward orientation and efficiency gains are of particular interest in SSA where the governments have adopted the export-led growth approach as their main development strategy, and juggle multiple trade partners without a clear vision. Moreover, one element of export behaviour of SSA manufacturing firms not discussed widely in the literature relates to the role of export destination. Our analysis is centred on three questions. First, do exporting enterprises exhibit evidence of superior performance relative to non-exporters? Second, if so, what are the sources of the relatively good performance traits of exporting firms? Finally, does the export market destination matter? The first two questions have been previously addressed by others. The last question constitutes the main contribution of the current paper.

In this paper we use a novel manufacturing firm-level survey dataset of 2009 in 19 SSA countries and the main findings of the paper are the following. SSA manufacturing firms that export differ from those that do not along a variety of dimensions: they are larger as measured by employment and output, more productive, more capital intensive, and pay higher wages. Moreover, foreign ownership and past export experience enhance the propensity to export, while continuing exporters outperform switching exporters. Coming finally to the role of export-market destination in the relationship between enterprise performance and exporting, three basic pieces of evidence emerge from the dataset. Firstly, in terms of improvements in total factor productivity (TFP), China is an excellent destination. Our estimates suggest that firm-level TFP increases by 0.21% for each 1% increase in China’s export intensity. Secondly, exporting to India enhances the wage rate, labour productivity and TFP. Quantitatively, a 1% increase in the share of output exported to India increases the wage rate, labour productivity and TFP by 0.28%, 1.03% and 0.11%, respectively. Thirdly, the South Africa export-market destination impedes capital intensity, i.e.,. a 1% increase in the share of output exported to South Africa decreases the capital/labour ratio by 0.60%. Finally, the scale of exporting, and not merely the export destination pattern, influences the success of exporters.

The rest of the paper is organized as follows. In Section 2, we outline the empirical model. In Section 3, we describe the main characteristics of the data and present some basic evidence on the magnitude of performance differences between exporters and non-exporters. In Section 4, we provide empirical evidence. The concluding remarks are provided in Section 5.

2. Methodology

First, we provide evidence on the magnitude of the performance gap between exporters and non-exporters for a variety of firm attributes. To measure the export premium, if any, we follow the methodology introduced by Bernard and Jensen (1995) and Bernard and Wagner (1997). The model searches for an export premium, as a measure of the superiority of exporters relative to non-exporters, in terms of firm characteristics and performance. Our general estimating equation takes the following form:

where i, j, and r index firms, industries and regions, respectively, and y denotes the log of the performance characteristics examined to determine if there is a premium between exporting and non-exporting firms: capital intensity, output (sales), wages, labour productivity (value added per employee), and TFP. The premium is captured using a dummy variable, E_i, to reflect the current export status of the firm (0 for non-exporter, 1 for exporter). The literature has been particularly interested in the coefficient on E_i. The rationale behind this specification is that exporting leads to better performance. The export premium coefficient (β₁) thus captures the average percentage difference between exporters and non-exporters in the same industry and country. In particular, the export premium is computed from the estimated coefficient β₁ as, 100(exp(β₁) − 1) (see Wagner, 2007). S_i is the dummy variable for size that takes the value of one when the number of employees is above the median employment level across all firms in each industry and country, zero otherwise.² D_j and D_r represent the respective full sets of two-digit industry and region dummies, so that any effect common to industries and regions is controlled for, and ε_i is random error term.

We also investigate whether export intensity makes a difference. For that we employ the specification that allows for an export premium that varies with export intensity. In particular, we consider the following specification:³

where ES_i is the share of exports in total output. We use fixed effects regression techniques to estimate (1) and (2) separately for each of the five enterprise performance characteristics.

The previous development will help provide evidence on whether exporters perform better than non-exporters. We pursue this issue further and perform an analysis of the determinants of export behaviour. Following Sterlacchini (2001), Castellani (2002), and Girma et al. (2004), we perform both a test on the determinants of the probability of exporting and of the intensity of exporting activity. We model the export behaviour of a firm specifying the following PROBIT equation for the probability of being an exporter:

