Credit and the exchange rate

Credit and the exchange rate

Abstract

Financial innovations during the eighties have led to a change in the relationship between money and credit as well as GDP and credit. Therefore I modify the classical monetary model of the exchange rate for it to include non-GDP transactions and compare the predictive capabilities of credit variables with the well tested monetary variables on the exchange rate. Both the monetary and credit variables show only little predictive power, but are not outperformed by

monetary and output variables.

This paper tries to explore if there is a role for credit as a fundamental variable determining the exchange rate. The exchange rate is often seen as an asset price. This means a currency is held as an asset that provides a return. The general approach to asset pricing is to first determine the fundamental value drivers of the asset’s price. Then the present value of the expected fundamental value should equal the price of the asset. If at any point in time this is not the case arbitrageurs can close this gap by buying or selling the under- or overvalued currency.

What are the fundamentals that drive the exchange rate? Where company earnings are seen as a fundamental driver for a stock price, national output can be seen as the driver of the exchange rate. However periods of hyperinflation showed that the money supply is also reflected in the exchange rate. This is the line of thought that inspired the monetary fundamental models that considers national output and money supply the fundamental drivers of the exchange rates. These models also include an

1

unobserved variable because they found that while there is evidence for money and output influence this cannot explain all of the exchange rate movement. These models are the dominant exchange rate models today.

It is widely acknowledged that the bulk of the money supply is determined by bank or Monetary Financial Institution (MFI) credit creation. Recently Schularick and Taylor (2012) showed that while before world war II credit and money aggregates were closely related, however they decoupled after the war. Engel and West (2005) also mention this change: “it may be that the appropriate measure of the money supply has permanently changed because of numerous financial innovations over our sample, so that the M1 money supply series vary from the “true” money supply by some I(1) errors.”

Looking at the past decade, rapid credit growth in a country was accompanied by its currency appreciating. This is in sharp contrast to countries with rapidly increasing money supply where this was accompanied by hyperinflation and a depreciating currency.

Like money it seems that credit and GDP had a stable relationship that started to break down after the eighties with credit overtaking GDP (Bezemer, 2009; Schularick & Taylor, 2012; Werner, 2012). One reason GDP is thought of as a driver of the exchange rate is that it represents the number of transactions in goods and services. However the same financial innovations that decoupled credit and money could also affect the use of the national output variable to measure the total amount of transactions. For example many of the transactions in the housing and stock market were driven by credit and are not measured by GDP. This means that it is possible that credit can give us additional information on the demand for a currency.

Banks are often seen as a highly informed investor. They should only extend credit if they expect that future income is able to repay the debt. If total credit rises rapidly this means the banks expect this debt can be repaid with future income. This might signal other investors that banks expect high future income growth, this could lead to increased demand for this currency.

power. I will use the macro-economic database of Schularick and Taylor (2012) and combine their data with the annual exchange rates between fourteen OECD countries for the period of 1975 until 2008.

The rest of the paper will proceed as follows. Section I presents a literature review to give the reader an idea of previous authors that have tried to find out what drives the exchange rate and the role of credit in the economy. Section II I present a modification to the basic monetary exchange rate model by modifying the quantity of money equation so that it also includes non-GDP transactions. After that I present details of the data construction and descriptive statistics in Section III. Section IV deals with my empirical findings: it presents the variables and basic testing procedure as well as the results. Appendix A shows the detailed results of unit root tests. The paper ends with a conclusion.

I. Literature review

The monetary exchange rate model was developed during the seventies. It states that the fundamental values underlying the exchange rate are the log relative money stocks and the log real incomes. These early models emphasized the importance of money in determining the balance of payments if the exchange rate is pegged or the exchange rate itself if it is flexible (Frankel, 1976). He viewed the exchange rate as the relative price of two assets that is determined by the willingness of people to hold these assets. Dornbusch (1976) emphasized the importance of money and production as fundamentals that determine the exchange rate. An expected increase in prices of assets will produce higher returns on assets and therefore induce a capital inflow and appreciate the exchange rate.

While these models seemed to have a sound theoretical basis Meese & Rogoff (1983) showed that they had little predictive value and had difficulty outperforming a simple random walk model out of sample. Models in the 1990s tried to improve the original models by incorporating sticky prices and adding extra variables. However these did not lead to a consensus view on what drives the exchange rate (Cheung, Chinn, & Pascual, 2005).

the puzzle are that it is hard to predict the exchange rate using macroeconomic fundamentals and the exchange rate is more volatile than any fundamental can account for.

