Interest Rates, Credit and Macroeconomy – a Small Structural Model for the Lithuanian Case
Tomas Ramanauskas* Bank of Lithuania and Vilnius Gediminas Technical University
Abstract In this paper we build a simple structural model aimed at analysing how changes in the interest rate environment, associated with the ECB monetary policy decisions and financial liberalisation in Lithuania, affected developments in the domestic financial market and the rest of the economy. The paper assumes an explicit role for credit, which enables us to estimate the impact of credit constraints, on top of the conventional interest rate channel. The empirical investigation generally confirms that the eased monetary policy and financial liberalisation earlier this decade has contributed significantly to the economic boom in Lithuania, and the effect is most pronounced on nominal demand-side variables.
Key words: monetary policy transmission mechanism, interest rate channel, credit channel, financial accelerator, credit cycle, credit growth.
This version: November 2006 *
Opinions and views expressed in this paper are those of the author and do not necessarily represent the official position of the Bank of Lithuania.
1. Introduction
In perfectly functioning financial markets there is no clear role for credit – banks always supply the amount of credit required by firms and individuals, they charge interest that perfectly complies with riskiness of borrowers, and raising funds directly in financial markets is equivalent to bank credit. This seemingly allows one to effectively leave credit variables out of macroeconomic analysis. In contrast, a growing body of theoretical and empirical literature rests on assumptions of imperfect capital markets and informational asymmetries between lenders and borrowers, which not only resolves the “puzzle” of existence of financial intermediaries but also provides evidence that financial market imperfections might have nontrivial effects on various economic processes. The state of a financial system affects efficiency of capital allocation and is one of the factors determining whether economic agents face binding credit constraints; credit also plays an important role in amplifying economic cycles and is seen by some as a distinct monetary policy transmission channel. Since credit may potentially raise long-term economic growth via more efficient capital accumulation but on the other hand excessive credit growth may lead to periods of economic booms and recessions (and even crises), including credit aggregates in macroeconomic modelling is often desirable despite the likely endogeneity of credit processes. The need to explicitly address credit-related issues is especially evident in young market economies undergoing rapid financial liberalisation and characterised by low initial levels of financial intermediation. For instance, most Central and Eastern European (CEE) countries have recently experienced episodes of rapid credit growth and in the majority of those cases policy makers have tended to deal directly with excessive bank lending in order to avoid snowball effects for the economy in the form of inflation outbursts, widening of current account deficits, asset price booms and busts, over-investment, etc. Rather limited success in controlling credit growth by prudential measures supports the view of the endogeneity of credit but then again, this does not imply that credit developments cannot add interesting “nonlinearities” to economic developments. In this paper we focus on the ongoing episode of rapid credit growth in Lithuania, which is in many respects similar to other CEE countries but stands out (together with the other two Baltic states) in terms of a very rapid pace of credit expansion, very strong economic upturn, investment, consumption and asset price booms, relatively large external imbalances and sizeable pressures in the labour market. Against this backdrop, it seems very expedient to inquire into a plausible economic mechanism linking these buoyant 2
developments, which could shed some light on the drivers behind brisk economic growth, its sustainability and possibility of sudden unwinding of imbalances. Standard explanations of the Lithuanian growth phenomenon include capital accumulation, adoption of new technologies, labour productivity growth, competitive advantages related to relatively cheap labour force; some authors also emphasize the role of strong exports. These explanations seem valid but above all that it should be noted that in early 2000s Lithuania experienced quite drastic changes in the interest rate environment and bank lending behaviour. Huge, credit-driven domestic demand is clearly among the crucial factors driving the recent economic developments and shaping future economic outlook. Sustainability of domestic demand and ability of domestic producers to satisfy it in the medium- and long-term, in our view, are the most important questions here. Hence the main objective of the current paper is an empirical investigation of ways in which interest rates, credit aggregates and other financial variables reflecting changes in monetary and financial environment interact with various real and nominal economic variables, such as economic growth, consumption, investment, domestic prices, external sector developments, etc. For this purpose we build a small structural macroeconometric model which incorporates a stylised credit channel. Its main feature is that modelling aggregate investment we try to capture explicit dependence of investment on availability of external (and internal) financing and on incentives to invest, as measured by the external financing premium. We endogenise these variables in a parsimonious way, with interest rates left as the main exogenous variable. The model thus can also be used to study the plausible process of the monetary policy transmission in Lithuania without strictly disentangling monetary policy channels – strict separation of monetary policy transmission channels is hardly possible due to well known identification problems. We conduct a scenario analysis to investigate the effects on the model economy of a permanent marginal change in interest rates, as well as partial and total credit constraints. The model broadly supports the view that a marked change in the interest rate environment and financial liberalisation has significantly contributed to the recent economic upturn and widening of some imbalances related to excessive domestic demand. Model results also suggest that interest rate increases and/or a reversal of (unsustainable) strong net financial inflows from the banking sector to the nonfinancial sector could result in a cyclical downturn. The remainder of this paper is structured as follows. In section 2 we discuss the importance of credit in the macroeconomic and monetary policy analysis. We present the conceptual framework of the model in section 3. Section 4 contains model equations. 3
Scenario analysis and discussion of model results are presented in section 5. Section 6 concludes.
2. Importance of credit in macroeconomic and monetary policy analysis
Under the classical view of the monetary transmission mechanism, changes in interest rates induce – through changes in relative prices and alternative costs – incentives to change investment, production and consumption behaviour (see e.g. Bean et al., 2002). For instance, a decline in interest rates reduces the economic cost of capital, lowers the required minimum return on investment and the user cost of holding inventories, thereby stimulating investment and production. It also raises the price of future consumption relative to the price of current consumption and has a positive effect on the value of discounted lifetime income providing incentives for households to spend more today. In the case of an adverse change in monetary conditions, the disincentives to invest, produce and consume may be enough to bring down the overall level of economic activity but positive incentives owing to interest rate cuts may not suffice to significantly stimulate economic activity if access to external financing is impaired. Thus the importance of credit, as a major source of external financing, becomes apparent. If financial markets are imperfect – which they are – households and firms may face binding credit constraints and this already implies that the economy may react to a monetary shock differently than in the absence of those constraints. Also, the severity of borrowers’ credit constraints and banks’ ability to extend credit may change as a result of a change in the interest rate environment. Most of the literature on the credit channel concentrates on the latter two aspects (for general discussion see, e.g., Bernanke and Gertler, 1995, or Brunner and Meltzer, 1988). The amplification of the business cycle and monetary policy impulses through credit channel is often called the financial accelerator effect. Dependence of external financing premium (and possibility to acquire credit) on the financial position of the borrower is referred to in the literature as the balance-sheet (sub-) channel. Due to well known asymmetric information problems borrowers’ ability to get loans is restricted by their estimated net worth, which is often simply proxied by a collateral, such as real estate or government securities. Interest rate changes affect the net worth, and thereby credit constraints, of a firm or an individual by directly affecting interest rate expenses and by altering the value of some assets on their balance sheets (real estate, equity, government
4