In Eq. (3) the dependent variable takes the value of 1 if a given firm was exporting in 2009 and zero otherwise. The explanatory variables are the firm size (S_i) proxied by the dummy variable for size that takes the value of one when the number of employees is above the median employment level across all firms in each industry and country, zero otherwise; TFP; the cost of labour per worker (W_i); the ownership status of the firm (OW_i) measured using a dummy variable which equals 1 for foreign firms and 0 for indigenous firms⁴; previous exporting experience (PE_i) measured using a dummy for whether the firm exported in the previous year in order to capture the effects of persistence in export behaviour; and the age of the firm (AG_i). The error term is u_i. Regarding the use of an appropriate econometric method, in Eq. (3) the dependent variable is a (0/1) dummy suggesting a discrete response model. In fact, we use the PROBIT regression. The dependent variable in Eq. (3) is also measured using the export intensity defined as a variable between 0 and 1. The popular approach here is the TOBIT regression, i.e., the TOBIT model is used because the dependent variable (export intensity) is quantitative and limited. But, for the purpose of comparison we also use the LOGIT model. Wagner (2001) and Castellani (2002) noted that the TOBIT estimation method is not designed for variables which by definition are bounded between 0 and 1. Therefore, a more appropriate model is the quasi-likelihood estimation method (QLM) for fractional responses developed by Papke and Wooldridge (1996).⁵

The third issue to be examined in the paper is associated with the impact of export behaviour on the performance characteristics of enterprises. The UNIDO sample allows a classification of firms according to their trajectories in the domestic and export markets over the period 2006–2010 into four groups: continuing exporters, non-exporters, entering exporters, and exiting exporters. In particular, we address the question of whether more productive firms become exporters (an empirical model of self-selection in export markets). To examine the relationship between the exporting and subsequent enterprise performance, we follow previous literature (see e.g. Farinas and Martin-Marcos, 2007) and estimate the following equation for each of the performance characteristics in levels,

where y_i is the level of the performance characteristics of enterprises.⁶ CX_i is a dummy variable equal to 1 if the enterprise exported continuously (export persistence) and 0 otherwise. SW_i is a dummy variable equal to 1 if the enterprise switched export status, whether the enterprise entered the export market or exited from it. In particular, the dummy variable takes the value 1 if the enterprise either entered or exited the export market. Hence, it does not distinguish between the two, but simply defines those enterprises that ‘switched’ export status. The coefficients β₁ and β₂ thus capture the increase in growth rates of performance characteristics for exporting and switching enterprises, respectively, relative to those that remained non-exporters. Z_i is a vector of enterprise characteristics e.g. (i) a dummy variable for size (this dummy variable is equal to 1 if the enterprise employment is greater than 45), (ii) a dummy variable indicating if the enterprise is owned by foreigners, and (iii) the R&D intensity variable. D_j and D_r are as previously defined. Eq. (4) will be estimated separately for each of the enterprise characteristics using standard cross-section regression techniques.

The last issue to be explored is related to the connection between export destination and intensity premium. As UNCTAD (2002) notes, successful exporting involves more than just increasing international market shares, because greater export diversification, reflected by changing export destinations, could be an indication of the improved export propensity of enterprises. Moreover, and as argued by Ruane and Sutherland (2005), by incorporating export destination patterns and intensity into the analysis of export premia, account is taken of important components of the export behaviour of enterprises.

In sum, the number and type of export destination markets to which enterprises ship their output can be viewed as proxies for the strength of export activity. This issue is of considerable interest to SSA countries, as their export promotion strategies attempted to expand exports beyond the traditional and primary destination, the EU market. Following the methodological approach established by Ruane and Sutherland (2005),⁷ and restricting the regression sample to the firms that exported in 2009, the estimating equation takes the following form:

where y_i is defined as in Eq. (4), ED_i is the proportion of exports shipped to various destinations. For instance, if β_i > 0, then exporters to a specific market destination have superior performance characteristics. ES_i is the share of exports in enterprise output, and captures the expected premium accruing to enterprises that export more intensively than others. As before, we control for enterprise characteristics (the ‘Z’ variable) e.g. size, age, foreign ownership, R&D intensity as well as sector and region effects. For each export market destination, we use the seemingly unrelated regression model (SURE) of Zellner (1962) which allows us to simultaneously estimate the five equations linked to the five attributes of performance.

3. Data description

Our analysis draws on survey data recently collected by the UNIDO based on a common questionnaire applied to manufacturing firms in 19 SSA countries in 2009.⁸ Table 1 Panel A provides summary statistics for exporting and non-exporting firms. On average exporting firms are larger than non-exporting firms, employing 263 and 90 workers, respectively.⁹ SSA manufacturing firms that export are more capital intensive than non-export firms. Exporting enterprises are also larger in terms of average gross value added. For example, the group of exporting firms accounts for 78.6% of the output of SSA manufacturing, which means that exporters are bigger than non-exporters. Exporting enterprises equally pay higher wages than non-exporting firms. It also seems that exporters are on average more productive than non-exporters in terms of both value added per employee and TFP. For example, the value of gross value added produced by each employee for exporters is 4.1% higher than the industry mean, whereas that for non-exporters is 2% below. Likewise, the mean level of TFP is 4.8% above the industry mean for exporting firms and 6.8% below for non-exporting firms. At the median, the average level of TFP is 1.8% above the industry mean for the export group and 1.1% below for the non-export group.

