At What Levels of Financial Development Does Information Sharing Matter?

The purpose of this study is to investigate how increasing information sharing bureaus affect financial access. For this reason, we have employed contemporary and non-contemporary interactive Quantile Regressions in 53 African countries for the period 2004-2011. Information sharing bureaus are proxied with public credit registries and private credit offices. Financial development dynamics involving depth (at overall economic and financial system levels), efficiency (at banking and financial system levels), activity (from banking and financial system perspectives) and size are used. Two key findings are established. First, the effect of increasing private credit bureaus is not clearly noticeable on financial access, probably because private credit agencies are still to be established in many countries. Second, increasing public credit registries improves financial allocation efficiency and activity (or credit) between the 25th and 75th quartiles for the most part. As a main policy implication, countries in the top (or highest levels of financial development) and bottom (or lowest levels of financial development) ends of the financial efficiency and activity distributions are unlikely to benefit from enhanced financial allocation efficiency owing to increasing public credit registries.


Introduction
The purpose of this study is to assess how increasing information sharing offices affects financial access when existing levels of financial development are taken into account. Consistent with recent literature, less than 20 percent of African households have access to financial services (see IFAD, 2011;. According to the narrative, a great part of the population on the continent depends on the informal sector for financial services. Some factors limiting access to finance include: low population densities in many areas, poor facilities in transport and limited communication infrastructure. In regions where financial services from the formal sector are available, low-income households and small businesses are for the most part unable to meet some basic lending requirements like strict documentation and collateral provision. Even in situations where such conditions are met, cost barriers (like substantial transactions fees) and high minimum deposit of savings could still overwhelmingly restrict financial access.
The above concerns have led to issues of surplus liquidity or excess cash in African formal financial institutions (see Saxegaard, 2006;Fouda, 2009;Asongu, 2014a, p.70). The authors have suggested measures to curb both the voluntary and involuntary holding of excess liquidity by banks. First, voluntary holding of excess cash can be reduced by: (i) helping banks to track their positions at the central bank to prevent them from keeping reserves above statutory limits; (ii) consolidating institutions that are favourable to interbank lending and (iii) improving infrastructure to prevent remote bank branches from holding excess reserves essentially due to transportation problems. Second, involuntary keeping of surplus cash can be kept at minimum by: (i) reducing the incapacity of banks to lend in scenarios where interest rates are regulated, (ii) creating conducive conditions for commercial banks to invest surplus liquidity in bond markets; (iii) increasing investment avenues for regional banks via promotion of regional stock exchange markets and (iv) reducing lending contraction of banks through instruments that encourage competition and mitigate information asymmetry. This line of inquiry is closest to the last point of the first strand.
Over the past decade, information sharing bureaus have been introduced across the African continent in order to enhance financial access by limiting information asymmetry.
Unfortunately, recent empirical literature has been based on the assumption that information sharing agencies may not be increasing financial access as theoretically anticipated (see Triki & Gajigo, 2014). For instance, Asongu et al. (2016) have concluded that the effects of information sharing bureaus have been negative for the most part on financial development dynamics of depth, allocation efficiency and activity. Moreover, as we shall demonstrate in the literature review which follows, there has been very limited scholarly focus on the role of information sharing agencies on financial access in the African continent.
This study addresses highlighted gaps by investigating whether increasing information sharing bureaus could enhance financial access when it matters. The interest in considering initial levels of financial development arises because the findings of Asongu et al (2016) Henderson et al., 2013). In the light of the above insights, the Quantile regression empirical strategy is adopted because existing studies on information sharing have examined the relationship between information sharing bureaus and financial development by involving parameter estimates at the conditional mean of financial development variables (Triki & Gajigo, 2014;Asongu et al., 2016).
The emphasized research gaps are addressed by answering the following question: how does increasing information sharing bureaus affect financial access when its existing level matters in Africa? It is important to address this research inquiry because the findings should inform policy makers on how financial access barriers can be lifted to enable households and small corporations to maximise their savings and earnings for more productivity, more employment and higher economic growth. Hence, the contribution of this study is to complement existing literature by investigating how increases in information sharing offices influence financial development when existing levels are considered in the modelling exercise. One of the main results stemming from the econometric analysis is that increasing public credit registries improves financial efficiency in the middle of the financial development distribution. This result is intuitive because for poorly developed financial systems, increasing information sharing bureaus may in some respect decrease the pace of development, whereas for more developed financial systems, the impact of information sharing may already have been taken into account.

