Social cohesion and firms' access to finance in Africa

Yabibal Mulualem Walle Working Paper Social cohesion and firms' access to finance in Africa IDOS Discussion Paper, No. 9/2022 Provided in Cooperation with: German Institute of Development and Sustainability (IDOS), Bonn Suggested Citation: Yabibal Mulualem Walle (2022) : Social cohesion and firms' access to finance in Africa, IDOS Discussion Paper, No. 9/2022, ISBN 978-3-96021-188-4, German Institute of Development and Sustainability (IDOS), Bonn, https://doi.org/10.23661/idp9.2022 This Version is available at: https://hdl.handle.net/10419/263265 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Social Cohesion and Firms’ Access to Finance in Africa Yabibal M. Walle IDOS DISCUSSION PAPER 9/2022 Social cohesion and firms’ access to finance in Africa Yabibal M. Walle Bonn 2022 Yabibal M. Walle is a researcher in the “Transformation of Economic and Social Systems” programme at the German Institute of Development and Sustainability (IDOS). E-mail: [email protected] Published with financial support from the Federal Ministry for Economic Cooperation and Development (BMZ) Suggested citation: Walle, Yabibal M. (2022). Social cohesion and firms’ access to finance in Africa (IDOS Discussion Paper 9/2022). Bonn: German Institute of Development and Sustainability (IDOS). https://doi.org/10.23661/idp9.2022 Disclaimer: The views expressed in this paper are those of the author(s) and do not necessarily reflect the views or policies of the German Institute of Development and Sustainability (IDOS). Except otherwise noted, this publication is licensed under Creative Commons Attribution (CC BY 4.0). You are free to copy, communicate and adapt this work, as long as you attribute the German Institute of Development and Sustainability (IDOS) gGmbH and the author(s). IDOS Discussion Paper / German Institute of Development and Sustainability (IDOS) gGmbH ISSN 2751-4439 (Print) ISSN 2751-4447 (Online) ISBN 978-3-96021-188-4 (Print) DOI: https://doi.org/10.23661/idp9.2022 © German Institute of Development and Sustainability (IDOS) gGmbH Tulpenfeld 6, 53113 Bonn E-Mail: [email protected] http://www.idos-research.de IDOS Discussion Paper 9/2022 III Abstract Social cohesion has recently gained increasing attention in academic and policy circles. Apart from being a necessary feature of stable societies per se, social cohesion is also a key factor for sustainable economic development. One potential means through which social cohesion could foster economic development is by enhancing financial development. In this paper, we examine whether social cohesion is significantly associated with firms’ access to finance in Africa. To this end, we use a recently constructed dataset on social cohesion in Africa, which is based on the Afrobarometer survey and the Varieties of Democracy database. The dataset contains indices for the three pillars of social cohesion – trust, inclusive identity and cooperation for the common good. Combining this dataset with that of the World Bank Enterprise Surveys, we build a sample which covers more than 12,500 firms and 27 African countries. Our results show that all three components of social cohesion are positively associated with at least one measure of firms’ access to external finance. In particular, trust – but not inclusive identity and cooperation for the common good – is significantly associated with the likelihood that firms have a checking or savings account, or are financially constrained. When we measure access to finance with respect to having a line of credit or a loan from a financial institution, all the three pillars of social cohesion, including inclusive identity and cooperation for the common good, are related to access to finance. The results are robust to addressing endogeneity concerns using a heteroskedasticity-based identification strategy. Overall, our results suggest that improving social cohesion (e.g. through social protection, education, strengthening civil society organisations) could do more than hold society together; it could also promote access to finance, growth of firms, and thus economic development and job creation. Keywords: Access to finance, social cohesion, trust, cooperation for the common good, identity, Africa IDOS Discussion Paper 9/2022 IV Acknowledgements I would like to thank Tilman Altenburg, Kathrin Berensmann, Clara Brandi, Francesco Burchi, Helmut Herwartz, Julia Leininger, Armin von Schiller and Christoph Sommer for their valuable comments and suggestions. Financial support from the German Federal Ministry for Economic Cooperation and Development (BMZ) is gratefully acknowledged. Bonn, August 2022 Yabibal M. Walle IDOS Discussion Paper 9/2022 V Contents Abstract III Acknowledgements IV Abbreviations VII 1 Introduction 1 2 Social cohesion and financial development: theoretical background 3 2.1 Defining social cohesion 3 2.2 Social cohesion and financial development 5 2.2.1 Trust and access to finance 5 2.2.2 Cooperation for the common good and access to finance 6 2.2.3 Identity and access to finance 7 3 Data and empirical strategy 8 3.1 Data 8 3.1.1 Social cohesion indicators and other country-level controls 9 3.1.2 Access to finance and other firm characteristics 12 3.2 Model specification 14 3.3 Estimation strategy 15 4 Results and discussion 15 4.1 Baseline results 15 4.2 Sensitivity analysis 20 5 Conclusions 20 References 22 Appendix: Results using the heteroskedasticity-based identification strategy 25 IDOS Discussion Paper 9/2022 VI Figures Figure 1: Constitutive elements of social cohesion 4 Tables Table 1: Afrobarometer and WBES surveys 9 Table 2: Summary statistics: country-level indicators 11 Table 3: Pearson’s correlation coefficient among country-level indicators 12 Table 4: Summary statistics: access to finance and other firm-level indicators 13 Table 5: Pearson’s correlation coefficient among access to finance indicators and firm-level characteristics 14 Table 6: Social cohesion and access to finance: likelihood of having a checking or savings account 17 Table 7: Social cohesion and access to finance: likelihood that a firm has a bank loan/ line of credit 18 