Types of bank loans and their impact on economic development: a case study of the Czech republic

34 2017, XX, 4 Ekonomie DOI: 10.15240/tul/001/2017-4-003 Introduction In this article, we discuss the relationship of banks, or loans provided by them, and economic development. We decided to investigate this relationship as banks in today’s economies play a signifi cant role as a vital institution in the fi nancial markets, where there is a distribution of monetary funds from surplus entities to defi cit entities. A necessary precondition of a functioning economy, in its present mostly mixed form, is a functioning and stable banking system. Currently, uncovered money is to a great extent the money generated by private banks, mainly in the form of loans. In these turbulent times, when economies are growing and declining at faster intervals than were customary in previous decades, this is very much a contemporary issue. It is due to the fact that the rate of growth/decline in lending, due to the importance and size of the fi nancial markets and the form of the uncovered money issued, signifi cantly infl uences the economic cycle. As reported by Černohorský (2015), banks provide loans to businesses and households for their consumption and investment and thus support the economy. What is important is the duration of a loan, as particularly long-term investments contribute to long-term economic growth. For this reason, we decided to examine the impact of total loans, as well as dividing them into loans to non-fi nancial businesses, loans to households, mortgage loans and consumer loans. The importance of credit access to banks is also compounded by the fi nancial and economic crisis which most of the developed countries experienced in recent years. The consequence of this crisis today is an abnormal situation on the fi nancial markets, which is refl ected in negative interest rates, foreign exchange intervention and quantitative easing by central banks. They are trying to use these unconventional monetary policies to restore the impaired credit channel of the monetary policy transmission mechanism. The main idea of this article is expressed by the hypothesis that the development of various types of bank lending has a positive effect on economic development. We will examine the validity of this hypothesis using selected statistical and mathematical methods as presented below. The aim of this article is to assess the impact of the development of different types of loans in the banking sector on economic development, based on the example of the Czech Republic. In achieving this set goal, we shall begin with the hypothesis that economic performance increases with the growth of the rate of various types of loans. 1. Theoretical Background In the past, the relationship of the fi nancial system and economic growth was examined by a number of renowned economists. Schumpeter (1912) emphasised the strong infl uence of banks on economic growth by encouraging innovation. Providing loans for these innovations and investments leads to the growth of business operations and thus to economic growth. In contrast, Lucas (1988), in more modern times, refers to an excessive infl uence of banks on the economy in a negative sense. Robinson (1952) considers that banks have a passive impact on economic development. As can be seen, even in the past these distinguished economists held different views on the impact of the banking sector on economic development. It is still the same today, as evidenced by subsequent research. Our paper is based on the transmission mechanism of monetary policy as it is understood both in economic theory and applied in practice in the enforcement of monetary policy by central banks. In particular, TYPES OF BANK LOANS AND THEIR IMPACT ON ECONOMIC DEVELOPMENT: A CASE STUDY OF THE CZECH REPUBLIC Jan Černohorský EM_4_2017.indd 34EM_4_2017.indd 34 13.12.2017 12:53:2813.12.2017 12:53:28 35 4, XX, 2017 Economics we follow the logic of the credit channel of the monetary policy transmission mechanism. This is based on the change in the interest rates set by the central bank, which affect interbank interest rates and in the end also the market interest rates offered to clients. If we consider declining interest rates, this results in a higher demand for loans from banks by companies and households. These loans are used for corporate investment and household consumption (or from a macroeconomic point of view also for investments through the purchase of real estate). At the same time, the amount of money in circulation is growing. The increase in consumption and investment thus contributes to the growth of the economy. In the case of increasing interest rates, the change in the given variables is the opposite or it may, for example, result not only in a decrease, but also in a decline in the growth rate, of the given quantities. A defi nite factor in support of this process is the central banks monitoring and striving to infl uence the rate of lending to a certain extent in order to support economic growth or to ensure that the economy doesn’t get overheated. The effect of the amount of money issued as bank loans on economic development is highlighted by the main proponent of monetarism, Milton Friedman (1968). As well, Friedman and Schwartz (1963) came to the conclusion that the correlation coeffi cients between the change