5
The rest of the study is structured as follows. Section 2 discusses the background, theoretical underpinnings and empirical literature. The data and methodology are covered in Section 3. The empirical results are presented and discussed in Section 4. Section 5 concludes with future research directions.

Background
Information sharing bureaus or credit reference agencies (private credit bureaus and/or public credit registries) are institutions that collect information on the obligations of commercial and individual borrowers from various sources, namely: direct investigation and public sources (for businesses), banks and credit card companies (for individuals) and retail lenders . Once the data is collected, the information is consolidated after cross-checking for a comprehensive credit report. Such a report is useful for potential creditors.
Data from a credit history report can encompass both positive and negative information: (i) positive information (entailing details on all opened and closed credits and repayment behaviour) and (ii) negative information (which is default data for the most part).
Information sharing bureaus are essential to enhance financial access in any economy because they enable the mitigation of information asymmetry that restricts lenders from offering credits. On the one hand, adverse selection issues are attenuated with information from credit histories. On the other, moral hazard issues are also controlled by limiting default rates and increasing repayment rates. Ultimately, the incremental volume in lending is essential for sectors with limited financial access like micro, small and medium corporations.
Prior to 2008, information sharing bureaus were confined to a few countries in the Organisation for Economic Cooperation and Development and Latin America (see Mylenko, 2008). However, the growth of information and communication technology has considerably increased the presence of such information credit offices in Sub-Saharan Africa, North Africa and the Middle and Eastern Europe. In Sub-Saharan, with the exception of South Africa, very few countries possessed well functioning information sharing bureaus before 2008. Some nations like Mozambique, Nigeria and Rwanda have instituted credit offices with the prime objective of boosting banking sector supervision. Greater articulation is made on higher lending rates and due to lack of appropriate technology and incentives, such credit offices did not provide timely and 6 accurate information most of the time. Before 2008, numerous initiatives were implemented throughout Africa in order to institute private credit bureaus given demands for data by supervisors to consolidate risk management practices on the one hand and on the other from financial institutions. In response, many countries introduced information sharing bureaus, namely: Ghana, Nigeria, Tanzania, Uganda and Zambia.

Theoretical highlights
As documented by Claus and Grimes (2003), two principal strands exist in the literature on theoretical underpinnings for a linkage between financial intermediaries and information sharing. The first strand shows liquidity provision by financial institutions while the second focuses on the capacity of financial intermediaries to modify the risk characteristics of assets.
Both strands are founded on the essential economic role of financial intermediation which is to augment efficiency in allocation by reducing the cost of conveying mobilised deposits from depositors to borrowers. The theories underlying the mission of financial intermediation are based on the literature of imperfect information in the market. Accordingly, the primary task of financial intermediation is to reduce transaction and information costs arising from information asymmetry between lenders and borrowers. Therefore, the establishment of information sharing bureaus is a channel through which the reduction of information can be consolidated. The underpinnings are consistent with pioneering literature on the relevance of information sharing in financial intermediary efficiency, notably, on: models of credit rationing (see Williamson, 1986;Stiglitz & Weiss, 1981;Jaffee & Russell, 1976), ex-post and ex-ante information asymmetry (Diamond & Dybvig, 1983), communication on potential borrowers to investors by banks (Leland & Pyle, 1977) and diversification with financial intermediaries (Diamond, 1984).
The association between information sharing bureaus and financial access can be seen from the view of adverse selection (from lenders) on the one hand and the perspective of moral hazard (from borrowers) on the other. Information sharing agencies provide lenders of financial institutions with borrower information and credit histories which enable the reduction of substantial interest rates that were previously the consequence of adverse selection on the part of financial institutions. When borrowers are granted credit, they automatically become liable to moral hazard because their economic activities related to granted credit could be concealed in order to reduce compliance with their financial obligations towards the bank or lender. It is 7 therefore the responsibility of credit bureaus to discipline borrowers on the severe consequences of not complying with their periodic financial obligations. It is within this framework that information sharing bureaus reduce moral hazard in borrowers, essentially by educating them on the perils of debt defaults and resorting to the informal financial sector as a sustainable alternative to the formal banking sector.

