Explanatory variables

As explained above, the control variables considered in this chapter are determined on the basis of the Monti-Klein model of the banking firm. Most of the explanatory variables applied in this chapter are the same as those used in Chapter 3¹⁴, excepting only the measures of loan demand elasticity and the merger dummy variable. In Chapter 3, the GDP growth rate is used as a proxy for the bank’s loan demand elasticity. Unfortunately, given that 1997 is the earliest year in our sample - the pre-treatment year in our analysis - no data of the GDP growth rate for 1997 are available.

Therefore, an alternative measurement is used in this chapter. Following Demirguc-Kunt and Huizinga (1998) and Focarelli and Panetta (2003), we consider the GDP per capita at nominal values, GDPCAP, as a proxy for the loan demand elasticity. This is the ratio which divides the value of the GDP by the country’s population (in thousands) for the same year. The values of the GDP and the population total were obtained from the ECB (2002, p.65; and 2006, p.65). For the merger dummy variable, as previously stated, the larger the number of observations, the more reliable the analysis. Therefore, to increase the number of observations, we apply in this chapter a different merger dummy variable. Following Sapienza (2002), we use the accumulative merger dummy variable, MDUMMY, which has a value equal to 1 in all the calendar years after a merger, and otherwise is zero. The coefficient of this dummy shows the counterfactual of merger effects.

In addition, there is an additional control variable to those shown in Chapter 3: EDEP, the ratio of deposits to GDP, used as a control variable for the elasticity of supply for

14 The full specifications for these variables are provided in Chapter 3.

deposits in the interest margin regression. Because the price elasticity of the deposit supply depends on cyclical macroeconomic factors, the potential impact on the price elasticity of the economic cycle can be captured by including the GDP as an explanatory variable (Fuentes and Sastre, 1998, p.9). Although there are alternative measurements, such as the interest insensitivity in deposit accounts used in the study of Khawaja and Din (2007), determining the interest spread in Pakistan’s banking industry, due to the limitations of the data which do not separate types of deposit account according to interest sensitivity, the total ratio of deposits to GDP, EDEP, is considered in this analysis as a proxy for the elasticity of supply for deposits. The data of aggregate deposits of each nation and the GDP were obtained from the ECB (2002, pp.57, 65; and 2006, pp.57, 65).

As in Chapter 3, following Ayadi and Pujals (2005), the ratio of deposits to total assets, DEPTA, is used to capture the bank deposit characteristics. Bank total deposit is transformed over the bank’s total assets to remove the nominal variations which could result from the size differences among banks. The data of deposits and total assets are from balance sheet statements obtained from the BankScope database. The cost-to-income ratio, COST, is the reported cost-to-income-ratio taken from income statements. This ratio is used by several studies, for example, those of Corvoisier and Gropp (2001), Focarelli and Panetta (2003), Gambacorta (2004), Ayadi and Pujal (2005) and Altunbas and Ibanez (2008), to measure the marginal cost of issuing loans and also to control for the difference in bank efficiency and productivity. As in Chapter 3, the default risk could be indicated by the ratio of loan loss provision to net interest revenue, LLOSS, obtained from bank income statements. Although there is some debate about the appropriateness of using loan loss provision as a credit risk

measure, since it is impossible to obtain information on NPLs in several European countries, the ratio of loan loss provision to net interest revenue, which is the figure most widely published in Europe (Altunbas and Ibanez, 2008, p.220), is applied in this chapter as a proxy for the bank’s default risk. In addition, the validity of loan loss provision as a default risk measurement is confirmed by the studies of Ahmed (1999), Demirguc-Kunt and Huizinga (1998), Fisher, Gueyie and Ortiz (2000), Ismail and Lay (2002), Nys (2003), Ayadi and Pujal (2005) and Altunbas and Ibanez (2008).

Following Nys (2003) and Mercieca, Schaeck and Wolfe (2009), the ratio of net loan to total deposit and short-term borrowing, NLD, is used to measure a bank’s liquidity risk. This ratio is reported in the bank income statements obtained from the BankScope database. It shows the extent to which a bank is dependent on volatile liabilities. The higher ratio indicates higher dependency and the high risk of suffering from unexpected deposit withdrawal, or, in other words, high liquidity risk. The three-month inter-bank interest rate, SHRT3, is considered a measurement for the market interest rate. It is an annual average amount quoted as a percentage, obtained from the Eurostat database. This three-month rate is the interest rate applied between banks with an original maturity of three months. The same three-month inter-bank interest rate is also used in the studies of Nys (2003), Panetta, Schivardi and Shum (2004) and Banal-Estonol and Ottaviani (2007) to capture a substitution effect, that is, substituting marketable assets for loans. The market structure is proxied by the five-firm concentration ratio, CR5. This ratio was obtained from the studies of the ECB (2002, p.54; and 2006, p.54). It is the sum of the market shares, in terms of total assets, of the five largest banks in a national market. We use this ratio following the reasoning by Moschandreas (2000, p.14), who suggests that the five firm concentration ratio is commonly used to investigate bank market competition in the

UK, and by other studies examining the European banking market structure such as those by Fernandez de Guevara, Maudos and Perez (2005), Groeneveld and Boonstra (2005), Baert and Vander Vennet (2009) and Casu and Girardone (2005, 2009). In addition, as suggested by Sapienza (2002, p. 345), there could be a relationship between the merger effect and market competition. Therefore, the interaction term for these two variables, MCR5, is also included as a proxy for this correlation.

Moreover, concerning the DID method, we have to consider the time effect. This effect can be controlled by the dummy variable, YDUMMY. This is the post-merger index, which is the binary index equalling 1 if the year is the post-merger period, and zero otherwise. The coefficient of this variable presents the time trend effect, in the absence of M&As. Finally, the DID effect of bank M&As is examined by the dummy variable, DID. This is the M&As effect, which equals 1 if a bank engages in M&As in the post-merger year, and zero otherwise. Its coefficient presents the true effects of bank mergers, comparing pre- and post-merger behaviour and between merged and non-merged banks.