Table 8: Social cohesion and access to finance: likelihood of experiencing financial constraint 19 Tables in Appendix Table A1: Social cohesion and access to finance: likelihood of having a checking or savings account (heteroskedasticity) 25 Table A2: Social cohesion and access to finance: likelihood that a firm has a bank loan/ line of credit (heteroskedasticity) 26 Table A3: Social cohesion and access to finance: likelihood of experiencing financial constraint (heteroskedasticity) 27 IDOS Discussion Paper 9/2022 VII Abbreviations CSO Civil society organisation DIE German Development Institute / Deutsches Institut für Entwicklungspolitik IDOS German Institute of Development and Sustainability / Deutsches Institut für Entwicklung und Nachhaltigkeit OECD Organisation for Economic Co-operation and Development UNDP United Nations Development Programme V-Dem Varieties of Democracy WBES World Bank Enterprise Surveys IDOS Discussion Paper 9/2022 7 et al. (2004) use electoral participation and blood donation as two important measures of social capital because both actions are driven only by social pressure and internal norms, and not by other legal or economic incentives. In the social cohesion literature, these two indicators would likely fall under the category of “cooperation for the common good”. Thus, we might interpret the findings of Guiso et al. (2004) as evidence not only of the effects of trust but also of cooperation for the common good on financial development. Indeed, Knack and Keefer (1997) argue that “norms of civic cooperation can be linked with economic outcomes in the same ways as trust”. These authors contend that cooperative norms act as constraints on narrow self-interest, leading individuals to contribute to the provision of public goods of various kinds. Internal sanctions (e.g. guilt) and external sanctions (e.g. shame and ostracism) associated with norms alter the costs and benefits of cooperation. It is noteworthy, however, that the social capital literature on cooperation tends to emphasise social pressure, expected (individual and social) benefits and internal norms as main motivations for cooperation. On the contrary, the “cooperation for the common good” component of social cohesion stresses the fact that people cooperate “for the common good” beyond their individual interests and “despite incentives for non-cooperation” (King, Samii, & Snilstveit, 2010, p. 337). Thus, while cooperative norms may facilitate or provide incentives for cooperation for the common good, they are not themselves a sign or measure of cooperation for the common good. In general, in a society with a higher degree of cooperation for the common good, public goods are provided more effectively, and free riders are more likely to be punished by community members through a complex set of sanctions (Ostrom, 1990). This could facilitate financial exchanges and thus promote the development of strong financial markets and institutions. Therefore, we expect that firms in countries with higher levels of cooperation for the common good are, on average, more likely to have access to finance. 2.2.3 Identity and access to finance Of the three social cohesion components, inclusive identity is by far the least studied in terms of its connection to a country’s level of financial development. However, there are several indications that inclusive identity could potentially affect financial development. The first one relates to the fact that the prevalence of discrimination is a key manifestation of a lower degree of inclusive identity. Several studies have shown that discrimination (e.g. against black workers in the United States (US), low caste people in India, or Jews in Nazi Germany) can lead to substantial economic losses, be it in terms of inefficient allocation of talent (Thorat & Newman, 2007; Hsieh, Hurst, Jones, & Klenow, 2019), or the loss of qualified business leaders (Huber, Lindenthal, & Waldinger, 2021). There is also ample evidence that discrimination in other areas of the economy extends to access to finance. For example, using the 1998 and 2003 US Small Business Finance Survey, Aseidu et al. (2012) find that, after controlling for a variety of factors that influence loan decisions, Black-owned firms have a 36.9 percentage points higher probability of loan denial than White male-owned firms. Similarly, Raj and Sasidharan (2018) document for the period 2006 to 2007 that Indian firms owned by socially disadvantaged groups (so-called Scheduled Castes and Scheduled Tribes) have a significantly lower probability of receiving formal credit, all other factors remaining equal. Hence, it is plausible to think that access to finance for both firms and households will be higher in countries where different identities are tolerated and minorities are not discriminated against. A second mechanism through which inclusive identity might influence access to finance is related to individuals’ sense of belonging to the nation state. In a socially cohesive society, people place a higher value on their national identities over their group identities, or at least the latter do not take strong precedence over the former (Langer et al., 2017). As a result, people IDOS Discussion Paper 9/2022 8 in such societies view themselves as participants in a shared national project (Langer et al., 2017), and this could have important implications for the stability of the state, the efficient provision of key public goods – including financial infrastructure – and the strength of institutions. These effects could, in turn, influence the development of financial institutions and, thus, firms’ access to external finance. 