securities, etc.). For a more thorough discussion of the balance-sheet channel see e.g. Gertler and Gilchrist (1994) or Bernanke and Gertler (1995). The other sub-channel of the credit channel, namely, the bank-lending channel is less intuitive and more controversial. The bank-lending channel is the mechanism which may force banks to contract supply of loans in response to raised monetary policy base rates. One plausible explanation of this is that an important resource of bank funds – sight deposits and other most liquid liabilities – might contract due to higher alternative costs and lower demand for loans (slower money creation process). Another explanation relates to bank solvency constraints – an increase in interest rates may lower banks’ capital adequacy ratios, through a negative impact on the price of securities held by them, and impose binding constraints on loanable funds (Pedersen, 2003). For traditional, and probably somewhat obsolete explanations of bank liquidity drain and loan supply effects related to interest rate rises, see e.g. Bernanke and Gertler (1995), or Romer and Romer (1990) for evidence against the traditional view of the lending channel. It should be noted, however, that rises in interest rates provide incentives for the nonfinancial sector to save more and this should hamper functioning of the bank-lending channel. The financial accelerator does help to fill gaps in understanding the monetary policy transmission mechanism and real business cycles of mature economies. Arguably it is even more relevant for emerging economies experiencing financial liberalisation and opening up of financial markets. From the perspective of economic growth theory, emerging economies are characterised by relatively scarce capital, so removal of credit constraints potentially opens many productive investment opportunities and may put the economy on a higher and steeper economic growth path. Also, those economies might have high unemployment rates, low income level and suppressed domestic demand – in the sense some poverty trap – which could be effectively eliminated by stimulating demand with the help of bank credit. Another explanation for relative importance of credit channel in emerging economies is that in underdeveloped financial markets bank credit can be by far the most important source of external financing. Finally but probably most importantly, financial liberalisation and effective elimination of binding credit constrains potentially has a huge impact on real estate prices. A mere decline in interest rates raises the fundamental (though not necessarily market) price of housing property relative to other goods because housing usually is the most expensive household’s purchase and its acquisition is most dependent on borrowing conditions. Residential property markets are also characterised by a relatively long and hazardous price discovery process. Prices do not usually immediately jump in response to, 5
say, interest rate falls, as there is no universal asset pricing model, which economic agents could easily refer to. Rather, more favourable financing conditions change the supply-demand balance in the housing market, by inducing excessive demand, which in turn triggers persistent movement in supply. So in emerging economies, improved borrowing conditions combined with income growth prospects can greatly boost demand for housing and trigger very strong increases in the real activity of construction-related economic sectors and positive spillovers to other sectors via increased domestic demand. If residential property supply is sluggish and households are financially naïve, the situation may easily turn into the asset price bubble, credit boom and property-oriented investment boom. This may end up in overheating economy or even serious economic and financial crises due to overborrowing and overinvestment (see McKinnon and Pill, 1997 for a theoretical discussion). Having reviewed quite some reasons why credit developments are very tightly bound to economic developments in emerging economies, it is natural to look at empirical evidence on this relationship. In general, however, there is little consensus on whether financial liberalisation, capital account opening and eased credit constraints foster economic growth (for the overview of empirical literature and discussion see Fratzscher and Bussiere, 2004). This could be at least partly linked to the abovementioned overborrowing syndrome. Fratzscher and Bussiere check a logical hypothesis that there is a trade-off over time between financial openness (liberalisation) and growth. They find evidence that following financial liberalisation, countries that experience large capital inflows, current account deficits and investment boom do grow faster initially, the effect vanishes, however, in the medium-term (in some five years). In other words, financial liberalisation and removal of strict credit constraints may lead to many sorts of booms and they tend to go together. Detken and Smets (2004) study properties of asset price booms and find that they are typically associated with a substantial positive output gap (i.e. economic boom), more intensive housing investment, easy monetary conditions, high money and credit growth rates. Some of such booms are followed by busts but they are difficult to discern ex ante – just those so called “high-cost” booms (i.e. followed by strong recessions) are characterised by big booms in the real estate sector and they last longer. Once credit growth has captured close attention of academics and central bankers, a strain of empirical literature aimed at explaining credit growth has emerged. It is mainly simple one-equation models in which credit growth is explained by some real sector and financial variables. For instance, Calza et al. (2003) link real loans in the euro area to real GDP growth, nominal lending rate and inflation rate. Cottarelli et al. (2003) try to estimate 6
equilibrium levels of credit to GDP ratios for non-transition economies by regressing them on real GDP, inflation, liberalisation index, measure of accounting standards, etc., and project those for European transition economies. Brzoza-Brzezina (2005) examines dynamics of real loans in a group of three new EU member countries and its dependence on real GDP and real interest rates. Backe et al. (2006) examine credit growth determinants in CEE countries and try many different specifications of potential sets of determinants, which include real per capita GDP, bank credit to the government sector, short and long nominal lending rates, inflation, housing prices, etc. The problem with these models is that, as was argued above, in these regressions many explanatory variables, most notably, GDP and asset prices, are not truly exogenous – credit developments affect them through the financial accelerator. In our view, it is important to attempt to build more elaborate structural models to be able to study the interrelation between credit and the real economy. Though credit growth in Lithuania has been extremely buoyant since around 2002 and the financial accelerator seems to have been strongly pronounced, the ongoing processes have not yet been deeply analysed and well understood. For example, there is no consensus among economists on whether this episode of rapid credit growth1 is a natural consequence of financial convergence process or it is overborrowing implying the elevated level of risks of economic recession and financial instability. Also, there are no estimates in the literature of the possible financial accelerator effect on real activity. A noteworthy reference for seeing individual country experiences with rapid credit growth (including Lithuania) in the larger, namely, CEE context is Enoch and Otker-Robe (forthcoming). Structure and evolution of financial systems in the Nordic-Baltic region is analysed in the IMF’s forthcoming financial system assessment programme of the Nordic and Baltic states. Ramanauskas (2005, 2006, forthcoming) examines drivers behind rapid credit growth in Lithuania and discusses risks to macroeconomic and financial stability associated with the credit boom. A paper by Kuodis and Vetlov (2002) is the first documented effort to describe and model the monetary policy transmission mechanism in Lithuania. They analyse, however, the pre-boom period characterised by severe liquidity constraints, and naturally the credit channel is essentially out of scope of their analysis. Kohler et al. (2006) construct a panel regression for the three Baltic states to test for the existence of the bank-lending channel. They regress nominal loan growth rates on nominal GDP growth rates, monetary policy indicators (money
1
Bank loan portfolio to the private nonfinancial sector grew from 13.4 percent of GDP at end-2001 to the estimated 44.4 percent of GDP at the end of the third quarter of 2006. Bank loan portfolio growth grew by 40 % on average during this period.
7
market rates) and examine whether size, liquidity and capitalisation of banks may be responsible for a different impact of a monetary shock on credit growth. Their results confirm that loan growth rates are influenced by the European monetary policy indicator (as opposed to domestic market rates). They also find that less liquid banks react more strongly to a change in monetary policy indicators, and the qualitatively similar finding is for small banks. These two effects are in line with the bank-lending channel. In contrast, banks with weaker capital adequacy react less strongly to a change in monetary conditions, which does not support the bank-lending channel. These findings have to be interpreted with caution, however, because of the identification problem: Kohler et al. analyse bank loan portfolio dynamics rather than bank loan supply. Due to strong ties of many Baltic banks with foreign resource-affluent parent banks, loan supply – at least in certain periods – was inelastic and credit developments arguably were mostly demand-driven.