Engel and West (2005) presented a new approach, and suggested the current exchange rate is an asset price that is determined by the current level of fundamentals as well as expectations of future fundamentals. If the expectations are discounted with a rate close to one and if the fundamentals are non-stationary this should lead to random walk like behavior of the exchange rate. There is some evidence that the first assumption (Engel, Mark, & West, 2007; Engel & West, 2005) and second assumption (Sarno & Sojli, 2009) are indeed valid. Furthermore Cerra and Saxena (2010) tested this model and found strong evidence for cointegration between nominal exchange rates and monetary fundamentals for a large dataset of 98 countries. More recently Balke, Ma and Wohar (2012) found that many of the long run exchange rate movements can be explained by directly observed monetary fundamentals and money demand shifters. However they found it difficult to explain short run movements.

Many different macro-economic variables were investigated in exchange rate models but to my knowledge total MFI credit was never an included variable. The most obvious reason for this is the close relationship of money and credit. Most modern money is created when monetary financial institutions extend loans. However not all MFI loans are backed by a monetary liability. Schularick and Taylor (2012) examined the historical relation between credit and ‘broad money’ and found that after world war II these two variables diverge for a wide array of countries.

A second argument not to include credit is its close relation to output as loans are supposed to improve economic growth. Bezemer (2009) and Werner (2012) investigated this relationship and made a distinction between productive and non-productive credit. They found a link between credit provided to industry and GPD growth for the United States and Japan. They found that total credit and GDP had a steady relationship but credit started to outpace GDP after the financial liberalization during the eighties.

during and after the eighties. Perhaps the existing monetary exchange rate model can be improved by adding a credit variable as a fundamental?

II: An adjusted monetary exchange rate model

One of the most well-known and empirically tested asset pricing exchange rate models is that of Engel and West (2005). They imply that the exchange rate is the present discounted value of a linear combination of expected economic fundamentals and unobservable shocks. The model presented below uses the notation of one of its most recent interpretations (Balke, et al., 2012).

(1)

where is the spot exchange rate, is the current value of observed fundamentals, and is the discount factor. Finally the term includes current and expected future values of unobserved fundamentals, such

as risk premia or money demand shocks, and possibly non-fundamental determinants of the exchange rate. The observable fundamental variables in a simple monetary model, which is a variant of the Cagan (1956) model, are: _{. Here is the log of the home money supply,}

is the log of the home price level, is the level of the home interest rate, is the log output and _is

an unobservable shock to money demand.

As noted in the literature review rising total MFI credit allows for increased amount of transactions in a certain currency without an actual increase in output. I assume the original quantity theory of money equation, which states that effective money in circulation should be equal to the total value of the transactions in the economy , still holds. Here M is the total money supply, V the

velocity in which it circulates, P the price level and Q the amount of transactions. This means that the modification of no longer holds because the amount of transactions outside of GDP, for example transactions in assets, has increased (Werner, 2012).

equal to C, where C is the total amount of credit outstanding. However as we have seen in the literature review the relationship between credit and money has changed and credit has overtaken money. This means that banks have paired loans with other non-deposit liabilities which will not show up in the current measures of the money supply. This leads to his modification of the quantity theory of money equation: . Where C replaces M as an estimator of the true money supply because it includes non-deposit

bank liabilities. Also the total number of transactions now includes asset transactions, this means . Here is the price of GDP transactions, A is the number of Asset transactions and is the

price of assets transactions.

If this modification to the core of monetary theory is valid it leads to a simple modification to the classic monetary explanation of the exchange rate.

₍₂₎

Where is the total amount of credit and the amount of asset transactions. The parameters of the foreign country are _{respectively.}

The nominal exchange rate is equal to its purchasing power parity value plus the real exchange rate.

(3)

Where is the deviation from uncovered purchasing power parity. Besides the uncovered purchasing

power parity, interest parity is described as

(4)

Where is the deviation from rational expectations uncovered interest parity.

These equations are rearranged to form

(5)

rate parity: . Finally the discount

factor is written as

. Substituting this into equation 5 leads to

(6)

Finally if equation 6 is applied repeatedly the model can be solved as equation 7, with the assumption of no explosive solution.

(7)

Which is similar to equation (1) in that it states that the exchange rate is equal to the value of the current fundamentals, the present value of the expected future fundamentals and the present value of the unobserved and non-fundamental determinants of the exchange rate.

III. Empirical findings

To test if adding the credit variable can improve the original monetary model, the empirical approach of Engel and West (2005) serves as inspiration. If the above presented model is correct is the present value of current and future values of the observable fundamental as well the unobservable fundamental . If this is the case in reality than this could be supported by finding Granger Causality from the lag of to . In addition if holds some information in addition to the past values of than one can support the model

if one finds Granger Causality from the lags of to . But because exchange rate models allow for an unobservable variable failure to find Granger causality does not mean the model does not hold. However Granger causality still supports the view that the model holds.