3 Data and empirical strategy 3.1 Data We combine two main data sources for this study. To measure firms’ access to external finance, we utilise firm-level data from the World Bank Enterprise Surveys (WBES) collected between 2009 and 2020. All of these surveys were conducted using the so-called “global methodology”, which is designed to enable cross-country comparisons.9 The firm-level data are combined with country-specific measures of trust, cooperation for the common good and identity levels, which are computed based on the fourth (2008), fifth (2012–2013) and sixth (2014–2015) rounds of the Afrobarometer survey and the V-Dem expert-based data of the corresponding years. It is noteworthy that, to minimise endogeneity concerns, and noting that it may take some time for social cohesion to have a significant impact on access to finance, we have deliberately taken the social cohesion data measured one to five years prior to the access to finance measures. This restriction ultimately leaves only 27 countries and 12,101 to 12,523 firms, depending on the estimated regression model. The Afrobarometer surveys and waves of WBES covered by the study are presented in Table 1. Finally, data on country-level controls are taken from Nunn (2008). 9 Specifically, our data are from the 26 October 2021 version of the WBES indicators database, downloaded from www.enterprisesurveys.org. Details on the survey methodology can be found on the website. IDOS Discussion Paper 9/2022 9 Table 1: Afrobarometer and WBES surveys Source: Author, based on Afrobarometer surveys, WBES and Nunn (2008). 3.1.1 Social cohesion indicators and other country-level controls To measure social cohesion and its components, we rely on the definitions and measurements suggested by Leininger et al. (2021). We chose this dataset primarily because, to our knowledge, it is the only dataset on social cohesion in Africa that contains estimates for each of the three pillars of social cohesion as well as for its sub-components, allowing us to examine the relationship between each of the components and sub-components and firms’ access to finance in Africa. In addition, this dataset is also unique in that it covers a large number of African Afrobarometer World Bank Enterprise Surveys Country Year Year Number of firms % Benin 2014 2016 142 1.13 Botswana 2008 2010 262 2.09 Burundi 2012 2014 156 1.25 Cameroon 2015 2016 301 2.40 Cape Verde 2008 2009 135 1.08 Côte d’Ivoire 2014 2016 322 2.57 Eswatini 2015 2016 117 0.93 Ghana 2012 2013 690 5.51 Guinea 2015 2016 136 1.09 Kenya 2014 2018 962 7.68 Lesotho 2014 2016 125 1.00 Liberia 2015 2017 146 1.17 Madagascar 2008 2013 362 2.89 Malawi 2012 2014 409 3.27 Mali 2014 2016 167 1.33 Morocco 2015 2019 785 6.27 Mozambique 2015 2018 589 4.70 Namibia 2012 2014 473 3.78 Niger 2015 2017 128 1.02 Nigeria 2012 2014 1,999 15.96 Senegal 2013 2014 537 4.29 South Africa 2015 2020 1,046 8.35 Tanzania 2012 2013 598 4.77 Togo 2014 2016 147 1.17 Uganda 2012 2013 658 5.25 Zambia 2014 2019 578 4.62 Zimbabwe 2014 2016 554 4.42 Total 12,524 100.00 IDOS Discussion Paper 9/2022 10 countries, which is crucial for conducting the meaningful empirical analysis envisaged in this study. Furthermore, some of the existing objective proxies for social cohesion are either its drivers or its consequences, e.g. the degree of ethnic fractionalisation or the share of the middle class (Easterly et al., 2006), and thus they may not always exhibit a robust correlation with the “true” level of social cohesion in a country (Van der Meer & Tolsma, 2014). In contrast, and as a third major advantage, the IDOS dataset provides a direct measure of social cohesion that – apart from cooperation for the common good, which is partly built on V-Dem expert data – is entirely based on representative surveys of individuals’ perceptions of various aspects of social cohesion. Works already using this dataset include those by Burchi et al. (in press), who analyse the relationship between social cohesion and human development, and Burchi and ZapataRomán (in press), who examine the relationship between inequality and social cohesion. While referring interested readers to Leininger et al. (2021) for more details, we provide here a brief description of how the three indices and the four sub-indices are constructed in the IDOS dataset. Trust: This index is constructed from two sub-indices: social trust and institutional trust. Social trust, in turn, is built on the positive responses of survey respondents to the popular survey question to measure social trust: “Generally speaking, would you say that most people can be trusted or that you must be very careful in dealing with people?” Institutional trust, on the other hand, is calculated as the arithmetic mean of trust in parliament, trust in the courts and trust in the police. The geometric mean of the two sub-indices is used to arrive at IDOS’s overall trust indicator. All data for the trust measure are taken from the Afrobarometer surveys. Cooperation for the common good: This indicator also consists of two sub-indices: horizontal cooperation and vertical cooperation. Horizontal cooperation, in turn, is measured using three indicators. The first indicator, from the Afrobarometer, concerns membership of voluntary, nonreligious associations or organisations. Because membership of some associations that focus on a particular ethnic group may not reflect cooperation for the common good of society as a whole, a number of adjustments and weightings were made to arrive at the final indicator of membership of organisations. The second indicator, taken from the V-Dem database, is an expert evaluation of the degree of participation of citizens in civil society organisations (CSOs). The third indicator of horizontal cooperation is again derived from Afrobarometer and measures whether respondents have joined others to raise an issue with the government in the past year. As with membership of organisations, this indicator is also adjusted to account for the fact that teaming up with people from other ethnic groups represents a higher level of cooperation for the common good than raising an issue together with members of one’s own ethnic group. Vertical cooperation measures the strength of state–society cooperation. Two groups of indicators are used to build the vertical cooperation sub-index. The first group includes perception data from the Afrobarometer regarding the frequency of attending meetings and contacting local government councillors, members of parliament, officials of a government agency and traditional leaders. The second group of indicators involves expert data from V-Dem regarding the level of state