3. Conceptual framework of the model
All single-equation models aimed at analysing relationship between credit growth and real economic activity suffer from several serious problems. First and most severe is endogeneity – the financial accelerator effect implies that credit fosters GDP, asset price growth etc., and these economic developments in turn stimulate credit. Secondly, credit aggregates, such as the credit to GDP ratio, arguably are too broad: one has to discern credit to households from credit to the corporate sector or, say, credit financed with domestic saving from that financed with foreign funds because implications for real economic processes may be very different. Then, credit conditions affect both demand- and supply-side of the economy, which cannot be captured in single-equation models with only GDP included. Finally, those models do not provide economic rationale behind relationship between credit and other economic variables. One of distinctive features of our model is the credit aggregate that we choose to analyse. A very simple example may illustrate why working with a stock or flow of bank loans (to the private sector) may be inappropriate. If credit is fully financed with domestic time deposits then an increase in bank credit would not necessarily be associated with a rise in domestic demand because for some people and firms to be able to borrow more others would have to save more. In contrast, if banks can raise funds abroad or can lend part of funds, which are held in bank accounts in the form of sight deposits (that is, if banks can effectively 8
“create” money), there can be a strong association between an increase in credit and an increase in domestic demand. It is obvious that analysis that concentrates on just bank asset side and neglects the liability side misses important information and may hardly shed light on the relation between credit and the real side of the economy. Hence we choose to work with net credit flows from the banking sector to the nonfinancial private sector. Net credit is a very intuitive notion and it simply reflects the net amount of funds transferred from the banking sector to the nonfinancial private sector in a given period. These funds are ready and meant to be invested or spent on consumption, thus it is very natural to examine empirical association between this kind of credit variables and, say, aggregate investment or consumption. We distinguish between credit to firms and credit to households because both their drivers and implications for other economic processes are different. So to get net credit to households, from (a change in) bank loans to households we deduct (a change in) household time deposits and net interest paid to banks. The same logic holds for firms. A noteworthy facet is that only time deposits are deducted whereas a change in sight deposits does not affect net credit flows. From a household’s or firm’s position, a time deposit is a “real” financial transfer to the banking sector and it means deferred consumption or investment whereas a sight deposit is just a convenient form of holding liquid funds, it is not really a saving and may be consumed right away. Consequently an increase in time deposits reduces net financial flows from banks to the private sector but an increase in sight deposits does not. The very nature of banking intermediation implies that the net credit variable should drift to some negative but moderate value in the long run – banks simply charge higher interest rates on loans than they pay for raising financial resources. In contrast, during financial liberalisation or credit boom periods, net financial flows from banks to the private sector can be positive and large. That is exactly what we observe in the Lithuanian data that we have (see Figure 1). Moreover, at this stage there is not even an evident tendency for the net credit flows to drift back to zero or some negative value. In other words, we can suspect that the economy is being stimulated strongly by willingness and ability to borrow, and we can estimate the stimulating impact but it is extremely difficult to tell from the limited amount of data when and how fast this impact is going to subside. Credit conditions affect both demand- and supply-side of the economy. As was mentioned above, demand for some consumer goods, such as housing, cars and other durables, may be very sensitive to financing conditions and credit constraints. The same can be said about demand for investment goods, such as production equipment. Hence there obviously exists a link between credit and aggregate demand. On the other hand, investment 9
in production equipment and other business facilities raises the overall level of productive capacity of the economy thereby shifting the aggregate supply curve. The reaction of supply to the credit-fuelled demand shocks determines whether the economy experiences economic upsurge and whether it is sustainable or leads overheating pressures, rising prices, and deteriorating external accounts. An analysis of these issues necessitates a clear distinction between demand- and supply-side developments, which requires dealing with the identification problem. In our model we deal with the abovementioned identification issues in a simple way. The model’s key relationship is the identity defining the GDP deflator, which simply is nominal GDP divided by real GDP. By the national accounts’ identity, nominal GDP comprises domestic demand and net exports. Hence we model explicitly the more important elements of domestic demand, that is, aggregate investment and private consumption, and add net credit variables to the list of conventional explanatory variables. The supply side of the economy is essentially reflected by real GDP (especially bearing in mind that GDP comprises not only final goods and services sold but also accumulation of inventories). Real GDP dynamics is modelled by employing the usual unrestricted Cobb-Douglas type production function, and the supply side is affected by credit conditions and credit flows indirectly – via accumulation of capital (determined on the demand side in the model). Domestic price level and to some extent net exports adjust to equate demand and supply and close the model. Hence the very basic framework of the model may be illustrated with the help of the following equation: (1)
DEFL =
CONS ( NCH ,⋅) + I ( NCF ,⋅) + G (⋅) + NX (⋅) , YR ( I ,⋅)
where DEFL, CONS, I, G, NX, YR denote GDP deflator, private consumption, aggregate investment, government consumption, net exports and real GDP, respectively; NCH and NCF stand for net credit flows to households and firms. It should be noted, however, that the NCH variable turned out insignificant in the consumption equation, thus effectively the only place where credit enters the model is the investment equation but then there are a lot of potential second-round effects – credit-related higher level of income fosters private and government consumption, higher profitability leads to stronger investment, demand pressures also temporarily increase tightness of the labour market and negatively affect net exports. Finally, credit is endogenised and depends on the level of income and external financing conditions, allowing for the possibility of formation of a credit cycle.
10
4. Data and model equations
Any empirical analysis aimed at establishing the role of credit in determining economic developments in Lithuania is severely aggravated by the usual problem of short time series, incomplete financial and macroeconomic data, methodological changes in published statistics and structural changes in the economy as a whole. Therefore the model should primarily be seen as a plausible (and broadly supported by data) mechanism of the business and credit cycle formation or as a means to think about the credit boom episode in a more or less structural way, rather than a tool for robust quantitative predictions and policy prescriptions. Due to the very limited data set we could not deal with simultaneity issues or rely on full-information estimation methods, hence all model equations are estimated by using simple OLS methods. Most regressions take the error-correction form, which proved convenient bearing in mind potentially temporary nature of monetary and credit stimulation and differences in short- and longer-term relationship between analysed variables. Most data series are integrated of order 1 but standard stationarity tests lack power due to a small number of observations and are not reported. The sample period spans from Q3 1999 to Q1 2006 and consists of 27 quarterly observations. All variables are seasonally adjusted by using Census X11 procedure, except some financial variables, namely, net credit flows, loan and deposit rates and the nominal exchange rate. Variables expressed in monetary terms are expressed in millions litas. The difference operator ∆ denotes the difference between two subsequent quarters. Demand side
We start modelling the demand side with the consumption equation. It is basically a very simple Keynesian type consumption function: nominal consumption is a function of nominal income level proxied by nominal GDP. The constant term in the linear equation is statistically insignificant, and income elasticity is large and close to unity in the longer-term, broadly in line with Friedman’s permanent income hypothesis (permanent changes in the income level lead to similar changes in consumption). To distinguish between effects of real, productivity-based changes in the level of income and credit-fuelled demand for consumption, nominal GDP is expressed in terms of real GDP and GDP deflator. The resulting errorcorrection regressions are reported below (t statistics are shown in brackets): 11
(2)
log(CONS ) = 0.954 ⋅ log(YR ) + 1.190⋅ log( DEFL) + u1 , ( 2560.30 )
(12.38 )
R 2 = 0.99, DW = 1.73, (3)
∆ log(CONS ) = 0.892⋅ ∆ log(YR ) + 0.728⋅ ∆ log( DEFL) − 0.755⋅ u1,t −1 , ( 6.63)
( 3.85)
( −3.88 )
R 2 = 0.41, DW = 1.73, Consumption dynamics tracks quite closely income growth, and neither the standard Euler equation, nor credit-augmented variants of consumption function seem to fit empirical data. Lack of direct relationship between net credit flows to households and consumption can at least partly be explained by the fact that during the sample period the bulk of bank loans to households were housing loans, and housing acquisition is shown as investment rather than consumption in national accounts. Since nominal government consumption has been a roughly constant fraction of GDP during the period of investigation, it is simply modelled as a linear function of nominal GDP: (4)
G = 1392.087+ 0.091⋅ Y N , (14.24 )
(13.04 )
R 2 = 0.86, DW = 1.89.