Empirical literature
A considerable bulk of empirical studies on information sharing has been oriented towards the impact of creditors' rights to more data on the one hand and the impact of reducing information asymmetry among creditors on the other. The former orientation has for the most part focused on the influence that stronger creditors' rights have on, inter alia: capital structure (El Ghoul et al., 2012); bankruptcy (Cleassens & Klapper, 2005;Djankov et al., 2011) and more risk-taking by financial institutions (Houston et al., 2010;Acharya et al., 2011). This last orientation has revolved around investigating how sharing information consolidates credit availability (Djankov et al., 2007;Brown et al., 2009;Triki & Gajigo, 2014), mitigates rates of default (Jappelli & Pagano, 2002), reduces credit cost (Brown et al., 2009 ), affects syndicated bank loans (Ivashina, 2009;Tanjung et al., 2010), impacts antitrust intervention (Coccorese, 2012) and affects corrupt-lending (Barth et al., 2009). It is apparent from the above literature that inquiries have for the most part been dedicated to developed countries which have comparatively less severe barriers to financial access. Accordingly, while a substantial body of studies has been oriented toward the Organisation of Economic Cooperation countries on the one hand and on the other, emerging nations in Asia and Latin America, very little scholarly work has been devoted to Africa: a continent with substantially higher constraints to financial access .
Macroeconomic evidence on the influence of reducing information asymmetry has been investigated by Galindo and Miller (2001) who concluded that developed countries with credit registries are associated with lower levels of financial restrictions in comparison to their less developed counterparts with credit bureaus. Specifically, public credit registries that are performing well contribute considerably to reducing the sensitivity of decisions in investment for 'cash flow availability'; a characteristic proxy for financial constraint. 8 A combination of private credit bureaus and public credit offices were employed by Love and Mylenko (2000) with firm-based data from the World Bank Business Environment Survey.
They investigated whether financial access constraints are negatively related to credit registries.
The findings show that private credit bureaus are linked to higher financial access whereas public credit registries have no significant impact on decreasing constraints in financial access. Barth et al (2009) have investigated the effect of (i) information sharing and (ii) borrower and lender competition on 'lending corruption' through information sharing bureaus using the World Bank Business Environment Survey from fifty-six nations. The data set consisted of 4000 corporations and private credit in one hundred and twenty-nine nations. Two main findings are established. First, corrupt-lending is reduced by competition in banking and reducing information asymmetry. Second, competition among firms and the legal environment have had considerable effect on corrupt-lending. Triki and Gajigo (2014) have investigated two principal issues, namely (i) the impact of information sharing bureaus on corporations' access to finance and (ii) the effect of public credit registries' design on the rate of constraint on financial access. The following key findings are apparent. First, access to finance is comparatively greater in countries with higher private credit bureaus relative to countries with public credit registries or no information sharing office.
Second, there is considerable heterogeneity in financial access and on the design of information sharing bureaus with public credit agencies.
Information sharing thresholds have been investigated by Asongu et al (2016). It was established that information sharing bureaus have negative effects on financial depth, with the impact from public credit registries comparatively more noticeable. Private credit bureaus have a higher negative impact on banking system efficiency whereas public credit registries have an insignificant effect. Information sharing bureaus have negative effects on financial activity, with the impact from public credit registries being comparatively higher. The positive influence of private credit bureaus on financial size is comparatively low.