repression toward CSOs and the degree to which CSOs are consulted by policy-makers. Inclusive identity: Due to lack of appropriate data to measure inclusive identity in Africa, IDOS’s measure of inclusive identity relies on a single question from the Afrobarometer surveys that compares the respondents’ feelings towards their superordinate national identity vis-à-vis their ethnic identity. Accordingly, a country with more respondents either with strong feelings only for their national identity or stronger feelings for their national identity than for their ethnic identity receives a better ranking of inclusive identity. Finally, for reasons of comparability among the various components of social cohesion, and as we do not know their “true” scales, all the indices and sub-indices were rescaled to take values IDOS Discussion Paper 9/2022 11 between 0 and 100.10 Table 2 documents summary statistics for the seven IDOS measures of various aspects of social cohesion and the four country-level controls: absolute latitude, the percentage of adherence to Islam, a dummy for French legal origins, and the intensity of slave trade (e,g. Pierce & Snyder, 2018; Levine et al., 2020). Table 2: Summary statistics: country-level indicators Author, based on Leininger et al. (2021) (social cohesion indicators) and Nunn (2008) (country controls). Table 3 documents the Pearson’s correlation coefficient between each of the social cohesion indicators and four country-level controls. These correlations reveal several noteworthy relationships. First, the correlation between the two sub-indices for trust (social and institutional) is low at 0.36, highlighting the importance of using each sub-index in the regression analysis we will perform in the next section. Moreover, the index for overall trust seems to be dominated by social trust, as both are correlated with a coefficient of 0.93, while the correlation coefficient between institutional trust and the overall trust is relatively low at 0.571. Second, unlike the case of trust, the correlation between the sub-indices for the cooperation for the common good is relatively high at 0.592. Moreover, with correlation coefficients of 0.856 (with vertical cooperation) and 0.891 (with horizontal cooperation), the comprehensive indicator of cooperation for the common good is highly correlated with its sub-components, allowing it to adequately represent both its sub-components in the upcoming regression analyses. Third, while trust and identity have a significant and positive correlation (0.446), cooperation for the common good is not significantly correlated with either trust (0.131) or identity (-0.050). This underscores the risk of using any of the three components as the sole indicator of social cohesion in Africa. As for the correlations between social cohesion indicators and country controls, it is interesting to see that former French colonies have a higher level of social (but not institutional) trust and national identity. Moreover, the percentage of Islamic population has also a significant (albeit at the 10%) and positive correlation with the level of social trust. The remaining country controls are not significantly correlated with the country’s level of social cohesion. In particular, contrary to the assumption made in Pierce and Snyder (2018) and Levine et al. (2020), the intensity of pre-colonial slave trade is not significantly correlated with any of the components or sub-components of social cohesion. This may reflect the fact that social cohesion, while very 10 For re-scaling, we use the so-called min-max normalisation. In particular, the new re-scaled value is obtained as New value= ((value – min) / (max – min))*100, where min and max represent the observed minimum and maximum values within each index and sub-index. Indicator Obs Mean SD Min Max Trust 27 33.35 25.3 0 100 Trust: social 27 27.55 23.6 0 100 Trust: institutions 27 48.04 25.6 0 100 Cooperation 27 35.67 25.5 0 100 Cooperation: vertical 27 43.86 26.5 0 100 Cooperation: horizontal 27 39.08 26.3 0 100 Identity 27 53.19 25.4 0 100 Absolute latitude 27 14.60 9.26 .2 33 French 27 0.48 .509 0 1 % Islamic 27 24.86 33.5 0 99 Slave trade 27 3.64 3.87 -2.3 8.8 IDOS Discussion Paper 9/2022 12 persistent, is not immutable. The various social, economic, and political changes that countries have experienced over the past 500 years seem to be more important to social cohesion in Africa today than the slave trade that took place between 1400 and 1900. Table 3: Pearson’s correlation coefficient among country-level indicators Note: Correlations are computed based on 27 observations (countries). Author, based on Leininger et al. (2021) (social cohesion indicators) and Nunn (2008) (country controls). 3.1.2 Access to finance and other firm characteristics We measure the degree of firms’ access to external finance by means of three alternative indicators. First, we consider an indicator of whether or not a firm has a checking or savings account (fin15). Similarly, our second indicator is a dummy variable which takes on the value 100 when the firm has a line of credit or a loan from a financial institution (fin14) and zero otherwise. Accordingly, both fin14 and fin15 will be our “direct” measures of the actual level of firms’ access to formal finance. As a third and “indirect” measure of access to finance, we generate an indicator that shows if a firm was financially constrained in the fiscal year prior to the survey (finConstr). Based on the WBES dataset, we consider a firm to be financially constrained (finConstr =100) if its loan application was rejected in the last fiscal year (fin21=100) or if it reports needing a loan (fin20=0) but has not applied for it (fin21 is missing). Hence, constrained firms are those that need finance but were denied it (“formally constrained”) or those that did not apply for it because they were discouraged from applying for fear of rejection (“informally constrained”).11 Summary statistics for all access to finance indicators along with other firm-level characteristics are provided in Table 4. In particular, 84.91% of the firms have a checking or savings account. This relatively high