Moderating effect of inflation on the finance–growth nexus: insights from West African countries

The positive impact of finance on growth has been reported in the literature (see Beck et al. 2000; Levine et al. 2000; Muhammad et al. 2016; Rousseau and D’onofrio 2013; Stolbov 2017). However, other views, such as demand-following, feedback and neutrality hypotheses are also gathering momentum in recent studies (see Acaravci et al. 2009; Apergis et al. 2007; Demetriades and James 2011; Gozgor 2015; Kalaitzoglou and Durgheu 2016; Peia and Roszbach 2015).

The level of financial development, level of per capita income, quality of institutions, financial integration, financial structure and political structure are some of the variables highlighted in the literature that moderate the impact of finance on growth (see Gehringer 2013; Law et al. 2013; Law and Singh 2014; Rioja and Valev 2004a, b).

For an excellent survey on the linear and nonlinear impact of inflation on economic growth, see these studies Barro (1996), Bruno and Easterly (1998), Eggoh and Khan (2014), Fischer (1993) and Gillman and Kejak (2005).

The inflation rates data were obtained from the World Economic Outlook (2016), while the data on credit to private sector relative to GDP were obtained from the World Development Indicators (2016).

Hung (2003) used the endogenous growth model to provide the theoretical background concerning the influence of inflation on the relationship between finance and growth. He posited that in countries with relatively high inflation rates, financial development raises inflation and diminishes economic growth, whereas financial development reduces inflation and spurs growth in countries with relatively low initial inflation.

In the finance-growth literature, many studies have used real GDP per capita as the dependent variable with varying connotations such as economic growth, output or economic development (see Chortareas et al. 2015; Demetriades and Hussein 1996; Demetriades and Law 2006; Law et al. 2013; Wolde-Rufael 2009). Specifically, Law et al. (2013) showed that irrespective of whether the growth rates of real GDP per capita or real GDP per capita are used as the dependent variable in the examination of the finance–growth nexus, the estimation results are essentially similar.

The PMG and MG estimators are appropriate for this study given the length of time series (T) and number of countries (N). These techniques have been employed in recent studies (see Demetriades and Law 2006; Iwata et al. 2011; Kim and Lin 2010).

See Demetriades and Law (2006), Iwata et al. (2011), Pesaran and Smith (1995), and Pesaran et al. (1999) for more information on the derivation of the MG and PMG estimators.

See Wooldridge (2012) for more details concerning the IV approach. The order condition for identification states that for each included endogenous variable, there should be at least one instrument (exogenous variable) that satisfies an exclusion restriction so that the equation can be identified.

SUR was proposed by Zellner (1962). This empirical strategy was also used by Bittencourt (2011). Pesaran (2006) revealed that parameters estimates could be substantially biased, and their sizes could be distorted if cross-sectional dependence is overlooked. To test for the presence of cross-sectional dependence among the countries in the panel, this study uses the Lagrange Multiplier (LM) test proposed by Breusch and Pagan (1980); the scaled CD_LM and general CD tests developed by Pesaran (2004) as well as the bias adjusted LM test developed by Pesaran et al. (2008). The results of the test statistics are 940.6***, 51.937***, 18.6*** and 51.7***, respectively. All the tests are statistically significant at the 5% level, thereby rejecting the null hypothesis of no cross-sectional dependence. This implies that cross-sectional dependence is present, and, hence, justifies the use of SUR.

For more information on the computation of marginal effects, see some empirical studies (Baltagi et al. 2009; Brambor et al. 2006; Law et al. 2017; Moradbeigi and Law 2016).

We chose to evaluate the marginal effects using the minimum, mean, and maximum levels of inflation rates since the marginal effects vary within the sample depending on the level of the inflation rate (see Baltagi et al. 2009).

The regression that included dummy variables to control for structural breaks has similar results.

Result is available on request.

Irrespective of whether or not the data are averaged over a 5-year non-overlapping period, Checherita-Westphal and Rother (2012) showed that the instrumental variables (IV) techniques produce similar estimation results in terms of the sign and significance of the coefficients.

We thank the anonymous reviewer for this comment. We acknowledge that the parameters in Eq. (1) could be estimated using different estimators. We chose to use the MG and PMG approaches because of the integration and cointegration properties of the variables in our model, as well as to obtain the inherent advantages in these estimators. The variables in our model are a mixture of I(0) and I(1), and the time period (T) is larger than the number of cross sections (N). Hence, we are of the opinion that it would be prudent to employ the MG and the PMG, which are based on the ARDL, and can be used irrespective of the order of the integration of the variables. Nonetheless, we are aware that the use of the MG and PMG estimators may not be able to fully address the issue of endogeneity as adequately as compared to the generalized method of moments (GMM). However, for this study, we are constrained in this respect because the GMM approach requires a criteria of N > T. So, in order to address the issue of endogeneity, we have chosen to employ the IV approach as a compensating measure. Although we have included time trends in our IV regression, we are acutely aware that some of these instruments do not vary over time. Hence, we should conservatively advise that the findings should be interpreted cautiously. Furthermore, we suggest that any future research undertakings should possibly try to increase the number of sample countries being researched or that such future studies should utilize a five-year average of the data to enable it apply to the GMM system, which is far more potent in addressing the question of endogeneity.

In other countries where the coefficients of the linear financial development and interaction terms are positive, the average credit to private sector and inflation rates are, respectively, presented in parenthesis: Niger (10.79 and 3.20%), Cape Verde (26.93 and 6.25%), Benin (18.25 and 4.08%), Burkina Faso (14.28 and 3.76%). Conversely, in other countries where the coefficients of the interaction terms are negative or statistically insignificant, the average credit to private sector and inflation rate are, respectively, presented in parenthesis: Guinea (4.42 and 18.85%), Guinea-Bissau (6.71 and 28.31%), Gambia (12.18 and 9.17%), Nigeria (14.92 and 19.74%), Liberia (11.24 and 10.36%), Sierra Leone (4.20 and 33.67%), Togo (21.81 and 4.62%).

The eight UEMOA members (Benin, Burkina Faso, Cote d’Ivoire, Guinea-Bissau, Mali, Niger, Senegal, and Togo) have one common regional central bank, one common stock market, one common currency as well as one legal/regulatory framework for the banking system for all member countries. Conversely, the other eight non-UEMOA members (Cape Verde, Gambia, Ghana, Guinea, Liberia, Mauritania, Nigeria, and Sierra Leone) have different central banks, currencies, stock markets and banking legal/regulatory frameworks.