in money and the nominal output range from 0.79 to 0.92 per survey period. They quantify the time delays in the effectiveness of monetary policy in a range of 12-24 months, with the maximum growth in the amount of money being in advance of 18 months ahead of the peak of economic growth. The minimum amount of money growth, according to his calculations, will be refl ected in the economy in the form of a recession earlier, with an interval of approximately 12 months. Also, the money supply is understood as an autonomous exogenous quantity given by the central bank, which affects other macroeconomic variables. The logic of this approach is supported by Miskhkin (2016), who, in addition to the credit channel, also defi nes other channels of monetary policy action. Kaufmann and Kugler (2010) also estimate real GDP on the basis of the development of M3, including the aspect of cointegration of the given variables. This idea is supported by Holtemöller (2004), who sets the time delay of the monetary policy tools on product changes at six quarters. For this he uses integration and cointegration analysis. Currently, the connection between bank performance (measured, for example, in the form of lending rate) and economic performance is much closer. This is due to the enormous scale of globalised and also local fi nancial markets due to the size of the economies and their impact on business activities. Today there is a higher degree of interconnectivity of fi nancial markets and economic development. The fi nal proof is certainly the fi nancial crisis in the USA. It developed primarily in the banking sector and spilled over into a public fi nance crisis and an economic downturn in the economically important countries in the world and Europe. Therefore examining the relationship between bank lending and economic development has gained importance. Among the various works various indicators of lending are used to measure economic performance. These contributions can be divided into three basic groups. The most signifi cant in terms of numbers is the group of economists who believe in the positive impact of bank loans on economic development. Levine and his coeconomists in their works (Levine & Zervos, 1998; Beck, Levine, & Loayza, 2000; Beck & Levine, 2004) examined various combinations of the effects of the liquidity of stock markets and banks (collectively, fi nancial intermediation) on economic growth, capital accumulation and increased productivity. All the above, according to the authors, is positively infl uenced by the activities of banks. Armeanu et al. (2015) tested the effects of credit expansion on sustainable economic growth. They see a greater effect with loans to legal entities rather than to natural persons. The importance of loans to legal entities (companies) acts over a longer period, because their investments lead to further growth. Banu (2013) focused on the question of whether an economy, specifi cally the Romanian economy, would be capable of economic growth in the absence of lending. Without the loans provided to the private sector, the Romanian economy would not grow, as no new projects would arise. Conversely, very low dependence was found between loans to the public sector and economic growth. Kelly et al. (2013) began with a range of data from 10 years for the economy of Ireland, which was signifi cantly affected EM_4_2017.indd 35EM_4_2017.indd 35 13.12.2017 12:53:2913.12.2017 12:53:29 36 2017, XX, 4 Ekonomie by the fi nancial crisis specifi cally because of the banking sector. Despite this signifi cant fl uctuation, Kelly fi nds a positive impact of lending activities on growth in the economy. As well, Ermisoglu et al. (2013) investigated whether data on loans would be an appropriate forecast for the development of gross domestic product (GDP). They stressed the importance of loan data in terms of a minimum delay. Again, they found a positive effect; i.e., they state that using the variable “credit incentives” in GDP prediction models increases their accuracy. The results of a further study by Cetorelli and Gamber (2001) show that the banking sector facilitates access to credit for “young” fi rms, thereby supporting the pace of economic growth, as investments by new fi rms are more likely to be involved in innovative technologies. Bencivenga and Smith (1993) conclude that the banking sector can also reduce excessive credit limitation through reduced monitoring costs and thus ensure accelerated economic growth in a country. Levine (2005) shows the link between the operation of the fi nancial system and economic growth. On the other hand, there are studies that show a negative relationship between bank loans and economic development as measured by GDP growth. Leitao (2012) came to this conclusion based on an analysis of macroeconomic variables (economic growth, trade balance and infl ation) and bank loans. He concluded that infl ation is negatively correlated with economic growth. The main idea behind the study is that excessive credit growth tends to weaken a banking system and increase infl ationary pressures, thereby undermining economic growth. Mian et al. (2015) based their study on an analysis of the relationship between household debt and GDP. According to their results, the growth of household debt in relation to GDP predicts a lower growth in production and higher unemployment rates in