Table I:

Exchange rate Fundamentals

Fundamental Notation

Monetary fundamental (M-M*)

Output fundamental (Y-Y*)

Credit fundamental (C-C*)

Consumer price fundamental (P-P*)

Short term interest fundamental (I-I*)

Money output fundamental (M-M*)-(Y-Y*)

Credit output fundamental (C-C*)-(Y-Y*)

Money credit fundamental (M-M*)-(C-C*)

a Variable definitions: Foreign variables are denominated with an “*”.M =log value M1; Y = log value real GDP; C = log value MFI credit; P = log CPI; I = short term interest rate.

A. Data and basic statistics

Almost all of the data used comes from Schularick and Taylor (2012). This includes the narrow money supply M1, CPI price level, short term interest rates and aggregated year end amount of outstanding Monetary financial institution loans to domestic Households and non-financial businesses in the domestic currency. I convert all data except the interest rates by taking logs and multiplying by 100.

includes only commercial banks and no savings banks or credit co-operatives. Finally this measure of credit does not include credit to the financial sector, which has dramatically increased since the eighties (Bezemer, 2009).

I obtained the annual year end exchange rates of these countries against the British Sterling and the US dollar from the bank of England. This data spans from 1975 to 2008. This makes the sample size per country 33. For the countries that joined the Eurozone data spans from 1975 to 2001 making their sample size 26.

Tables II and III provide basic statistics for all variables that are tested as fundamental and the actual exchange rates. The first rows of both tables show that exchange rates for both the UK and the USA are serially uncorrelated and quite volatile. All exchange rate standard deviations range between 7,63% and 14,94%. First order autocorrelations are small and only significant at the 10% level in the case of Sweden and at the 5% level in the case of Italy and Spain.

Table II

Basic statistics UK

Rejections at 1% (***), at 5% (**) and at 10% (*) H0: no serial correlation

Fundamental AUS CAN DEN FRA GER ITA JPN NET NOR SPE SWE CHE GBP USA

Mean 0,75 -0,35 -1,47 0,66 -1,94 3,18 -4,72 -1,58 -0,34 0,75 0,75 -3,77 0,00 -1,04 Δs Stdev 12,37 12,79 11,11 9,34 11,37 8,80 14,94 11,21 8,41 9,84 9,84 12,47 0,00 13,50 p1 -0,10 -0,11 -0,14 -0,21 -0,14 -0,39** 0,02 -0,16 -0,18 -0,36** -0,28* -0,08 0,00 -0,02 Mean -3,39 -1,17 -2,87 0,18 -1,31 -2,97 -0,81 -0,89 0,24 -6,12 0,92 0,39 0,00 -2,35 Δ(M-M*) Stdev 6,11 4,70 6,74 3,90 3,89 4,37 6,07 5,65 10,26 16,73 2,99 11,00 0,00 11,22 p1 0,18 0,27 0,39** 0,43** 0,56*** 0,64*** 0,41** 0,08 -0,35* -0,13 0,17 -0,05 0,00 -0,24 Mean 0,00 0,24 0,28 0,39 0,47 0,30 0,10 0,27 -0,52 -0,39 0,43 1,00 0,00 0,28 Δ(Y-Y*) Stdev 2,00 1,72 1,75 1,56 2,03 1,95 2,35 1,85 2,21 1,74 1,70 2,17 0,00 1,76 p1 0,04 0,06 0,28* 0,50** 0,04 0,46** 0,67*** 0,44** 0,44** 0,49** 0,38** 0,34** 0,00 0,18 Mean -1,60 2,35 1,05 3,02 5,83 1,34 6,81 2,07 -0,19 -2,73 1,14 6,61 0,00 3,71 Δ(C-C*) Stdev 4,51 7,01 7,92 6,09 6,21 5,92 4,29 6,87 6,80 7,40 6,50 5,13 0,00 4,65 p1 0,33* 0,27 0,59*** 0,44** 0,77*** 0,31 0,69*** 0,79*** 0,54*** 0,12 0,15 0,52*** 0,00 0,63*** Mean -0,13 1,04 0,81 0,80 2,67 -1,60 3,36 2,28 0,56 -2,04 0,22 2,98 0,00 1,05 Δ(P-P*) Stdev 2,64 2,56 2,27 2,38 3,67 2,35 2,29 3,34 3,01 2,66 2,24 3,89 0,00 2,65 p1 0,52*** 0,33** 0,45*** 0,55*** 0,73*** 0,75*** 0,71*** 0,82*** 0,50*** 0,49** 0,22 0,72*** 0,00 0,50*** Mean -0,01 0,00 0,00 0,01 0,03 -0,02 0,04 0,03 -0,01 -0,02 0,00 0,05 0,00 0,02 Δ(I-I*) Stdev 0,02 0,02 0,03 0,03 0,02 0,04 0,03 0,02 0,03 0,04 0,02 0,02 0,00 0,02 p1 -0,08 -0,39 0,17 0,05 -0,12 0,24 -0,13 -0,01 0,04 0,08 -0,06 -0,26 0,00 -0,26 Mean -3,39 -1,41 -3,15 -0,21 -1,79 -3,27 -0,90 -1,16 0,76 -5,73 0,49 -0,61 0,00 -2,63 Δ(M-M*)-Δ(Y-Y*) Stdev 6,11 5,21 6,56 4,28 4,71 3,87 6,50 5,85 10,34 16,82 3,28 11,92 0,00 11,95 p1 0,18 0,10 0,31* 0,50** 0,41** 0,46** 0,52*** 0,13 -0,30 -0,05 0,17 0,05 0,00 0,14 Mean -1,60 2,10 0,77 2,63 5,36 1,04 6,71 1,80 0,33 -2,34 0,70 5,62 0,00 3,43 Δ(C-C*)-Δ(Y-Y*) Stdev 4,42 6,37 8,04 6,14 6,27 5,92 4,62 6,63 6,74 7,26 6,98 5,22 0,00 4,96 p1 0,23 0,36** 0,55*** 0,41** 0,60** 0,38*** 0,62*** 0,73*** 0,56*** 0,01 0,13 0,54*** 0,00 0,66*** Mean -1,79 -3,52 -3,93 -2,85 -7,14 -4,31 -7,61 -2,96 0,42 -3,39 -0,21 -6,22 0,00 -6,06 Δ(M-M*)-Δ(C-C*) Stdev 7,62 9,69 8,78 6,42 6,66 7,45 8,53 8,94 10,87 18,42 6,49 13,02 0,00 11,82 p1 0,06 0,27 0,61*** 0,21 0,73*** 0,47** 0,60*** 0,41** -0,16 -0,08 0,17 -0,06 0,00 0,17