Regarding the other characteristics, such as the R&D intensity, exporters have a higher R&D effort measured in terms of the ratio of R&D expenditures to value added. R&D effort is almost two times greater for exporters with respect to non-exporters in 2009. In particular, R&D effort is on average 2.8 points higher for exporting firms. Looking at firm age, once again we observe systematic differences between exporters and non-exporters and the average age difference is four years. With respect to foreign ownership, the degree of involvement of foreign-owned firms in the population of SSA manufacturing firms is high. The percentage of firms with a majority of foreign capital participation is greater in the group of exporting firms (nearly 56.8% of firms) whereas in the group of non-exporters, the rate of participation is approximately 29.4%. Pairwise t tests suggest that all of these indicators differ significantly between the two samples except in the both cases of labour productivity and R&D intensity. Summarizing, the raw data suggest a strong relationship between exporting and performance.

Table 1 Panel B shows that the proportion of exporting enterprises is nearly 31% and their corresponding export intensity is around 46%. Table 1 Panel C finally details the shares of output exported by SSA manufacturing firms to various destinations. The SSA destination has the lion’s share: approximately 36.9% of SSA manufactured exports are shipped within the SSA region. The historical economic ties and trade agreements (e.g. African Growth and Opportunities Act - AGOA with the US and Cotonou Agreements/Economic Partnership Agreement with the EU) have possibly given SSA enterprises relatively greater trade access to EU and US markets. For example, around 17.7% (with a corresponding export intensity of 40.8%) and 13.1% (with a corresponding export intensity of 10.7%) of SSA manufactured goods are exported to EU and US markets, respectively. In the remaining six export destinations of SSA enterprises, the proportion of exports shipped ranges from “other destinations” at 15.5% to India destination at 2.7%.

Firms’ characteristics by export-market destination are shown in Table 1 Panel D. The average firm size ranges from 215 employees in SSA exporters to 554 in US exporters. It is clear from the standard deviations that the size range is quite large. MENA, India and “other Asia” exporters have a higher level of capital intensity than exporters to the remaining export-market destinations. The activity of exporting internationally and within the SSA region is mainly dominated by older and foreign-controlled firms. The average firm age ranges from 27 years in “other destinations” exporters to 21.5 years in US exporters, whereas the average foreign ownership varies from 62.7% in India exporters to 51.8% in EU exporters. SSA manufacturing firms exporting to MENA and ‘other destination’ invest more in R&D (on average 11% and 9.8% of value added, respectively). In the remaining export-market destinations, the share of R&D expenditures in value added ranges from 7.9% in ‘other Asia exporters’ to 4.7% in US exporters.

4. Results

5. Conclusion

The objective of this paper, which uses a novel manufacturing firm-level dataset of 2009 in 19 sub-Saharan Africa (SSA) countries, was three-fold. The first was to determine the export premia and the impact of export behaviour on different performance measures, e.g., capital intensity, output, wage rate, labour productivity, and total factor productivity (TFP). The second was to assess the determinants of exporting. Finally, the study examined whether export-market destination mattered.

We find that SSA manufacturing firms that export differ from those that do not along a variety of dimensions: they are larger as measured by employment and output, more productive, more capital intensive, and pay higher wages. Moreover, foreign ownership and past export experience enhance the propensity to export, while continuing exporters outperform switching exporters. Regarding the role of export destination in the relationship between enterprise performance and exporting, three basic results emerge. Firstly, in terms of TFP, China appears to be the best destination. Secondly, exporting to India enhances the wage rate, labour productivity and TFP, with the labour productivity impact by far the largest among all the destinations. Thirdly, the South Africa destination impedes capital intensity. Finally, the scale of exporting matters for the performance of exporters. In particular, higher export intensity to the USA tends to be associated with improved performance, while the reverse appears to be the case with exports to other SSA countries.

From an industry policy standpoint, for a continent that is currently juggling multiple trade partners without a coherent and coordinated strategy, China and India export-market destinations could bolster the industrialization prospects of SSA countries. Meanwhile, the finding that export intensity is associated with particularly exports to the USA suggests that increasing volumes of exports to that country should be encouraged. Unfortunately, however, the reverse appears to be the case with exports to other SSA countries. The latter finding does not augur well for increased intra-Africa trade. Hence, SSA policymakers should recognise this opportunity and position themselves accordingly.