the medium term. As well, an increase in household debt will result in consumption growth and worsening current account balances as a result of the increased import of consumer goods. Koivu (2002) published a study based on data from 25 transitional economies in the years 1993-2000. In his work, he concluded that an increase in lending does not accelerate economic growth. The causes are a series of banking crises in these economies and fi scal restraint. He also stressed that these results are non-standard with economic fi ndings primarily due to specifi c conditions in transition economies. IbáñezHernández et al. (2015) came to the conclusion that the high growth in lending leads to instability in the fi nancial sector and thus negatively affects economic development. There are also studies that do not indicate any signifi cant relationship between the loans provided and economic growth. For example, Takats and Upper (2013) investigated the effect of bank loans on economic growth after the fi nancial crisis on the basis of data from 39 fi nancial crises that had been preceded by a credit boom. They found that a declining amount of bank lending to the private sector does not necessarily hinder economic recovery after a fi nancial crisis. In these crises, changes in the rate of bank lending, either in real terms or in relation to GDP, do not correlate with growth during the fi rst two years of recovery. In the third and fourth year, the relationship becomes statistically signifi cant, but still remains insignifi cant in economic terms. De Gregorio and Guidotti (1995) found a positive correlation between the growth rate of bank loans to the private sector and the growth of GDP, but the impact varies in different countries. In Latin American countries, the relationship is actually negative. Their rationale was the recent fi nancial liberalisation in these markets combined with a poor level of regulatory framework. They also emphasise that the main method whereby the growth of lending affects economic growth is primarily that these loans must be provided for effective projects; i.e., not a critical amount of these loans. Based on the list mentioned above, it is clear that studies are prevalent which confi rm the logic of the credit channel of the monetary policy transmission mechanism and show a positive relationship between the growth of lending and the growth of the economy. This corresponds to the current prevailing theoretical knowledge of bank contributions through money issuance by providing bank loans to grow the economy. As well, we are aware of the interdependence of the effects of bank loans and the development of the economy in both directions (i.e., acting as a multiplier and accelerator). However, in this article we have focused on the impact of bank loans on the development of the economy. A two-way relationship is also taken into account in the discussion of the results achieved. EM_4_2017.indd 36EM_4_2017.indd 36 13.12.2017 12:53:2913.12.2017 12:53:29 37 4, XX, 2017 Economics 2. Methods This article focuses on examining the relationship between two variables – loans granted and economic development. It is clear that in economic practice, there are a number of factors which affect economic development. We have drafted our analysis on the basis of the ceteris paribus condition, which simplifi es the real relationship, but is still suitable for examining the relationship of two variables. In the fi nal discussion, we also defi ne the factors that will otherwise defi nitely have an impact on the development of the economy. In this work we have decided to use cointegration analysis; i.e., a method that distinguishes short and long term relationships of time series. This is a relatively modern method used in many of the studies mentioned above and studies of central banks. The result of this is whether the time series are cointegrated or not. Cointegration means that the deviation in the directions of the development of the time series can only be short-term, and there is a limit beyond which the deviation may not continue. The time series are then in equilibrium and have a long-term relationship between them; i.e., they have a common element that can be examined (Arlt & Arlt, 2007). The advantage of this method over traditional statistical methods is that it identifi es any apparent regression. The analysis model selected is designed in accordance with professional analyses and based on the specifi c characteristics of the time sequence. The model is created for testing delays of the dependent variable of GDP and stationarity testing, including necessary adjustments of data by differencing. Cointegration is then tested and the fi nal test is to perform Granger causality. The fi rst step is the need to test the time sequence on the optimum order of delays for the dependent variable GDP. To determine the delay, we used a calculation using Akaike’s information criterion (AIC) in the equation below: (1) where M defi nes the number of parameters in the model, is the residual variance, and T is the number of observations. The best range of delay is the one where the information criterion reaches the lowest values. The test outputs of the best range of delay are applied in the following tests. An important prerequisite before testing cointegration is to verify the stationarity of the time sequence being input