figure for fin15 mirrors recent progress in expanding account ownership to firms and households in Africa (Demirgüç-Kunt & Klapper, 2012). However, this figure drops to just 20.21% when we consider the number of firms with a bank loan or a line of credit (fin14). This could be attributable to several factors, such as the high cost of credit, the high collateral 11 This operationalisation of financial constraint by considering both formally and informally constrained borrowers was first proposed by Jappeli (1990) and has been widely used in the literature (e,g. Guiso et al., 2006; Léon, 2015; Popov & Udell, 2012; Ferrando, Popov, & Udell, 2019). Trust Trust: social Trust: instit. Coop. Coop.: vertical Coop.: horiz. Identity Latitude French % Islamic Trust: Social 0.963*** Trust: inst. 0.571*** 0.360* Coop. 0.131 0.144 0.097 Coop.: vertical 0.008 0.012 0.111 0.856*** Coop.: horizontal 0.215 0.233 0.072 0.891*** 0.529*** Identity 0.446** 0.432** 0.346* -0.050 -0.065 -0.015 Latitude -0.042 -0.073 0.172 -0.213 -0.169 -0.199 0.144 French 0.401** 0.483** -0.078 0.027 -0.072 0.112 0.504*** -0.088 % Islamic 0.348* 0.330* 0.057 0.185 0.069 0.234 0.160 -0.036 0.474** Slave trade 0.157 0.182 -0.206 0.270 0.211 0.254 -0.037 -0.674*** 0.239 0.259 IDOS Discussion Paper 9/2022 13 requirements imposed by banks, the unavailability of enough savings on the part of the banks, or lack of viable investment projects by the firms. In line with the low fin14, the average number of firms that are financially constrained, i.e. firms whose loan applications were denied or firms who were discouraged from applying (finConstr), is high at 44.27%. Table 4: Summary statistics: access to finance and other firm-level indicators Indicator Obs Mean SD Min Max % firms with a checking or savings account (fin15) 12,523 84.91 35.8 0 100 % firms with a bank loan/line of credit (fin14) 12,101 20.21 40.2 0 100 % firms that are financially constrained (finConstr) 12,626 44.27 49.7 0 100 Log(age) 12,523 2.63 .781 0 5.0 Small firm 12,523 56.99 49.5 0 100 Exporter 12,523 12.81 33.4 0 100 Foreign ownership 12,523 15.70 36.4 0 100 Female in the top management 12,523 16.46 37.1 0 100 Log(manager’s experience) 12,523 2.54 .74 0 4.1 Source: Author, based on self-reported responses of managers of firms surveyed by WBES. With regard to the correlation between our access-to-finance indicators and firm-level characteristics, Table 5 shows that the two direct measures of access to finance (fin15 and fin14) are positively correlated with each other, as expected. Moreover, the index of financial constraint (finConstr) shows the expected negative and significant correlation with the two direct measures of access to finance (fin15 and fin14). Furthermore, most correlations between all three indicators of access to finance and firm-level characteristics have the expected signs. In particular, older, larger, and exporting firms, as well as firms led by experienced managers, have a higher unconditional probability of better access to finance (and a lower probability of facing financial constraint). However, the relationship between having a female manager and access to finance depends on the specific access to finance indicator being considered. In particular, while female-owned firms are more likely to have a checking or savings account than maleowned firms, the former are less likely to have a line of credit or a loan from a financial institution and are more likely to be financially constrained. IDOS Discussion Paper 9/2022 14 Table 5: Pearson’s correlation coefficient among access to finance indicators and firmlevel characteristics fin15 fin14 finConstr age small exporter foreign female fin14 0.143*** finConstr -0.079*** -0.141*** Age 0.069*** 0.048*** -0.076*** Small -0.148*** -0.174*** 0.153*** -0.225*** exporter 0.001 0.083*** -0.067*** 0.058*** -0.179*** Foreign 0.073*** 0.085*** -0.038*** 0.019** -0.199*** 0.259*** Female manager 0.017* -0.028*** -0.028*** -0.022** 0.084*** -0.040*** -0.080*** Manager’s experience 0.056*** 0.057*** -0.019** 0.440*** -0.122*** 0.020** 0.023** -0.090*** Note: Correlations are computed based on 12,523 observations (firms). Source: Author. 3.2 Model specification Our goal is to estimate the effect of a country’s level of social cohesion on firms’ access to finance in Africa. As our access to finance indicators are dichotomous, we estimate a series of probit models with the following empirical specification: FDisc =β0+β1Trustc+β2Cooperationc+β3Identityc+β4Xisc +β5Cc+β6ds+β7dPeriodt +εisc, (1) where FDisc stands for one of the three measures of access to finance for firm 𝑖𝑖, which belongs to sector s and is located in country 𝑐𝑐. The three main variables of interest are country-levels of trust (Trustc), cooperation for the common good (Cooperationc), and identity (Identityc). The firm-level characteristics we control for (Xisc) include firm size, a dummy variable for the gender of the manager, the number of years of experience of the manager in logarithms and the age of the firm in logarithms. Moreover, we include two dummy variables indicating whether the firm generates at least 10% of its annual sales from direct exports, and whether foreigners hold at least 10% of the firm’s shares. To account for industry-specific differences among firms, our estimation model in (1) also includes sector dummies (ds). Because the WBES surveys were conducted in different years spanning 2009 to 2020, we include time dummies for the periods 2009–2013, 2014–2015, 2016–2017 and 2018–2020 to account for global trends over time.12 The final term in the regression (εisc) is the white noise error term. Noting that firms that are located in the same country may have similar characteristics as they operate in similar business environments, we also cluster the standard errors at the country level. The vector Cc stacks four country controls: absolute latitude, the percentage of adherence to Islam, a dummy for French legal origins, and the intensity of slave trade (see, for example, Pierce and Snyder, 2018; Levine et al., 2020). Since the inclusion of country fixed effects is not possible because the model already contains country-level