Nominal investment dynamics is explained in the model by availability of external and internal financing (as measured by net credit flows and corporate earnings, respectively) and the external financing premium (EFP): (5)
I = 2431.748+ 0.423⋅ NCF + 0.638⋅ PROFITS − 33.657⋅ EFP + u2 , (12.10 )
( 3.24 )
( 5.93)
( −1.75 )
R 2 = 0.91, DW = 1.64, (6)
∆I = 0.328⋅ ∆NCF + 0.255⋅ ∆PROFITS − 0.698⋅ u2,t −1 , ( 3.74 )
(1.88 )
( −3.43)
R 2 = 0.37, DW = 1.75. One can see that less than a half of net credit flows to firms directly turn to investment in the longer-run. The figure is reasonable because part of credit may be used for business purposes other than investment; also, aggregate investment includes government investment (it is not possible to discern it from the overall figure), which is not directly related to credit or corporate earnings dynamics. Corporate profits in the investment function play a dual role: they are a source of internal financing but in addition to this they reflect returns on earlier investment and carry information on general investment outlook and investment perspectives, especially if firms are partly backward-looking. To capture explicitly information on investment environment, the investment equation includes the variable reflecting external financing premium, which is significant in the cointegration relationship but insignificant in the short-run equation. The external financing premium is defined as the difference between
12
the cost of external financing, i.e. average interest rates on loans for businesses, and the return on capital. Naturally, the larger this difference, the less incentives to invest. The return on capital, i.e. the appropriately scaled profit to capital ratio, is fully endogenised in the model, allowing for potential amplification of its effect on investment. This variable can also be seen as some proxy for equity prices bearing in mind that current returns are most relevant for asset pricing. Hence the EFP variable at least to some extent reflects Tobin’s q effect on investment. Net exports dynamics is linked to domestic demand and the nominal effective exchange rate (NER): (7)
NX = 1.148⋅ NER − 0.147⋅ (YN ,t −1 − NX t −1 ) + u3 , ( 3.06 )
( −5.57 )
R 2 = 0.54, DW = 1.60, (8)
∆NX = 2.972⋅ ∆NER − 0.743⋅ u3 , (1.85 )
( −3.98 )
R 2 = 0.46, DW = 1.74. The constant term in the cointegrating equation was insignificant, and in the short-term equation domestic demand is insignificant. The coefficient signs are as expected. An increase in domestic demand, ceteris paribus, should lower net exports due to stronger imports. An increase in the effective nominal exchange rate leads to improved terms of trade and in the short-term can positively affect net exports. We did not find any conclusive evidence on significance of real effective exchange rates, that is the theoretical price competitiveness and net exports relationship was not generally confirmed by the data, possibly due to the short sample period or due to structural changes in the trade structure. Note that domestic demand entering equation (7) is lagged one quarter. Though the contemporaneous relationship is also significant, the domestic demand variable is proxied by its lagged values in order to avoid explosive and multiple model solution paths (which can be the case if both domestic prices and net exports are free to adjust).
Supply side
To model the supply side we employ the usual Cobb-Douglas type aggregate production function. We let the data speak for themselves and do not restrict parameters to the case of constant returns to scale. One reason is that in the error-correction setting such parameter restrictions would lose economic meaning as residuals from the cointegrating
13
relationship enter the short-term equation. But more importantly, in our view the level of data aggregation is too high and data reliability (especially on capital dynamics) is too low to impose any theoretical restrictions, or, conversely, draw any inferences about income distribution from obtained estimates from the production equation. Moreover, in the light of economic growth theories, economic growth factors might be different in the short- and longterm – e.g. raw capital accumulation in the development point which is far from the steady state can be much more beneficial for the economy than in the vicinity of the steady state. We add a time trend to function as the Solow residual and to account for growth factors, such as the technological progress, which are not particularly relevant for the current analysis of short- and medium-term fluctuations. The estimation results are as follows: (9)
log(YR ) = 0.092⋅ log(TREND) + 0.374⋅ log( K ) + 0.729⋅ log( L) + u4 , ( 7.32 )
( 5.79 )
( 8.00 )
R 2 = 0.96, DW = 0.69, (10)
∆ log(YR ) = 0.037⋅ ∆ log(TREND) + 0.433⋅ ∆ log( K ) + 0.632⋅ ∆ log( L) − 0.413⋅ u4,t −1 , (1.70 )
( 3.91)
( 2.44 )
( −3.16 )
R 2 = 0.20, DW = 2.23. Interestingly, the sum of coefficients on capital and labour does turn out to be close to unity. Furthermore, these coefficients are very similar to those chosen by Vetlov (2004) based on expert judgement. It should be noted, however, that in the production function equation he employed capital obtained by using the perpetual investment method and data on investment dynamics, whereas we proxied capital with fixed tangible assets obtained from the corporate balance-sheet country-level statistics. Modelling of employment in this model is a bit simplistic because any serious theoretically founded modelling would require building a multi-country framework as the labour force migration has been among the most important factors determining developments in the Lithuanian labour market. Hence we do not attempt to explain the longer-term dynamics and simply fit a polynomial trend, and in the short-term we regress employed labour force on the autoregressive term and also link it to demand pressures proxied by nominal GDP (in the spirit of Okun’s law): (11)
L = 1481.318− 13.694⋅ TREND + 0.517⋅ TREND 2 + u5 , ( 93.11)
( −5.60 )
( 6.54 )
R 2 = 0.64, DW = 0.55, (12)
∆L = − 7.697+ 0.284⋅ ∆Lt −1 + 0.026⋅ ∆YN − 0.316⋅ u5,t −1 , ( −1.49 )
(1.56 )
( 2.09 )
( −2.13)
R 2 = 0.32, DW = 2.00.
14
Our results confirm that employment exhibits some inertia and covaries positively with the overall economic activity.
Endogenising net credit flows
As was argued above, net credit flows are expected to be mean-reverting processes. The sample period, however, is not sufficiently long to confirm this assertion empirically. This does not have serious negative implications for the purposes of our analysis but if some similar model was intended for forecasting, one should take into account a rising debtservicing cost and the mean-reversion property of credit flows, or, in other words, the fact of non-sustainability of strong positive credit lows into the private sector. The analysed episode of rapid credit growth in Lithuania is characterised by the housing boom and household borrowing. The corporate sector, especially certain sectors such as property developers, reacts to credit-fuelled demand and may also resort to credit resources. For this reason net credit flows to households enter as an explanatory variable in the equation explaining dynamics of credit flows to the corporate sector. Credit to the corporate sector is also found to depend on the degree of financial liberalisation and market tightness as measured by average bank margins, i.e. the difference between average firm loan and firm deposit rates: (13)
NCF = 701.646+ 0.839⋅ NCH − 102.472⋅ ( FLRATES − FDRATES ) + u6 , ( 2.32 )
( 2.13)
( −1.80 )
R 2 = 0.69, DW = 1.50, (14)
∆NCF = 0.640⋅ ∆NCH − 0.872⋅ u6,t −1 , (1.81)
( −4.14 )
R 2 = 0.46, DW = 1.95. All coefficient signs are as expected. Credit conditions affect net credit flows in the longertem, while short-term fluctuations in bank margins do not seem to have a clear impact on credit to the corporate sector. Similarly, net credit to households depends on market conditions (the difference between average household loan and deposit rates). Households’ financial position proxied by lagged nominal GDP also enters the cointegrating relationship: (15)
NCH = − 106.453⋅ ( HLRATES − HDRATES ) + 0.043⋅ YN ,t −1 + u 7 , ( −9.75 )
(11.51)
R 2 = 0.80, DW = 1.79,
15
(16)
∆NCH = − 73.60⋅ ∆( HLRATES − HDRATES ) − 0.877⋅ u 7,t −1 , ( −1,58 )
( −4 , 35 )
R 2 = 0.41, DW = 1.99. In general credit flows seem to depend on market conditions and income dynamics, as is welldocumented elsewhere in the literature on credit growth.