to the model. Stationarity of the time series is required in order to estimate the regression model. In the case of non-stationary data and modelling using the least squares method, the analysis could have distorted outcomes and could present an apparent regression. In the case of non-stationary time sequence, adjustment should be made using differentiation of the original data. A stochastic process is a timeordered set of random variables, which in theory may be viewed as a function of mean value, variance, covariance and correlation functions. A stochastic process is thus referred to as stationary if the characteristics of the random variable are time constant. These conditions are formally written as follows (Arlt et al., 2007):  Mean value function: (2)  Variation function: (3)  Covariance function: (4)  Correlation function: (5) where Xt is the dependent variable, E(Xt) is the mean value D(Xt) is the variance. Stationarity verifi cation is performed using the extended Dickey-Fuller test (ADF test) to test the hypothesis of the existence of a unit root. The test is based on regression of the fi rst differences of the time sequence based on their own delayed values, or the delayed differences. In practice there are three forms of the ADF tests: without a constant, with a constant, and last is with a constant and a trend. The selection is made through the lowest Akaike criteria. Evaluation is based on an assessment of the null hypothesis when it is tested at a signifi cance level of 0.05, if the time sequence has a unit root. Then we may say that the time sequence is non-stationary. EM_4_2017.indd 37EM_4_2017.indd 37 13.12.2017 12:53:2913.12.2017 12:53:29 38 2017, XX, 4 Ekonomie Verifi cation of the null hypothesis is evaluated based on the calculated p-values. In testing, we assume that the generating process has the form (Arlt et al., 2007): (6) where we test that Ø = 0 (variable contains a unit root), Xt is the dependent variable, p is a delay and et is a residual component. In the event that it is a non-stationary time sequence, it is necessary to adjust the time sequence by using the fi rst difference. Based on the new values of the time sequence we decide on its stationarity. If the input time sequences are nonstationary and after adjustment by differentiation, they acquire stationarity of the same order, it is possible to perform a cointegration analysis. When the above conditions are met, the Engle-Granger test (EG test) will be applied on the time sequence to determine the cointegration of the time sequence. This test is based on testing the estimated residues of the cointegrating regression for the presence of a unit root. Cointegration regression will be performed using the smallest squares method. In accordance with the Engle-Granger theories in the next step, random components are tested using the ADF test for the presence of unit roots. Evaluation of this test is identical to the ADF test mentioned previously, including the selection of the type of regression model by the lowest AIC. We will test the null hypothesis that the time sequences are not cointegrated at a signifi cance level of 0.05. If the p-value for the residues tested is higher than the level of signifi cance of 0.05, we will not reject the null hypothesis and the tested time sequences are not cointegrated. The variables are then tested for a possible mutual causal link between the tested series on the basis of Granger causation. The coeffi cient of determination observed, or the corrected (adjusted) coeffi cient of determination describes the closeness of the connection. The resulting value can be interpreted in terms of percentage, while indicating what percentage the changes in the response variables are dependent on changes in the explanatory variables. The coeffi cient of determination indicates the quality of the regression model; expressed more precisely, it indicates what percentage of variance of the response variables is explained by the model and how much remains unexplained. The fi nal test is to test the causal link between the time sequences, using the Granger causality analysis. Granger defi ned the concept of causality in the practical use of vector autoregression models (VAR models) for restricted and unrestricted regression. The basic idea is that when a series X affects a series Y, then the series X should improve predictions for the series Y (Hendl, 2012). VAR models are based on a comparison of residues of individual models differing in the number of delays. The most suitable model is chosen of a type that has a minimum value of AIC. For the Granger causality test we will use the null hypothesis that the variable X does not affect the variable Y under Granger’s conditions. The basic models take the following form (Hušek, 2007): (7) (8) where αi and βi are the coeffi cients of the variables, Xt and Yt are time sequence variables, p is the delay and ut is the random component. The fi rst equation estimates the dependent variable based on its own delayed values, the second equation adds to its own delayed values the delayed values of the fi rst variable. The test is conducted using VAR models in which an interaction of up to eight delays is tested. We reject the null hypothesis if the p-value is less than the signifi cance level of 0.05. Hnízdo (2015) further explains the issue in detail. 