measures of social cohesion, the country controls are supposed to minimise the risk that our social cohesion indicators might pick up other country-level differences. However, the fact that we have only 27 countries and thus 12 The use of year dummies was not possible because some years are represented by single countries, leading to multicollinearity problems. IDOS Discussion Paper 9/2022 15 limited degrees of freedom makes it difficult to include other country controls, as they would cause the problem of multicollinearity among the explanatory variables. The four country controls used in this paper are those that are known to determine a country’s institutional, economic, and financial development and have been checked to ensure that they do not cause multicollinearity in our model. Moreover, it is noteworthy that social cohesion itself could be affected by other, deeper, factors, such as geography and culture, and hence controlling for such variables could attenuate the strength of the association between social cohesion and firms’ access to finance. 3.3 Estimation strategy Our baseline estimation method is to perform a probit regression on the model in (1). The possibility of reverse causality from firms’ access to finance to the degree of social cohesion in a country could be expected to be minimal on two grounds. First, social cohesion is measured at the country level while access to finance is measured at the firm level. Second, the surveys used to create the social cohesion indicators are conducted one to five years before the Afrobarometer surveys, which we use to measure firms’ access to finance. Still, our probit estimation could suffer from endogeneity biases arising from an omitted variable that affects both a country’s level of social cohesion and firms’ access to finance. To address this concern, one strategy we follow is to include in our regression model (1) as many country-specific controls as the multicollinearity problem allows, along with industry and time fixed effects. As a second strategy, we test the robustness of our probit estimates by means of an instrumental variable estimation. However, finding a suitable external instrument that affects firms’ access to finance only through its effect on social cohesion is difficult. As an alternative to this, we employ the heteroskedasticity-based identification strategy proposed in Lewbel (2012). This approach allows identification by means of internal instruments without imposing any exclusion restrictions. In what follows, we provide a brief intuitive explanation of this approach and refer interested readers to Lewbel (2012) and Baum and Schaffer (2012). We begin by grouping the variables into three: the dependent variable (Y1), the endogenous variables (Y2) and the exogenous variables (X). To construct the instruments, we first regress each endogenous variable in Y2 (social cohesion in our case) on the exogenous variables X and obtain the vector of residuals 𝑉𝑉 �. Then, we obtain the instruments as the product of the de-meaned exogenous variables and the residuals from the regression of the endogenous variable as (X − 𝑋𝑋 �)𝑉𝑉 �, where 𝑋𝑋 � is the mean of 𝑋𝑋. For the instruments to be valid, the residuals 𝑉𝑉 � have to be heteroskedastic. Recent applications of this identification strategy to address endogeneity problems can be found in Mallick (2012), Arcand, Berkes and Panizza (2015) and Tran, Walle and Herwatz (2020). 4 Results and discussion 4.1 Baseline results We proceed in several steps to examine the relationship between social cohesion and firms’ access to finance in Africa. Tables 6, 7 and 8 show the marginal effects from the estimated probit models. The three tables differ in terms of the particular access to finance measure used: dummy variables for having a checking or savings account fin15 (Table 6), having a bank loan or a line of credit fin14 (Table 7), and having been financially constrained in the fiscal year prior to the survey finConstr (Table 8). In all the three tables, the first specification includes all the three core social cohesion components simultaneously while each of the remaining seven IDOS Discussion Paper 9/2022 16 specifications contains only one of the main or sub-components of social cohesion. Robustness check results obtained by applying the heteroskedasticity-based identification strategy are provided in Tables A.1, A.2 and A.3 of the Appendix. Unless otherwise stated, our discussion of statistical significance refers to the 5% level. Table 6 presents probit estimates for the likelihood of a firm having a checking or savings account (fin15). In the specification with all the three main social cohesion indicators (column (1)), we see that trust enters positively and significantly at the 1% level of significance. On the contrary, coefficient estimates for cooperation and identity are not statistically significant. This leads us to tentatively conclude that trust is the most important, if not the only important, component of social cohesion that is positively associated with firms’ access to finance in Africa. Since there is some correlation between the three components (at least between trust and identity), one could argue that the components may share common information and therefore our regression analysis may face difficulty in disentangling separate effects, given the limited degrees of freedom (27 countries). In the next columns, we use each main and sub-component as a sole indicator of social cohesion. Columns (2) to (8) show that while all the social cohesion indicators are positively related to fin15, statistical significance of coefficients differs across the indicators. Consistent with the results in column (1), all trust indicators (overall, social and institutional) enter positively at the 1% level, while the coefficient estimates for inclusive identity are significant at the 10%. On the contrary, both the overall index of cooperation for the common good and its sub-indices do not show any significant association with firms’ access to finance in Africa. Hence, while these results clearly corroborate existing literature on the role of trust in financial