Adding auxiliary regressions and closing the model
At this point there is still one crucially important link missing in the model. On the demand side, we analyse, inter alia, investment (taken from national accounts and expressed in nominal terms), whereas on the supply side, capital (from the corporate balance-sheet data) enters the production function. So obviously we need to link these two variables. The change in the capital stock is equal to investment minus depreciation. For simplicity and a better fit, we assume a constant absolute depreciation (a constant rate of depreciation also gives plausible estimates), so we get a very simple equation of motion of capital: (17)
∆K = −1524.123+ 0.845⋅ I , ( −1.74 )
( 2.88 )
R 2 = 0.23, DW = 2.04. Actually the fit is quite poor, which is natural, given the very different nature of data on capital and investment and the fact that investment data also comprises government investment whereas the capital variable reflects only the corporate sector data. Corporate earnings are among drivers determining investment dynamics in the model. It is also a very procyclical variable that plays an important role in forming business and credit cycles. So if it is left exogenous, some interesting dynamics is deprived from the model. We relate corporate profits (from the corporate income statemets’ data) to the model’s closest proxy for national disposable income, which is the difference between nominal GDP and government expenditure: (18)
PROFITS = − 1718.721+ 0.235⋅ (YN − G ) + u8 , ( −6.55 )
(10.15)
R 2 = 0.78, DW = 0.96, (19)
∆PROFITS = − 0.518⋅ u8,t −1 , ( −3.43)
R 2 = 0.29, DW = 2.10. Finally, the model is closed using the definition of the GDP deflator (see equation (1)). The GDP deflator is the balancing variable, which is derived with no direct linking to the
16
corresponding empirical data, thus it is very interesting to compare its dynamics to the actual dynamics. This can serve as some indication of model’s appropriateness. It should be noted beforehand that in the baseline scenario the system is driven by a very small number of exogenous variables, such as interest rates and the nominal effective exchange rate, so towards the end of the sample period the estimated baseline variable can follow a very different path from the actual one. Also, the GDP deflator is a relatively vague and unreliable statistic in our case, which makes its estimation a daunting task. Hence in Figure 2 we do observe some differences between the actual and estimated dynamics. The GDP deflator (and demand pressures) is overestimated in the initial stage of the credit boom, i.e. between 2002 and 2004, and underestimated afterwards. Interestingly the model predicted a demand-driven increase in domestic prices, which was followed by a quite similar actual price hike. So if it is just a question of timing, judging from the figure one could expect a significant slowdown in domestic price growth, related to more balanced nominal consumption and investment growth.
5. Scenario analysis and discussion of model results
Once the above equations are put together in a model, a scenario analysis can be conducted. We analyse four scenarios. Scenario 1 assumes a permanent favourable shift in external financing conditions whereby from Q1 2000 both household and business loan rates permanently drop by 1 p.p. relative to the actual path (deposit rates remain unchanged). Scenario 2 models implications of the assumption that firm and household loan and deposit rates are fixed at the Q1 2000 level and are constant throughout the analysis period. Scenario 3 assumes imposition of permanent credit constrains starting from Q1 2000, which means holding bank deposits and bank loans constant but allowing interest rates to follow the actual path (this implies a very moderate and gradual change in net credit flows due to easing interest rate servicing burden related to interest rate declines). Scenario 4 assumes “total” credit constrains and differs from scenario 3 in that interest rates are also left constant at prohibitively high levels that prevailed in Q1 2000. Financial liberalisation and associated declines in bank interest margins can have a quite sizeable impact on the economy in the medium-term. The impact of a 1 p.p. drop in bank interest margins, assumed in scenario 1, on other economic variables is reported in Table 1 and Figure 3. The first thing to note is that most effects do not vanish during the analysed 17
medium-term period: a tightening of bank margins (due to lower loan rates) has a strong and lasting stimulus on corporate profits, net credit flows, capital accumulation processes, somewhat more moderate on nominal consumption, nominal and real GDP, it strongly negatively affects net export developments. On the other hand, it leads to a very moderate and short-lived increase in employment in the model, and the impact on the GDP deflator, though quite pronounced, vanishes in some five years. Importantly, a 1 percentage point decline in bank interest margins in Q1 2000 contributed on average during the subsequent six-year period to stronger growth of real GDP by 0.3 p.p., nominal GDP – by 0.6 p.p., corporate earnings – by 1.9 p.p., capital stock – by 0.7 p.p., nominal consumption – 0.7 p.p., GDP deflator – by 0.4 p.p. (see Table 1). It took 2 to 9 quarters for the impact to fully materialise. For instance, net credit flows or employment variables were among the fastest to react to a favourable interest rate change, whereas it took more than two years before profits or net exports felt the strongest impact. In assessing the impact of the very significant change in the interest rate environment over the analysed period, one has to bear in mind that during this period bank margins gradually narrowed by some 5 p.p. (i.e. five times the simulated shock). Scenario 2 is likely to capture the full impact of the change in the interest rate environment in on the economy, thus it is probably the most important case to examine. The effects are sizeable for almost all analysed variables but generally more significant for nominal variables – see Table 1 and Figure 4. To be more specific, should the interest rates have stayed at the Q1 2000 levels, corporate earnings would have grown on average slower by 8.1 p.p., trade deficit – by 6.1 p.p., investment – by 4.4 p.p., consumption – by 2.7 p.p., nominal GDP – by 2.5 p.p., prices of domestic producers – by 1.7 p.p. Real GDP growth and employment would have been, respectively, 0.8 and 0.1 p.p. weaker. The negligible effect on the employment dynamics captured by the model is hardly surprising bearing in mind that employment growth was quite slow (as opposed to the decline in unemployment) and was to a large extent explained in the model by a time trend. Analysis of scenario 2 reveals that the largest part of extremely buoyant credit dynamics can be explained by a mere change in the interest rate environment (see dynamics of credit flows in Figure 4). Nonetheless, ability to borrow is a integral part of the monetary transmission mechanism and restrictions imposed on the ability to borrow alter the impact of the interest rate changes as is discussed below. To conduct scenarios 3 and 4, we exclude the net credit flows equations from the model (so it is a different, simpler model), assign an exogenous path for the net credit to the corporate sector in scenario 3 and, on top of that, assume constant interest rates in scenario 4. Scenario 3 allows to assess how financial liberalisation and a change in banks’ willingness to 18
finance the private sector affected macroeconomic developments. Figure 5 and summary indicators provided in Table 1 reveal that the impact on many variables is quite significant and again, it is most pronounced in the case of nominal demand-side variables. For instance, removal of heavy credit constraints contributed most sizeably to growth of corporate earnings (on average by 2.4 p.p. per year), trade deficit (1.7 p.p.), investment (1.5 p.p.), consumption (1.0 p.p.), nominal GDP (0.9 p.p.) and the GDP deflator (0.6 p.p.). On the other hand, the contribution to real GDP and employment growth was found (somewhat at odds with a priori expectations) to be relatively modest – only 0.2 and 0.1 p.p. per year, respectively. Analysis of scenario 4 is aimed at assessing how a change in the interest rate environment coupled with active bank lending affected the Lithuanian economy. The effect is qualitatively similar but, of course, considerably stronger compared to scenario 3. As can be seen from Figure 6 and Table 1, interest rate declines and easing credit constraints most significantly contributed to annual growth rates of corporate profits (by 6.0 p.p.), trade deficit (4.3 p.p.), investment (3.3 p.p.), consumption (2.1 p.p.), nominal GDP (1.9 p.p.), GDP deflator (1.3 p.p.), whereas impact on real GDP and employment average annual growth was moderate – 0.6 and 0.1 p.p., respectively. This scenario is principle very similar to scenario 2 and produces very similar, only weaker results. Comparing these results one has to take into account that scenarios 2 and 4 represent two different models, the latter being simpler due to removed equations, with some feedback model links broken. The above scenario analysis confirms that interest rate developments and the availability of credit clearly are among factors playing crucial role in shaping today’s economic situation in Lithuania. As was expected, there is a stronger link with the nominal demand side, whereas model results suggest that real economic growth is less influenced by financial market developments than we initially thought. Nevertheless, taking into account the many data and model limitations (e.g. the well-known difficulty to realistically model aggregate production using simple two-factor functions), we still tend to interpret this as the model deficiency rather than conclusive evidence that robust economic growth is relatively little affected by credit. In any case, model confirms that there are credit-related overheating pressures, which is very important to take into account in the context of active policy debate over whether the ongoing credit growth is excessive. Actually, no econometric model can directly and convincingly answer this question because there are not enough empirical data and no prior experience of a full credit cycle in the case of Lithuania. The current model only helps to assess the extent to which current economic developments are conditional on financial market developments (and this extent is relatively large, especially on the demand 19
side). Since the Lithuanian financial market is being directly affected by the ongoing ECB’s monetary policy tightening process and strong positive net credit flows (which in principle are unsustainable in the long-run for the reasons discussed above) may subside due to the ending real estate boom and other factors, policy makers have to take into account possible implications of this for the business cycle. The question the model cannot answer, however, is on the possible asymmetry of the discussed business cycle amplification mechanism – it might well be the case that further interest rate rises or, say, a sudden reversal of net credit flows might have a substantial impact on the real economic activity, whereas during the economic upturn this impact was found to be only moderate.