3. Data Data on loans are taken from the Czech National Bank (ČNB) database. These are loans to non-fi nancial businesses, loans to households, mortgage loans, consumer loans and total loans. All variables are in the form of relative annual change in a quarterly frequency. These data are seasonally adjusted for the time period 2004-2015. The time series is based on fi nancial market developments and changes in the Czech economy. One reason for setting this series is the fact that before 2004 there were signifi cant changes in the banking sector in the Czech Republic, which changed the ownership structure, and the government had intervene to EM_4_2017.indd 38EM_4_2017.indd 38 13.12.2017 12:53:2913.12.2017 12:53:29 39 4, XX, 2017 Economics stabilise the banking sector and clear the debts of declining large banking companies. The data are shown in the following table. Development of the economy is measured using standard indicators of gross domestic product in the Czech Republic, reported in the statistics of the Czech Statistical Offi ce and also referred to in the database of the Czech National Bank. Again, for comparative purposes, these variables are in the form of relative annual changes in a quarterly frequency. These data are shown in the following fi gure (Fig. 1). Fig. 1 shows the strong growth of all components of loans granted in 2005-2008, household loans and mortgages from an earlier period. This is related to the rapid growth of the economy, including exports, which are, among other things, driven by investment activity. This is widely funded by loans provided to businesses as well as by mortgage loans. In addition, interest rates have fallen sharply at this time, contributing to a greater willingness particularly among households to incur debts. Another factor is undoubtedly the demographic development, where a signifi cant portion of the population in this period dealt with their housing needs. After 2008, on the other hand, the effects of the fi nancial crisis in developed countries begin to show. These have manifested themselves in the Czech Republic in the economic crisis and the decline in investment and credit activity resulting from a crisis of confi dence; i.e., due to the caution of banks in granting loans. Variables are examined in absolute values and in inter-annual changes. Tab. 1 shows descriptions of the variables used. Fig. 1: Development of GDP and selected types of loans (annual change in %) Source: Czech National Bank (2016) EM_4_2017.indd 39EM_4_2017.indd 39 13.12.2017 12:53:2913.12.2017 12:53:29 40 2017, XX, 4 Ekonomie 4. Results As mentioned above, we are interested in whether there is cointegration among the selected variables. That is, whether the given time series evolve similarly over the long term. This is, from an economic point of view, the underlying idea of the credit channel of the monetary policy transmission mechanism. In the short term, based on the interpretation of the cointegration analysis, some discrepancies may occur. Based on the model chosen, an optimum order of delay is tested, as well as data stationarity, cointegration test and subsequently the Granger causality test is performed. For this model, we fi rst determined the optimum delay order on the basis of AIC (according to Formula 1) for GDP in absolute terms as well as inter-annual changes. The results are shown in the Tab. 2. Based on the lowest value of AIC, we can conclude that for the dependent variable GDP, the optimum delay is that of the second order. The semi-annual delay identifi ed will be refl ected in subsequent tests. We will proceed to verify the stationarity of the time sequence. The fi ndings of stationarity in the time series will be made using the extended Dickey-Fuller test (see formula 6). A null hypothesis is used for the ADF test when the time sequences tested are not stationary. Stationarity test results of the series in absolute values and inter-annual changes to GDP and total loans are shown in the Tab. 3. The results of the ADF unit root test indicate that the original data for all time sequences are non-stationary. Non-stationarity of the time series means that apparent correlation could occur for the correlation analysis. Stationarity for all time sequence was achieved only after their differentiation and the time sequences are therefore integrated in stage I (1); see Part 2 of the Tab. 3. Based on the results shown above, we can proceed to the cointegration test. The cointegration test is performed using the Engle-Granger test (see formula 2). This test requires non-stationarity of the original time sequence and the same degree of integration. Both conditions are shown in Tab. 3. The null hypothesis for cointegration is that the time sequences tested are not cointegrated. The type of test chosen is based on the lowest value of the Akaike criterion. Testing cointegration relationships for variables in absolute values is performed using the EG test, where the model is chosen with a constant and trend based on the lowest value of Variable Macroeconomic value Unit Source Y Gross domestic product bn. CZK ČNB Uc Total loans bn. CZK ČNB ∆Y GDP growth % ČNB ∆Uc Total loan growth % ČNB Source: own Order of Delays AIC for ∆Y 13.41229 22.94612 32.99098 43.03142 Source: own Tab. 1: Defi nition of the variables used in the analyses Tab. 2: Results of optimum order of delay EM_4_2017.indd 40EM_4_2017.indd 