Data
As has been said, this study examines a panel of 53 African countries with data for the Database of the World Bank. The periodicity is constrained by data availability. Consistent with Asongu et al. (2013), four financial development variables are used, namely: depth, efficiency, activity and size. First, financial depth embodies (i) overall-economic depth (M2/GDP) 1 representing the monetary base plus demand, savings and time deposits and (ii) financial system deposits (Fdgdp). Distinguishing these two measurements is important because a substantial bulk of the monetary base in developing nations does not circulate within the formal banking sector.
Second, financial allocation efficiency measures the ability of financial intermediaries to transform mobilised deposits into credit for economic agents. Two measurements of efficiency are used, namely (i) banking-system-efficiency (with bank credit on bank deposits: Bcbd') and (ii) financial-system-efficiency ('financial system credit on financial system deposits: Fcfd').
Third, financial activity is measured as the ability of financial institutions to provide credit to economic agents. Two indicators are also used for this dimension of financial development, namely (i) banking system activity (with 'private domestic credit by deposit banks: Pcrb') and (ii) financial system activity (with 'private credit by domestic banks and other financial institutions: Pcrbof"). Fourth, financial size is the as the ratio of 'deposit bank assets' to 'total assets' ('deposit bank assets on central bank assets plus deposit bank assets': Dbacba). It is important to note that financial ratios which are dependent variables are mostly dimensions identified by the Financial Development and Structure Database of the World Bank.
Consistent with recent information asymmetry literature, information sharing bureaus are measured with public credit bureaus and private credit registries (Triki & Gajigo, 2014;Asongu et al., 2016). Asongu et al. (2016) have documented six distinguishing characteristics between public credit bureaus and private credit registries, notably: purpose, coverage, status, ownership, data sources used and access. First, whereas private credit registries are made-up of public institutions that are constituted within the framework of supervising the banking sector, public credit bureaus are created in response to the need of and demand for information on borrowers in the banking market. Hence, data from private credit registries, usually employed to examine the credit-worthiness of clients, could also be acknowledged as a collateral benefit or by-product of private credit registries. Second, while the coverage engendered by private credit registries is restricted in terms of information (or data) type and history provided, public credit bureaus extend beyond the scope of large corporations and include small and medium size enterprises 10 (SMEs) that are characterised with richer data and longer histories. Third, whereas public credit bureaus are fundamentally focused on profit-making, private credit registries are not primarily established for profit-making. Fourth, on the issue of ownership, whereas private credit registries belong to governments and/or central banks, the ownership of public credit bureaus revolve outside highlighted establishments (central banks and governments) to include lenders, lenders' associates and independent third parties. Fifth, while the data used by private credit registries is sourced from non-bank and bank financial establishments, data from public credit bureaus entails: private credit registries, tax authorities, courts and utilities to sources employed by private credit registries for information. Sixth, access to private credit registries (public credit bureaus) is restricted to providers of information (open to all lender types).
The control variables are also consistent with the recent information asymmetry literature (Asongu et al., 2016), namely: inflation, public investment, GDP growth, trade and foreign aid.
First, foreign aid like remittances (Aggarwal et al., 2011;Efobi et al., 2014) could increase financial development if it is not associated with activities that decrease their flow within a country such as funds captured by developed countries for consultancy services and deposited by corrupt officials from developing countries in tax havens that are under the jurisdictions of developed countries.
Second, there is an abundant supply of literature which has established a positive growthfinance relationship (see Saint-Paul, 1992;Greenwood & Jovanovic, 1990;Owosu & Odhiambo, 2014;Nyasha & Odhiambo, 2015ab). According to them, economic growth is linked to decreasing cost in financial intermediation which is the outcome of higher compensation that entails growing financial resources that are devoted for the purpose of investment. Moreover, the importance of income levels in financial development has been established in both broad (Levine, 1997) and African-specific (Asongu, 2012) studies. Whereas Asongu has shown that countries with high income are linked to greater financial development levels in Africa, it has been concluded by Jaffee and Levonian (2001) that higher income countries are associated with more developed banking system structures. It is important to balance the engaged narrative with the fact that growth may be linked to financial crises that ultimately reduce financial development (Asongu, 2016).
Third, there is a branch of the literature sympathetic to the perspective that policies which are friendly to openness (like trade) are positively linked to financial development (see Huang & Temple, 2005;Do & Levchenko, 2004).
Fourth, the nexus between investment and financial development has been assessed by Huang (2011) who found a positive connection.
Fifth, some main domestic macroeconomic policies, such as the keeping of inflation levels stable/low are needed for financial development (Huang, 2011;Boyd et al., 2001;Huybens & Smith, 1999). Accordingly, Huybens and Smith (1999) and Boyd et al. (2001) have respectively shown theoretically and empirically that nations with chaotic/high inflations are very likely to be associated with smaller, less efficient and less active financial institutions.
It is important to note that the expected signs of covariates cannot be established without uncertainty because the corresponding financial development variables are conflicting by definition and measurement. For example, financial efficiency is measured as the ratio of financial activity (credit) on financial depth (deposits). The definitions and sources of the variables are disclosed in Appendix 1, the summary statistics in Appendix 2 and the correlation matrix in Appendix 3.

