Working Paper Series: Special Edition of 2016 to 2018 Interns

growth, audit, exporter, duration) that lenders look for when assessing firms’ creditworthiness. Country is a set of variables representing the country characteristics. They are GDP per-capita and real interest rate 5 . GDP per-capita measures the wealth and depth of financial development of a country (Beck & Cull, 2014), while real interest rate measures the level of stability in the economy. From henceforth this shall be called model 1. , � + In this model demand is the outcome variable and is dichotomous in nature identifying whether the ℎ firm in country j demands credit or not. Similar to model 1 firm specific represents those key policy variables as well as those that speak to the characteristics of the firms (technical, foreign, larger, size, legal, education). Credit assessment in this model is the set of those characteristics (innovation, overdraft, collateral, development, growth, audit, exporter, duration) that determine the firm’s creditworthiness. Country variables in this model are the same as those identified in model 1. Hereafter, this model shall be referred to as model 2. It should be noted that higher the real rate of interest the lower the demand for credit would be (Nistor & Popescu, 2013). Collinearity test is also employed to determine the extent of correlation among the explanatory covariates. If there is high correlation among the explanatory variables then most likely the model would produce unstable results and not be a good fit. As such, the variables were subjected to collinearity test. Those found to be correlated were removed from the models. A benchmark of r > 0.5 was used to determine which variables should be booted from the model (see table 21-24) 6 . For the sub-sample of ECCU countries in model 1, the variables collateral and real were dropped. For model 2 the variable real was dropped as it showed a high level of correlation with the other covariates. 2. Pr ( ) = � ,

5 The real interest rate can be computed by subtracting the inflation rate from the nominal rate of interest. In the literature it is often postulated that inflation and interest rate move together thus making the real interest rate stable in the long-run. 6 Pairwise correlation test is employed due to its ability to take into account observations with missing data points.

21