development and access to finance (e.g. Guiso et al., 2004, An et al., 2021), they do not yet provide strong evidence of whether the broader social cohesion prevailing in a country (including cooperation and inclusive identity) is related to firms’ access to finance in Africa. For trust, the estimated coefficient is also large in economic terms. For example, given that our social cohesion indicators are scaled to take values between 0 and 100, the estimated coefficient in row (1) and column (2) of Table 6 imply that firms in the country with the highest trust level are 21.9 percentage points more likely to have a checking or savings account than firms in the country with the lowest trust level, other things constant. Turning to other firm controls, we obtain results that align well with those in the existing literature. In particular, small firms (i.e., firms with less than 20 workers) are more likely to be financially constrained and, in this particular case, less likely to have a checking or savings account – a finding that has been documented in several studies (Beck & Demirgüç-Kunt, 2006; Beck, Demirgüç-Kunt, & Maksimovic, 2005). The other feature of firms that is significantly linked to access to finance is their ownership. As expected, firms that are at least partly foreign-owned have a significantly higher probability of owning a checking or savings account (Beck & Demirgüç-Kunt, 2006; Beck et al., 2005). Moreover, firms managed by experienced managers are, as expected, more likely to have a checking or savings account. Coefficient estimates for other firm-level characteristics are not statistically significant, and hence we refrain from discussing them. IDOS Discussion Paper 9/2022 23 Jappelli, T. (1990) “Who Is Credit Constrained in the U.S. Economy?” Quarterly Journal of Economics, 105, 219–34. King, E., Samii, C., & Snilstveit, B. (2010). Interventions to promote social cohesion in sub-Saharan Africa. Journal of Development Effectiveness, 2(3), 336–370. Knack, S., & Keefer, P. (1997). Does social capital have an economic payoff? A cross-country investigation. Quarterly Journal of Economics, 112(4), 1251–1288. Konte, M., & Ndubuisi, G. (2021). Financial constraint, trust, and export performances: Firm-level evidence from Africa. Journal of Institutional Economics, 17(4), 583–605. Langer, A., Stewart, F., Smedts, K., & Demarest, L. (2017). Conceptualising and measuring social cohesion in Africa: Towards a perceptions-based index. Social Indicators Research, 131(1), 321–343. La Porta, R., Lopez‐de‐Silanes, F., Shleifer, A., & Vishny, R. W. (1997). Legal determinants of external finance. The Journal of Finance, 52(3), 1131–1150. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., & Vishny, R. W. (1998). Law and finance. Journal of Political Economy, 106(6), 1113–1155. Leininger, J., Burchi, F., Fiedler, C., Mross, K., Nowack, D., von Schiller, A., … Ziaja, S. (2021). Social cohesion: A new definition and a proposal for its measurement in Africa (Discussion Paper 31/2021). Bonn: German Development Institute / Deutsches Institut für Entwicklungspolitik (DIE). Léon, F. (2015). Does bank competition alleviate credit constraints in developing countries? Journal of Banking & Finance, 57, 130–142. Levine, R. (2005). Finance and growth: theory and evidence. Handbook of Economic Growth, 1, 865–934. Levine, R., Loayza, N., & Beck, T. (2000). Financial intermediation and growth: Causality and causes. Journal of Monetary Economics, 46(1), 31–77. Levine, R., Lin, C., & Xie, W. (2020). The African slave trade and modern household finance. Economic Journal, 130(630), 1817–1841. Lewbel, A., 2012. Using heteroscedasticity to identify and estimate mismeasured and endogenous regressor models. Journal of Business & Economic Statistics, 30(1), 67–80. Mallick, D., (2012). Microfinance and moneylender interest rate: Evidence from Bangladesh. World Development, 40(6), 1181–1189. Nunn, N. (2008). The long-term effects of Africa’s slave trades. Quarterly Journal of Economics, 123(1), 139–176. OECD (Organisation for Economic Co-operation and Development). (2011). Perspectives on global development 2012: Social cohesion in a shifting world. OECD Publishing. Ostrom, E. (1990). Governing the commons: The evolution of institutions for collective action. Cambridge: Cambridge University Press. Pierce, L., & Snyder, J. A. (2018). The historical slave trade and firm access to finance in Africa. The Review of Financial Studies, 31(1), 142–174. Popov, A., & Udell, G.F. (2012). Cross-border banking, credit access, and the financial crisis. Journal of International Economics, 87(1), 147–161. Raj, R. S. N., & Sasidharan, S. (2018). Does the caste of the firm owner play a role in access to finance for small enterprises? Evidence from India. Developing Economies, 56(4), 267–296. Rajan, R. G., & Zingales, L. (2003). The great reversals: The politics of financial development in the twentieth century. Journal of Financial Economics, 69(1), 5–50. Robinson, J., 1952. The generalization of the general theory. In: The rate of interest and other essays. London: Macmillan. Schiefer, D., & Van der Noll, J. (2017). The essentials of social cohesion: A literature review. Social Indicators Research, 132(2), 579–603. Stulz, R. M., & Williamson, R. (2003). Culture, openness, and finance. Journal of Financial Economics, 70(3), 313–349. Thorat, S., & Newman, K. S. (2007). Caste and economic discrimination: Causes, consequences and remedies. Economic and Political Weekly, 42(41), 4121–4124. IDOS Discussion Paper 9/2022 24 Tran, V. T., Walle, Y. M., & Herwartz, H. (2020). The impact of local financial development on firm growth in Vietnam: Does the level of corruption matter? European Journal of Political Economy, 62, 101858. UNDP (United Nations Development Programme). (2020). Strengthening social cohesion: Conceptual framing and programming implications. New York: UNDP Van der Meer, T., & Tolsma, J. (2014). Ethnic diversity and its effects on social cohesion. Annual Review of Sociology, 40(1), 459–478. V-Dem dataset. Gothenburg: V-Dem Institute, University of Gothenburg. Retrieved from https://www.vdem.net/vdemds.html World Bank Enterprise Surveys (WEBS). Washington, DC: World Bank. Retrieved from https://www.enterprisesurveys.org/en/enterprisesurveys (accessed 5 November 2021) World Bank (2012). Global -2013: Rethinking the role of the state in finance. Washington, DC: World Bank. IDOS Discussion Paper 9/2022 25 Appendix: Results using the heteroskedasticity-based identification strategy Table A1: Social cohesion and access to finance: likelihood of having a checking or savings account (heteroskedasticity) Notes: Standard errors are clustered (by country) and robust to heteroskedasticity. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable takes on the value 100 if the firm has a checking or savings account and 0 otherwise. The models are estimated using the heteroscedasticity-based identification strategy of Lewbel (2012). For further notes, see Table 3. Source: Author. (1) (2) (3) (4) (5) (6) (7) (8) Trust: social Trust: institutions Coop.: vertical Coop.: horizontal Trust 0.252*** 0.281*** (0.069) (0.072) Cooperation -0.010 -0.150 (0.052) (0.107) Identity -0.012 0.100 (0.061) (0.078) Sub-index 0.256*** 0.138*** -0.038 -0.086 (0.072) (0.050) (0.092) (0.146) Log(age) 1.164 1.192 0.937 1.099 1.012 1.370 1.011 1.019 (1.013) (1.021) (1.003) (1.015) (1.021) (1.036) (1.015) (0.984) Small firm -0.087*** - 0.087*** - 0.089*** - 0.086*** -0.087*** -0.086*** -0.087*** -0.088*** (0.023) (0.023) (0.022) (0.023) (0.023) (0.023) (0.023) (0.022) Exporter -0.000 0.002 -0.015 -0.004 -0.005 0.004 -0.009 -0.011 (0.015) (0.016) (0.015) (0.015) (0.016) (0.016) (0.017) (0.017) Foreign ownership 0.039*** 0.039*** 0.039*** 0.036*** 0.040*** 0.031*** 0.035*** 0.037*** (0.011) (0.011) (0.011) (0.010) (0.011) (0.010) (0.010) (0.011) Female top manager -0.001 -0.001 0.006 0.005 0.000 0.003 0.005 0.005 (0.010) (0.009) (0.011) (0.010) (0.010) (0.010) (0.010) (0.011) Log(manager’s experience) 1.901*** 1.857*** 2.380*** 2.145*** 2.197*** 1.598** 2.176*** 2.293*** (0.673) (0.683) (0.776) (0.726) (0.726) (0.739) (0.727) (0.868) Observations 12,523 12,523 12,523 12,523 12,523 12,523 12,523 12,523 Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Period FE Yes Yes Yes Yes Yes Yes Yes Yes Adjusted Rsquared 0.102 0.102 0.0846 0.0980 0.0977 0.103 0.0929 0.0916 Clusters 27 27 27 27 27 27 27 27 IDOS Discussion Paper 9/2022 26 Table A2: Social cohesion and access to finance: likelihood that a firm has a bank loan/line of credit (heteroskedasticity) Notes: Standard errors are clustered (by country) and robust to heteroskedasticity. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable takes on the value 100 if the firm has a bank loan/line of credit and 0 otherwise. The models are estimated using the heteroscedasticity-based identification strategy of Lewbel (2012). For further notes, see Table 3. Source: Author. (1) (2) (3) (4) (5) (6) (7) (8) Trust: social Trust: institutions Coop.: vertical Coop.: horizontal Trust -0.041 0.226** (0.099) (0.095) Cooperation 0.174*** 0.130 (0.049) (0.102) Identity 0.122 0.300** (0.088) (0.117) Sub-index 0.271*** 0.038 0.300** 0.209** (0.090) (0.082) (0.145) (0.089) Log(age) 1.043 0.942 0.990 0.945 0.819 0.953 1.173 0.992 (0.897) (0.937) (0.883) (0.863) (0.932) (0.938) (0.924) (0.934) Small firm -0.110*** -0.113*** -0.111*** -0.112*** -0.113*** -0.113*** -0.109*** -0.110*** (0.017) (0.017) (0.018) (0.017) (0.017) (0.017) (0.016) (0.018) Exporter 0.074*** 0.069*** 0.070*** 0.072*** 0.065*** 0.066*** 0.077*** 0.073*** (0.019) (0.019) (0.019) (0.019) (0.020) (0.020) (0.019) (0.019) Foreign ownership 0.030 0.037 0.029 0.039* 0.039* 0.032 0.025 0.028 (0.022) (0.023) (0.021) (0.021) (0.023) (0.022) (0.023) (0.022) Female top manager -0.009 -0.012 -0.010 -0.005 -0.012 -0.008 -0.013 -0.010 (0.012) (0.013) (0.012) (0.013) (0.013) (0.012) (0.012) (0.012) Log(manager’s experience) 1.306 1.337 1.323 1.589 1.621 1.394 1.228 1.144 (0.919) (0.981) (0.898) (0.972) (1.005) (1.013) (0.885) (0.914) Observations 12,101 12,101 12,101 12,101 12,101 12,101 12,101 12,101 Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Period FE Yes Yes Yes Yes Yes Yes Yes Yes Adjusted Rsquared 0.0952 0.0813 0.0901 0.0870 0.0774 0.0848 0.0888 0.0903 Clusters 27 27 27 27 27 27 27 27 IDOS Discussion Paper 9/2022 27 Table A3: Social cohesion and access to finance: likelihood of experiencing financial constraint (heteroskedasticity) Notes: Standard errors are clustered (by country) and robust to heteroskedasticity. *** p<0.01, ** p<0.05, * p<0.1. The dependent variable, finConstr, takes on the value 100 if the firm faced financial constraint in the year prior to the survey and 0 otherwise. The models are estimated using the heteroscedasticity-based identification strategy of Lewbel (2012). For further notes, see Table 3. Source: Author. (1) (2) (3) (4) (5) (6) (7) (8) Trust: social Trust: institutions Coop.: vertical Coop.: horizontal Trust -0.302** -0.299** (0.144) (0.128) Cooperation -0.081 0.199 (0.119) (0.211) Identity 0.258 -0.034 (0.163) (0.089) Sub-index -0.249* -0.126 -0.042 0.246 (0.130) (0.142) (0.144) (0.240) Log(age) -2.186* -2.251* -1.954 -2.117 -2.063 -2.394 -2.140 -2.015 (1.311) (1.304) (1.307) (1.331) (1.309) (1.480) (1.345) (1.314) Small firm 0.115*** 0.116*** 0.119*** 0.115*** 0.115*** 0.115*** 0.115*** 0.119*** (0.011) (0.012) (0.012) (0.012) (0.012) (0.013) (0.012) (0.012) Exporter -0.068*** -0.071*** -0.052*** -0.063*** -0.064*** -0.072*** -0.064*** -0.050*** (0.014) (0.015) (0.014) (0.012) (0.013) (0.018) (0.014) (0.014) Foreign ownership -0.010 -0.017 -0.017 -0.013 -0.017 -0.009 -0.011 -0.018 (0.018) (0.018) (0.018) (0.021) (0.018) (0.023) (0.020) (0.020) Female top manager -0.027* -0.030** -0.038** -0.035** -0.032** -0.034** -0.034** -0.037** (0.015) (0.015) (0.015) (0.014) (0.014) (0.015) (0.013) (0.015) Log(manager’s experience) 1.129 0.963 0.340 0.657 0.605 1.155 0.704 0.213 (1.109) (1.176) (1.278) (1.184) (1.198) (1.248) (1.155) (1.288) Observations 12,626 12,626 12,626 12,626 12,626 12,626 12,626 12,626 Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Period FE Yes Yes Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.0543 0.0509 0.0380 0.0475 0.0534 0.0446 0.0495 0.0387 Clusters 27 27 27 27 27 27 27 27