Methodology
Consistent with the research question which is to assess the influence of information sharing offices on financial access when existing financial development levels matter, the study accounts for existing levels of financial development by employing a Quantile Regressions estimation technique which has been documented to account for initial levels in dependent variables (see Billger & Goel, 2009;Okada & Samreth, 2012;Asongu & Nwachukwu, 2017) when information sharing bureaus are examined throughout the conditional distributions of the outcome variable (Keonker & Hallock, 2001).
Previous literature on information sharing has examined the relationship between information sharing bureaus and financial development by engaging parameter estimates at the conditional mean of financial development variables (Triki & Gajigo, 2014;Asongu et al., 2016). Whereas mean impacts are relevant, the underlying literature is extended by employing Quantile Regressions that distinguish existing levels of financial access. Furthermore, while Ordinary Least Squares-related regressions are based on the hypothesis that financial access variables and error terms are normally distributed, such an assumption of error terms that are normally distributed does not hold with Quantile Regressions.
The Quantile Regressions models parameter estimates at numerous points of the conditional distirbution of financial access. Therefore, the technique is in conformity with the motivation of distinguishing nations with high, intermediate and low initial levels of financial development.
The  th quantile estimator of a financial access variable is derived by solving for the optimization problem in Eq (1) that is shown without subscripts for simplicity and ease of presentation.
Where unique slope parameters are modelled for each  th specific quintile. This formulation is analogous to in the Ordinary Least Squares slope where parameters are investigated only at the mean of the conditional distribution of financial development. For the model in Eq (2), the dependent variable i y is a financial development indicator while i x contains a constant term, foreign aid, trade, GDP growth, public investment and inflation. As specified in Eq. (2), the linearity in quantiles is appropriate under the assumption of homoscedasticity. This is essentially because if there is heteroscedasticity in the error process, then the quantiles will possess nonlinearities of different degrees. Table 1, Table 2, Table 3 and Table 4 respectively disclose results corresponding to financial depth, financial allocation efficiency, financial activity and financial size. Irrespective of tables, the left-hand-side (LHS) presents contemporary estimation whereas the right-hand-side (RHS) presents non-contemporary estimations. The motivation for lagging the independent variables on the RHS by one year is to account for endogeneity (see Mlachila et al., 2014;Asongu et al., 2016). The consistent variations in information sharing estimates between Ordinary Least Squares and Quantile Regressions (with respect to sign, significance and magnitude of significance) is a justification for the relevance of the problem statement, notably investigating the incidence of increasing information sharing offices for financial development when existing levels of financial development matter.