6. Conclusions
In this study we analyse how interest rates and credit variables interact with other important economic variables, such as real and nominal GDP, consumption, investment, inflation, net exports, domestic prices, etc. Since this interaction is undoubtedly complex, instead of developing a single-equation model as is now popular, we attempt to take a different route and present a simple structural model with an explicit role for credit. Due to severe data and model limitations, we see the current model primarily as a more or less structural way of thinking about the place of credit and other financial variables in the dynamically growing emerging economy, rather than a formal quantitative tool. Our approach is novel in several ways. First of all, we work with net credit flows to the private non-financial sector as opposed to very general credit variables like aggregate credit growth or bank loan portfolio to GDP ratio. Net financial transfers from the banking sector to the private sector, in our view, are very natural variables to link to aggregate investment and consumption, which is not necessarily so in the case of the standard credit variables. Another distinct feature of our model is that we explicitly introduce – somewhat loosely speaking – the supply- and demand-side of the economy in the model. The structure of the model is such that aggregate nominal investment equation includes among other variables net credit, which is a source of external financing. Thus stronger net credit would stimulate aggregate domestic demand and foster nominal GDP growth, as the first round effect. On top of that, via the standard aggregate production function investment is linked to the supply side, i.e. production of real output. Real GDP and domestic demand is balanced with the help of adjusting domestic prices (GDP deflator) and to some extent net exports 20
dynamics. Net credit flows are basically explained by external financing conditions and the financial position of borrowers (proxied by lagged GDP). A very small number of fully exogenous variables in the model enables it to capture second-round effects related to changes in the interest rate environment or borrowing conditions. The presence of both credit and interest rate variables in the model makes it possible to model two arguably most important monetary policy transmission channels in Lithuania, namely, the interest rate channel and the credit channel. We conduct analysis of four scenarios: (i) a marginal permanent change in external financing conditions, (ii) a case of constant interest rates fixed at the prohibitively high level of early 2000s, (iii) a case of borrowing constraints and interest rates following the original path and (iv) a case of “total” credit constraints that are characterised by both fixed net credit and high interest rates. Comparison of model’s baseline economic developments with those of the scenario of high interest rates in principle gives the feel of relevance of the strong changes in the interest rate environment and financing conditions in the first half of 2000s for explaining buoyant economic developments over the period. The impact of changes in the interest rate environment and external financing conditions on the economy is generally strong and more pronounced on nominal variables. The model reveals that the largest part of extremely buoyant credit dynamics can be explained by a mere change in the interest rate environment. Nonetheless, ability to borrow is a integral part of the monetary transmission mechanism and restrictions imposed on the ability to borrow alter the impact of the interest rate changes. The impact of the interest rate declines and the credit boom on the real economic growth is somewhat lower than a priori expected but this probably could be linked to the deficiencies of standard modelling of the aggregate production function (e.g. inclusion of a time trend, etc.) and other model limitations. Another fact that makes us think that the effect is underestimated is the huge actual impact of these financial developments on asset prices and an associated rise in property development related sectors. Property prices could not be modelled explicitly in the model due to the lack of reliable data but most probably they would add interesting dynamics on the real side of the model economy. Notwithstanding the limitations, model results are important for several reasons. First of all, they give the feel of the economy’s reaction to strong monetary policy impulses, witnessed during the period of interest. The model also shows that credit developments are to a large extent driven by interest rate and income developments, which implies that dealing with credit growth without a direct control over the policy rate is a formidable task for policy makers but does not imply that credit growth is not excessive. On the contrary, the strong 21
relation between credit and nominal variables, as well as prices of domestic producers suggests that earlier changes in the interest rate environment together with strong credit growth create significant overheating pressures and the possibility of disorderly unwinding of resulting imbalances may not be excluded.
22
Table 1. Summary statistics for the scenario analysis Scenario 1
Scenario 2
consumption GDP deflator
Nominal investment Government expenditure Capital stock
Employed labour Net exports
Corporate earnings Nominal GDP Real GDP
Scenario 4
Average
Average
Average
Average
Average
Average
Average
Average
difference
growth
difference
growth
difference
growth
difference
growth
from
difference
from
difference
from
difference
from
difference
baseline
from
baseline
from
baseline
from
baseline
from
scenario*
baseline
scenario*
baseline
scenario*
baseline
scenario*
baseline
(%)
scenario**
(%)
scenario**
(%)
scenario**
(%)
scenario**
(p.p.) Nominal
Scenario 3
(p.p.)
(p.p.)
(p.p.)
4.1
0.7
-15.3
-2.7
-5.6
-1.0
-11.7
-2.1
2.1
0.4
-9.6
-1.7
-3.6
-0.6
-7.4
-1.3
6.4
1.0
-23.4
-4.4
-8.5
-1.5
-18.2
-3.3
1.8
0.3
-6.8
-1.2
-2.5
-0.4
-5.2
-0.9
4.5
0.7
-11.4
-2.0
-2.9
-0.5
-7.7
-1.3
0.0
0.0
-0.6
-0.1
-0.4
-0.1
-0.6
-0.1
-9.7
-1.6
31.2
6.1
9.9
1.7
23.0
4.3
11.7
1.9
-39.6
-8.1
-13.8
-2.4
-31.0
-6.0
3.7
0.6
-14.1
-2.5
-5.2
-0.9
-10.9
-1.9
1.7
0.3
-4.9
-0.8
-1.4
-0.2
-3.4
-0.6
* Average difference from the baseline scenario is obtained by time-aggregating analysed variables (e.g. consumption under scenario 1 and under the baseline scenario) over the period from Q1 2000 to Q1 2006 and calculating the percentage difference between the two aggregates. ** Average growth difference from the baseline scenario is obtained by distributing the average difference indicator evenly through the analysed period (6 years). E.g. for nominal consumption under scenario 1, 1.041^(1/6)-1=0.007.
