Análisis de la encuesta dirigida a los estudiantes

The data period ranges from 2001:M1 to 2008:M12 and we have 96 available observations per bank. The reason for restricting the time period is because before 2001 the interest rate series for each bank were not available while at the beginning of 2009 a new accounting methodology was applied which distorts comparisons with the rest of the data series used in this analysis. Throughout the sample period we work with a balanced panel comprising 15 banks that have been operating continuously over the sample period. The SUR model is conducted on balanced panel data; although there are some recent developments in SUR methods for unbalanced data, these are still in the process of development.

The sample has been adjusted for mergers and acquisitions. The adjustment of banks‟ balance sheet items has been done by backward aggregation of the data series before the merger or acquisition occurred (for more details see section 5.3). Although this is the most commonly used approach in the literature (Ehrmann et al., 2003; Worms, 2003; Farinha and Marques, 2003; de Haan, 2003; Gambacorta, 2005; Havrylchyk and Jurzyk, 2005; Prutenau-Podpiera, 2007; Juks, 2004 and Benkovskis, 2008) and no other approach appears preferable, we have to be aware that this may give errors in the data because changes in the management of the merged bank and any gained know-how are not controlled for. In the case of merger we aggregate the data backwards as a weighted average of the value of the stock of loans and the respective interest rate of both entities. In doing this we have assumed, in the case of merger, that the management of both entities has also been merged and no single entity‟s retail rate setting strategy is taken as a

MACEDONIA – A SUR APPROACH

144 dominant one. This may be problematic in that after the merger the new entity may apply a completely different price setting strategy. Nevertheless, due to the relatively small cross-sectional sample and the need to work with balanced panel data, we think that this is the „second‟ best solution. In the case of acquisition, we have kept the lending rates of the acquiring bank before the acquisition has occurred, instead of backward aggregating the series as in the case of mergers. We argue that after the acquisition has occurred, the acquiring bank is likely to have maintained its previous retail rate setting strategy and has not changed or adopted the strategy of the acquired bank.

All variables in the model are expressed in nominal terms, except for the industrial production index which is in real terms. The balance sheet items such as total loans, long-term loans and long-term deposits are those of the non-financial private sector. Some of the balance sheet items15 such as total assets, total loans, long-term loans, gross deposits and liquidity have been seasonally adjusted by using the census X-12 additive method, which is the most commonly used seasonal adjustment method in the literature. The other available option was to include monthly dummies in each equation in order to control for the seasonality in the data. However, this method will reduce the degrees of freedom substantially and thus, we refrained from using this method.

In examining the determinants of banks‟ lending rate rigidity, we use the interest rate series on banks‟ outstanding loans for each bank separately. Interest rates on newly issued loans are likely to be much more responsive to changes in the „cost of funds‟ rate, but those data series are unavailable. However, in examining the effect of monetary policy changes, the retail rates of banks‟ outstanding loans are arguably preferable because they provide a more comprehensive picture about the cash flows of firms and households. If we had worked only with the interest rates of newly issued loans, then the pass-through coefficient would have been overestimated.

15 _{We have seasonally adjusted some but not all of the balance sheet items because a priori we do}

not expect a seasonal pattern in all balance sheet items such as: capital, NPL ratio etc. Additionally, we have checked whether there is a seasonal pattern in these balance sheet items where we might not expect seasonality and we did not find any.

MACEDONIA – A SUR APPROACH

145 Regarding the currency structure, we use an interest rate series of loans denominated in denars for the reasons explained in section 2.3.5. An additional reason for selecting the lending rates of loans denominated in domestic currency is because their reaction to changes in the domestic reference rate is the main focus of the monetary policy makers. Nevertheless, the interest rate series in denars also include the interest rates of foreign currency indexed loans and due to data unavailability we were unable to disaggregate them. Nonetheless, as can be seen on figure 1.22, these two interest rate series on aggregate level (the ones in domestic currency and the ones indexed to a foreign currency) have almost the same dynamics, except that latter have on average a lower level.

Concerning the frequency of the data, the interest rate series were available in both quarterly and monthly frequencies. However, due to the estimation method used (see section 3.3), as well as from the monetary policy makers perspective (see section 3.2), using higher frequency data is preferable. Therefore, we use monthly frequency data.

Regarding the construction of the interest rate series, the loan interest rates in denars from 2001 to 2003 include short-term loans with maturity up to 1 year for the corporate sector only, while for 2004 they include short-term loans with maturity up to 1 year of both corporate and household sector of all loan types by purpose. From 2005 to 2008, loan rates are constructed as weighted average of all maturities (short- and long-term) of both sectors (household and corporate). This methodological change in constructing the data series that occurred in 2005 is a limitation with the interest rate series. The chosen interest rate series includes the interest rates of loans on both a fixed and variable basis.

Investigating the interest rate adjustment separately for the household and corporate sectors, as well as the adjustment of different types of loans by purpose may be more appropriate because, according to the empirical literature presented in section 2.3, not all loan interest rates adjust equally to changes in the „cost of funds‟ rate (Sander and Kleimeier, 2004a, b; Lago-Gonzalez and Salas-Fumas, 2005; Egert et al., 2007; Sorensen and Werner, 2006). For example, loan rates on

MACEDONIA – A SUR APPROACH

146 household sector are typically found to adjust more sluggishly then the ones in the corporate sector, while interest rates on mortgage loans are found to adjust more sluggishly to changes in the referent rate compared to interest rates on consumer loans. Nonetheless, such disaggregated interest rate series are not available.

Apart from the interest rate series, there are also some limitations of the rest of the data series, i.e. banks‟ balance sheet items in terms of their reliability, methodological consistence and the way they have been collected and backward revised (see section 5.3). However, these are perceived as minor and unlikely to significantly affect the results.

A detailed description of each data series is presented in table 3.2. Table 3.2: Data description

Source: NBRM and SSO.

Variable: Description: Value: Source:

Lending rate Weighted average monthly loan rates for each bank separately %, annualised NBRM

'Cost of funds' rate Weighted average MBKS rate %, annualised NBRM

Bank size Log of total assets Nominal NBRM

Liquidity

Ratio of liquid over total assets. Liquid assets include: cash in vault at the NBRM + short term deposits in accounts in banks abroad + CB Bills and treasury bills with maturity up to 1 year + cash in vaults in domestic banks + short term restricted deposits in accounts in domestic banks + short term loans granted to domestic financial institutions (banks and saving houses).

Nominal NBRM

Capital Ratio of equity plus reserves to total assets. Nominal NBRM

Credit risk exposure Ratio of NPL to total loans. Nominal NBRM

Maturity-mismatch Ratio of long-term loans to long-term deposits and long-term

borrowings from abroad. Nominal NBRM

Relationship lending Ratio of long-term loans to total loans. Nominal NBRM Operational efficiency Ratio of administrative costs to total costs. Nominal NBRM Portfolio diversification Ratio of non-interest income to gross income. Nominal NBRM

Price changes Annual rate of inflation, measured by CPI. % SSO

Economic growth Annual rate of growth of IPI. % SSO

Market concentration Log of HHI and (HHI)2 Index

Author's own calculations upon the

MACEDONIA – A SUR APPROACH

147 The summary statistics of each variable as described in table 3.2 is presented in table below.

Table 3.3: Summary statistics

Source: Author‟s own calculations performed in EViews 6.