40 13.12.2017 12:53:3013.12.2017 12:53:30 41 4, XX, 2017 Economics AIC, which amounts to 1,121.736. The resulting calculated values for determining cointegration are listed in the Tab. 4. Based on tests, we will not reject the null hypothesis of non-cointegration of the time sequence, as the calculated all p-values are higher than the specifi ed signifi cance level of 0.05. For this reason, in both cases, the test series is non-cointegrated. In the following test, causality is performed using a VAR model (see formula 3 and 4). For the Granger tests of time causality, a null hypothesis is set, that the development of bank loans does not affect the economic cycle and therefore has no impact on the forecasts of the GDP. Tests are performed for eight quarterly delays, where any causality can be assumed. Test results for inter-annual changes of quarterly values are given in the following table (Tab. 5). Causal relations of the development of loans to GDP are shown at a signifi cance level of 0.05 at two, four and eight quarterly delays. Based on the calculated p-value, it is possible in three cases, to decide to reject the null hypothesis at a signifi cance level of 0.05. From the economic point of view it means that the development of bank loans in the form of relative annual changes (for total loans) Model p-value Evaluated result of ADF test Test with constant ∆Y0.6492 Time sequence is non-stationary ∆Uc 0.7910 Time sequence is non-stationary ∆Up 0.5196 Time sequence is non-stationary ∆Ud 0.8450 Time sequence is non-stationary ∆Uh 0.4006 Time sequence is non-stationary ∆Us 0.8225 Time sequence is non-stationary Test with constant First difference ∆Y0.01756 Time sequence is stationary First difference ∆Uc 0.00427 Time sequence is stationary First difference ∆Up 0.00007 Time sequence is stationary First difference ∆Ud 0.00010 Time sequence is stationary First difference ∆Uh 0.00068 Time sequence is stationary First difference ∆Us 0.00006 Time sequence is stationary Source: own Model Variable AIC p-value H0: 2 delays with constant ∆Y 239.4507 0.0693 Not rejected 1 delay with constant and trend ∆Up 232.1227 0.4894 Not rejected 1 delay with constant and trend ∆Ud 231.2742 0.2772 Not rejected 1 delay with constant and trend ∆Uh 228.4799 0.1836 Not rejected 1 delay with constant ∆Us 237.2326 0.1728 Not rejected Source: own Tab. 3: Results of ADF stationarity test for total loans and GDP Tab. 4: Results of the E-G cointegration test in annual changes EM_4_2017.indd 41EM_4_2017.indd 41 13.12.2017 12:53:3013.12.2017 12:53:30 42 2017, XX, 4 Ekonomie causally act within the meaning of Granger causality on the development of GDP, with a certain time lag. The following table (Tab. 6) shows the results of the Granger causality in quarterly inter-annual changes between developments in loans to non-fi nancial businesses and GDP growth. Causal relations of the development of loans to non-fi nancial businesses to GDP are shown at a signifi cance level of 0.05 at two, four and eight quarterly delays. Based on the calculated p-value, it is possible in three cases, to decide to reject the null hypothesis at a signifi cance level of 0.05. From the economic point of view it means that the development of bank loans in the form of relative annual changes (for loans to non-fi nancial businesses) causally act within the meaning of Granger causality on the development of GDP, with a certain time lag. The Tab. 7 shows the results of the Granger causality in quarterly inter-annual changes between developments in loans to households and GDP growth. Causal relations of the development of loans to households to GDP are shown at a signifi cance level of 0.05 at one, two, four and six quarterly delays. Based on the calculated p-value, it is possible in four cases, to decide to reject the null hypothesis at a signifi cance level of 0.05. From the economic point of view it means that the development of bank loans in the form of relative annual changes (for loans to households) causally act within the meaning of Granger causality on the development of GDP, with a certain time lag. The Tab. 8 shows the results of the Granger causality in quarterly inter-annual changes between developments in mortgage loans and GDP growth. Null hypothesis Delay p-value H0: ∆Uc does not causally act on ∆Y1 0.1843 Not rejected ∆Uc does not causally act on ∆Y2 0.0070 Rejected ∆Uc does not causally act on ∆Y3 0.2816 Not rejected ∆Uc does not causally act on ∆Y4 0.0086 Rejected ∆Uc does not causally act on ∆Y5 0.3925 Not rejected ∆Uc does not causally act on ∆Y6 0.4210 Not rejected ∆Uc does not causally act on ∆Y7 0.1402 Not rejected ∆Uc does not causally act on ∆Y8 0.0176 Rejected Source: own Null hypothesis Delay p-value H0: ∆Up does not causally act on ∆Y1 0.5110 Not rejected ∆Up does not causally act on ∆Y2 0.0039 Rejected ∆Up does not causally act on ∆Y3 0.4321 Not rejected ∆Up does not causally act on ∆Y4 0.0172 Rejected ∆Up does not causally act on ∆Y5 0.6775 Not rejected ∆Up does not causally act on ∆Y6 0.8680 Not rejected ∆Up does not causally act on ∆Y7 0.0769 Not rejected ∆Up does not causally act on ∆Y8 0.0458 Rejected Source: own Tab. 5: Results of Granger causality in annual changes – total loans Tab. 6: Results of Granger causality in annual changes – loans to non-fi nancial businesses EM_4_2017.indd 42EM_4_2017.indd 42 13.12.2017 12:53:3013.12.2017 12:53:30