23
Figure 1. Credit flows between banks and the nonfinancial private sector 10000
8000
1000
800
800
600
600 6000
400
400 200 200
4000
0
0 2000
-200
-200 0
-400 2000
2001
2002
2003
2004
2005
-400 2000
Loans to households Household time deposits
2001
2002
2003
2004
2005
2000
D(HLOANS)-D(HTDEPS) Net interest paid by households
20000
16000
2001
2002
2003
2004
2005
Net credit flow s to households
1600
1600
1200
1200
800
800
400
400
0
0
-400
-400
12000
8000
4000
-800
-800
0 2000
2001
2002
2003
2004
2000
2005
2001
2002
2003
2004
D(FLOANS)-D(FTDEPS) Net interest paid by firms
Loans to nonfinancial corporate sector Time deposits of nonfinancial fir ms
2005
2000
2001
2002
2003
2004
2005
Net credit flow s to firms
Figure 2. Actual and model GDP deflator dynamics 1.16 1.12 1.08 1.04 1.00 0.96 0.92 0.88 2000
2001
2002
2003
2004
2005
G DP deflator (actual) G DP deflator (model; baseline)
24
Figure 3. Impact of more favourable financing conditions on the model economy (scenario 1) CONSUMPTION
GDP DEFLATOR
INVESTMENT
11000 10000
1.15
4000
1.10
3600
1.05
3200
9000 1.00 6 5 4
8000
4
0.95
7000
3
0.90
3 2 1 0 2001
2002
2003
2004
4
1
2
2005
2400
6
2
0 2000
2800 8
2000
0 2000
2001
CONS (Scenario 1) CONS (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
2001
DEFL (Scenario 1) DEFL (Baseline) Percent Dev iation
CAPITAL
EMPLOYMENT
2003
2004
2005
NET CREDIT (HOUSEHOLDS)*
70000
8
2002
I (Scenario 1) I (Baseline) Percent Dev iation
60000
1480
600
1460
500
1440
400
1420 6
50000
4
40000
2
30000
0
.8
1400
.6
1380
.4
1360
300 200 100 0
.2
-100
.0
-200
-.2 2000
2001
2002
2003
2004
2005
-300 2000
2001
K (Scenario 1) K (Baseline) Percent Dev iation
2002
2003
2004
2005
NET EXPORTS**
600 400 200
2002
2003
2004
1400 -800 -1000
16 -1200
12 8
-1400
5
-400
-4
0 2000
2001
2002
2003
2004
2005
2000
16000
5
2200
12000
4
0.5
1
10000
G (Scenario 1) G (Baseline) Percent Dev iation
16000
2.4 2.0
14000 1.6 12000
1.2
10000 0.4 0.0
0 2005
18000
0.8 2
2004
2005
3
1.0
0.0
2004
18000
14000
1.5
2003
REAL GDP
NOMINAL GDP
2400
2002
PROFITS (Scenario 1) PROFITS (Baseline) Percent Dev iation
2800
2003
2001
NX (Scenario 1) NX (Baseline) Percent Dev iation
3000
2002
400
15 10
2005
600
20
0
GOV'T CONSUMPTION
2001
800 25
NCF (Scenario 1) NCF (Baseline) Percent Dev iation
2000
1200 1000
20
4
2.0
2005
1600
-200
2600
2004
CORPORATE EARNINGS
0
2.5
2003
-600
1000 800
2002
NCH (Scenario 1) NCH (Baseline)
1200
2001
2001
L (Scenario 1) L (Baseline) Percent Dev iation
NET CREDIT (FIRMS)*
2000
2000
2000
2001
2002
2003
2004
2005
2000
Y N (Scenario 1) Y N (Baseline) Percent Dev iation
2001
2002
2003
2004
2005
YR (Scenario 1) YR (Baseline) Percent Dev iation
* Percentage deviation for credit flows not reported since these variables change sign. ** Net exports are negative so percentage deviation should be interpreted in absolute value only.
25
Figure 4. Impact of interest rates left at prohibitively high levels (scenario 2) CONSUMPTION
GDP DEFLATOR
INVESTMENT
11000
1.2
10000
3600 3200
1.1
9000
4
8000
0
7000
-4
2800 1.0
5 0
10 0.9
2400
0 2000 -10
-5
0.8
-8
-20
-10 -12
-15 -20
-30
-16 2000
2001
2002
2003
2004
2005
-40 2000
2001
CONS (Scenario 2) CONS (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
2001
2002
DEFL (Scenario 2) DEFL (Baseline) Percent Dev iation
CAPITAL
2003
2004
2005
I (Scenario 2) I (Baseline) Percent Dev iation
EMPLOYMENT
NET CREDIT (HOUSEHOLDS)*
60000
1480
500
55000
400 1440
50000 45000 40000
5
300
0.4 1400
200
0.0
35000
0
30000
-5 -10
-0.4
0 -0.8
-15
100
1360 1320
-100
-1.2
-20 -25
-200
-1.6 2000
2001
2002
2003
2004
2005
-300 2000
2001
K (Scenario 2) K (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
2001
2002
L (Scenario 2) L (Baseline) Percent Dev iation
NET CREDIT (FIRMS)*
2003
2004
2005
NCH (Scenario 2) NCH (Baseline) Percent Dev iation
NET EXPORTS**
CORPORATE EARNINGS
1000
-400
1400
800
-600
1200 1000
-800
600 10 400 200 0 -200 -400 2000
2001
2002
2003
2004
0
10
600
-1200
0
400
-10
200
-10 -1400
-20
800
-1000
-20
-30
-30
-40
-40
-50
2005
-50 2000
2001
NCF(Scenario 2) NCF (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
NX (Scenario 2) NX (Baseline) Percent Dev iation
GOV'T CONSUMPTION
2001
2002
2003
2004
2005
PROFITS (Scenario 2) PROFITS (Baseline) Percent Dev iation
REAL GDP
NOMINAL GDP 2900
18000
18000
16000
16000
2800 2700 2600 2500
2
2400
0
2300
-2
14000 5
12000
12000
0
0
10000
-5
-4
14000 2
10000
-2 -4
-10
-6
-6
-8
-15
-8
-10
-20
-10
2000
2001
2002
2003
2004
G (Scenario 2) G (Baseline) Percent Dev iation
2005
2000
2001
2002
2003
2004
2005
2000
YN (Scenario 2) YN (Baseline) Percent Dev iation
2001
2002
2003
2004
2005
YR (Scenario 2) YR (Baseline) Percent Dev iation
* Percentage deviation for credit flows not reported since these variables change sign ** Net exports are negative so percentage deviation should be interpreted in absolute value only.