Financial development and Public Credit Registries
The findings are explained in three levels, namely, in terms of marginal effect, net effect and thresholds. The net effect of increasing public credit registries in the 0.10 th quintile on the LHS of Table 1 is computed with conditional and unconditional effects of public credit registries. Accordingly, the marginal or conditional effect (from the interaction) is 0.052 while the unconditional impact of public credit registries is -0.945. Hence, the corresponding net effect of increasing public credit registries is -0.832 ([2.155×0.052] + -0.945) 2 . Given that the conditional or marginal impact is positive, the correspond threshold in public credit registries at which the negative unconditional effect changes from negative to positive is 18.173 (0.945/0.052). The positive threshold is feasible because it is within the public credit registries range (minimum to maximum) disclosed by the summary statistics (0.00 to 49.80).
It is important to note that whereas the computation of net effects requires statistically significant unconditional and conditional effects, a threshold may be apparent even when only the conditional effect is significant. Consistent with Asongu and De Moor (2017), the notion of threshold is in accordance with Cummins (2000) on the minimum level/threshold in language proficiency before rewards are acquired in a second language. In addition, the definition of threshold is also supported by the critical mass theory which has been abundantly documented in the literature on economic development (see Roller & Waverman, 2001;Ashraf & Galor, 2013).
A recent application of the threshold or critical mass theory based on interaction variables can be 2 2.155 is the mean value of public credit registries. 14 found in Batuo (2015). Therefore, from the perspective of this study, the concept of threshold is not different from: (i) the minimum requirement for the reaping of positive effects (Cummins, 2000); (ii) conditions for U-shape and inverted U-shape (Ashraf & Galor, 2013) and (iii) critical mass for positive impacts (Roller & Waverman, 2001;Batuo, 2015).
Two key results can be established from Table 1  The following findings can be established from Table 2 on linkages between financial efficiency and public credit registries. In Panel A (on banking system efficiency) and Panel B (on financial system efficiency), most of the significant estimates are between the 25 th and 75 th quartiles, with positive net effects and negative thresholds that are not within range.
The main outcome from Table 3 on linkages between financial activity and public credit registries is shown in Panel A (on banking system activity) and Panel B (on financial system activity). It is noteworthy that most of the significant estimates are between the 25 th and 75 th quartiles, with positive net effects and negative thresholds that are not within range.    In Table 4 on the connections between financial size and public credit registries, most of the significant estimates are between the 25 th and 75 th quartiles, with positive net effects and negative thresholds that are not within range.
Most of the significant control variables have the expected signs. It is important to note that some of the signs may vary from one table to another because the financial development variables are by definition contradictory. The financial development variable of Table 2 (financial efficiency) is the ratio of the financial development variable in   Tables 1-4 on public credit registries, the marginal, threshold and net effects of private credit bureaus in Tables 5-8 are not clearly apparent. In order to examine why findings corresponding to private credit bureaus are not significant, we assess country-specific averages of information sharing offices which we disclosed in Appendix 4. From these country-specific averages, it is reasonable to infer that the findings on private credit bureaus are not very significant because of issues in degrees of freedom. Hence, the concluding implications that follow are essentially based on findings connected to public credit registries.

Concluding Implications, Caveats and Future Research Directions
The purpose of this study has been to investigate how increasing information sharing bureaus affect financial access. For this purpose, we have employed contemporary and noncontemporary interactive Quantile Regressions in 53 African countries for the period 2004-2011.
Information sharing bureaus are proxied with public credit registries and private credit bureaus.
Financial development dynamics associated with depth (at overall economic and financial system levels), efficiency (at banking and financial system levels), activity (from banking and financial system perspectives) and size are used. The following findings have been established.
First, the incidence of increasing private credit bureaus is not very apparent on financial access probably because private credit bureaus are still to be established in many countries.
Second, increasing public credit registries improve financial allocation efficiency and activity (or credit) between the 25 th and 75 th quartiles for the most part. This result is intuitive because for poorly developed financial systems, increasing information sharing bureaus may in some cases decrease the pace of development, whereas for more developed financial systems, information sharing may already have been taken into account. found that African countries with information sharing mechanisms for banks are associated with higher levels of financial development. The findings are also aligned with those of Galindo and Miller (2001) in the view that credit registries are more likely to enhance financial development compared to credit bureaus in less developed countries. On the other hand, our results appear not to be broadly in line with those of Love and Mylenko (2003) who argued that whereas the presence of private registries are associated with a higher share of bank lending and lower constraints on finance, public registries do not have a significant effect on financing constraints.
Our results do not also align with Triki and Gajigo (2014) who concluded that private credit bureaus are more positively sensitive to finance access, compared to public credit registries. Asongu et al. (2016) found that information sharing offices negatively affect financial access, for the most part while  concluded that financial development dynamics respond more positively to private credit bureaus relative to public credit registries.

20
There are three main shortcomings in the merit emphasis. First, we have not reported QR coefficients with which to substantiate the analysis in the study. Accordingly, we adopted a minimalist approach in the interpretation of estimated coefficients. The minimalist approach is based on the understanding that net effects are computed from corresponding significant increasing public credit registries is also worthwhile. A more robust quantile estimator with which to investigate these suggested lines of inquiry is that proposed by Canay (2011).
Accordingly, this estimator considers country-specific heterogeneity that is ignored in the current analysis.