26
Figure 5. Impact of credit constraints on the model economy (scenario 3) CONSUMPTION
GDP DEFLATOR
INVESTMENT
11000
1.15
4000
10000
1.10
3600
1.05
3200
9000
4
8000
4
0 7000
0
-4
-4 -8
0.95
10
0.90
0
0.85
-8 -12 2000
2001
2002
2003
2004
2005
2001
2002
2003
2004
2005
2000
EMPLOYMENT 50000 45000 40000
2
35000
0
30000
-2
2005
1520
-400
1480
-600
1400
1.0 0.5
-800
1360
-1000
0
-1200
-5
0.0 -0.5 -1.0
-15 -20
-2.0 2005
-1400
-10
-1.5 2004
2004
5
-6 -8
2003
1440
-4
-10
2002
NET EXPORTS*
55000
2003
2001
I (Scenario 2) I (Baseline) Percent Dev iation
60000
2002
-10
DEFL (Scenario 2) DEFL (Baseline) Percent Dev iation
CAPITAL
2001
2000
-30 2000
CONS (Scenario 2) CONS (Baseline) Percent Dev iation
2000
2400
-20
-12 -16
2800
1.00
-25 2000
2001
K (Scenario 2) K (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
2001
L (Scenario 2) L (Baseline) Percent Dev iation
GOV'T EXPENDITURE
2002
2003
2004
2005
NX (Scenario 2) NX (Baseline) Percent Dev iation
CORPORATE EARNINGS 3000
NET CREDIT (FIRMS) 1400
1600
1200
2800
1200
1000 2600
800
2 2400
0
800
600
10
400
400 2200
-2
0
-4
-10
-6
-20
-8 2000
2001
2002
2003
2004
0 -400
-30
2005
-800 2000
2001
G (Scenario 2) G (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
PROFITS (Scenario 2) PROFITS (Baseline) Percent Dev iation
2001
2002
2003
2004
2005
NCF (Scenario 2) Baseline
REAL GDP
NOMINAL GDP
4
18000
18000
16000
16000
14000
14000
12000
0
10000
-4
12000
1 0
10000
-1 -2
-8
-3
-12
-4 -5
-16 2000
2001
2002
2003
2004
Y N (Scenario 2) Y N (Baseline) Percent Dev iation
2005
2000
2001
2002
2003
2004
2005
Y R (Scenario 2) Y R (Baseline) Percent Dev iation
* Net exports are negative so percentage deviation should be interpreted in absolute value only.
27
Figure 6. Impact of credit constraints and prohibitively high interest rates on the model economy (scenario 4) CONSUMPTION
GDP DEFLATOR
INVESTMENT
11000
1.2 1.1
9000
4
8000 7000
2800 2400
10 0.9
2000
0
-4
-5 0.8
-8
-10
3200
1.0
0
5 0
4000 3600
10000
-10 -20
-15
-12
-20 -25
-30
-16 2000
2001
2002
2003
2004
2005
-40 2000
2001
CONS (Scenario 3) CONS (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
2001
DEFL (Scenario 3) DEFL (Baseline) Percent Dev iation
CAPITAL
2002
2003
2004
2005
I (Scenario 3) I (Baseline) Percent Dev iation
EMPLOYMENT
NET EXPORTS*
60000
1500
-400
55000
-600
50000
1450 -800
45000 40000
5
1400
1
-1000
0
35000
0
10
30000
1350
0
-5
-1200
-10
-1400
-20
-10
-1 -30
-15 -20
-2 2000
2001
2002
2003
2004
2005
-40 2000
2001
K (Scenario 3) K (Baseline) Percent Dev iation
2002
2003
2004
2005
2000
2001
L (Scenario 3) L (Baseline) Percent Dev iation
CORPORATE EARNINGS
GOV'T CONSUMPTION
2002
2003
2004
2005
NX (Scenario 3) NX (Baseline) Percent Dev iation
NET CREDIT (FIRMS) 1400
3000
1600
1200 2800
4
800
2200
-4
800
600
10 2400
0
1200
1000
2600
400
0
400
200
-10
0
-20 -30
-8
-400
-40 -50
-12 2000
2001
2002
2003
2004
-800 2000
2005
2001
2002
2003
2004
2005
2000
2001
PROFITS (Scenario 3) PROFITS (Baseline) Percent Dev iation
G (Scenario 3) G (Baseline) Percent Dev iation
NOMINAL GDP
2002
2003
2004
2005
Baseline Percent Dev iation
REAL GDP
LOAN RATES (FIRMS)
18000
18000
13
16000
16000
12
14000
14000
11 10
12000
5
2
12000
9
0
0
10000
8
10000
-5
-2
-10
-4
-15
-6
6
-20
-8
5
-25
-10 2000
2001
2002
2003
2004
YN (Scenario 3) YN (Baseline) Percent Dev iation
2005
7
4 2000
2001
2002
2003
2004
YR (Scenario 3) YR (Baseline) Percent Dev iation
2005
2000
2001
2002
2003
2004
2005
Baseline Percent Dev iation
* Net exports are negative so percentage deviation should be interpreted in absolute value only.
28
References
Backe, P., B. Egert and T. Zumer (2006). Credit Growth in Central and Eastern Europe: New (Over)Shooting
Stars?
Focus
on
European
Economic
Integration.
112-139.
Oesterreichichische Nationalbank, Vienna. Bean, C.,
J. Larsen and K. Nikolov (2002). Financial Frictions and the Monetary
Transmission Mechanism: Theory, Evidence and Policy Implications. ECB Working Paper No. 113. ECB, Frankfurt. Bernanke, B., and M. Gertler (1995). Inside the Black Box: the Credit Channel of Monetary Policy Transmission. NBER Working Paper No. 5146. NBER Cambridge, MA. Brunner, K., and A.H. Meltzer (1988). Money and Credit in the Transmission Process. American Economic Review Vol. 78 No. 2: 446-451. Brzoza-Brzezina, M. (2005). Lending Booms in the New EU Member States. Will Euro Adoption Matter? ECB Working Paper No. 543. ECB, Frankfurt. Calza, A., M. Manrique and J. Sousa (2003). Aggregate Loans to the Euro Area Private Sector. ECB Working Paper No. 202. ECB, Frankfurt. Cottarelli, C., G. Dell’Ariccia and I. Vladkova-Hollar (2003). Private Sector in the Central and Eastern Europe and the Balkans. IMF Working Paper No. WP/03/213. IMF, Washington. Detken, C., and F. Smets (2004). Asset Price Booms and Monetary Policy. ECB Working Paper No. 364. ECB, Frankfurt. Enoch, C., and I. Otker-Robe (forthcoming). Rapid Credit Growth in Central and Eastern Europe: Endless Boom of Early Warning. Palgrave Publishing. Fratzscher, M., and M. Bussiere (2004). Financial Openness and Growth: Short-Run Gain, Long-Run Pain? ECB Working Paper No. 348. ECB, Frankfurt. Gertler, M., and S. Gilchrist (1994). Monetary Policy, Business Cycles, and the Behaviour of Small Manufacturing Firms. The Quarterly Journal of Economics Vol. 109, No. 2: 309-340. Kohler, M., J. Hommel and M. Grote (2006). The Role of Banks in the Transmission of Monetary Policy in the Baltics. Discussion Paper No. 06-005. Centre for European Economic Research. Kuodis, R., and I. Vetlov (2002). Monetary Policy Transmission Mechanism in Lithuania. Monetary Studies 3: 27-63 (in Lithuanian). Lietuvos bankas, Vilnius. McKinnon, R., and H. Pill (1997). Credible Economic Liberalizations and Overborrowing. The American Economic Review, Vol. 87, No. 2: 189-193.
29
Pedersen, A.M. (2003). The Credit Channel in Monetary-Policy Analyses. Monetary Review 4: 85-101. Danmarks Nationalbank, Copenhagen. Ramanauskas, T. (2006). Blowing bubbles, enjoying economic growth. Baltic Economic Trends 1: 23-29. BICEPS and SSE Riga, Riga. Ramanauskas, T. (2005). Assessment of Bank Credit Growth from Macroeconomic Perspective. Monetary Studies 3: 78-97 (in Lithuanian). Lietuvos bankas, Vilnius. Ramanauskas, T. (forthcoming). Assessment of Credit Growth in Lithuania. In Enoch, C., and I. Otker-Robe Rapid Credit Growth in Central and Eastern Europe: Endless Boom or Early Warning. Palgrave Publishing. Romer, C., and D. Romer (1990). New Evidence on the Monetary Transmission Mechanism. Brookings Papers on Economic Activity, No. 1: 149-213. Vetlov, I. (2004). The Lithuanian block of the ECSB multi-country model. BOFIT Discussion Paper No.13. BOFIT, Helsinki.
30