a Variable definitions: Δs = percentage change in UK Sterling exchange rate (higher value indicates depreciation). Non UK variables are denominated with an “*”.ΔM =percentage change in M1; ΔY is percentage change in real GDP; ΔC = percentage change in MFI credit; ΔP = percentage change in consumer prices; I = short term interest rate. p1 = the first order autocorrelation coefficient.

Table III

Basic statistics USA

Rejections at 1% (***), at 5% (**) and at 10% (*) H0: no serial correlation

Fundamental AUS CAN DEN FRA GER ITA JPN NET NOR SPE SWE CHE GBP USA

Mean 1,78 0,69 -0,43 1,92 -0,68 4,45 -3,68 -0,31 0,69 4,39 1,78 -2,73 -0,19 0,00 Δs Stdev 11,42 7,63 12,61 13,10 12,89 13,34 13,07 12,88 11,65 13,86 13,63 13,18 16,94 0,00 p1 0,08 0,04 0,27 0,32 0,31 0,28 0,10 0,30 0,05 0,36* 0,25 0,18 -0,02 0,00 Mean -1,04 1,17 -0,53 2,52 1,03 -0,63 1,54 1,46 2,58 -3,77 3,27 2,74 2,35 0,00 Δ(M-M*) Stdev 13,09 11,34 13,67 12,69 11,98 12,34 13,60 13,29 12,87 21,99 11,45 13,68 11,22 0,00 p1 -0,29 0,10 -0,11 0,34 0,10 0,57*** 0,33 0,18 -0,40* -0,15 -0,95** 0,08 -0,24 0,00 Mean -0,28 -0,04 0,00 0,11 0,19 0,02 -0,19 -0,02 -0,81 -0,67 0,15 0,71 -0,28 0,00 Δ(Y-Y*) Stdev 2,18 1,77 1,92 1,99 2,11 1,91 2,30 1,96 2,09 2,14 2,22 2,30 1,76 0,00 p1 0,38* 0,69** -0,01 0,15 -0,27 0,16 0,41** 0,25 0,28 0,30 0,43** 0,38** 0,18 0,00 Mean -5,31 -1,37 -2,66 -0,69 2,12 -2,37 3,09 -1,65 -3,90 -6,44 -2,58 2,90 -3,71 0,00 Δ(C-C*) Stdev 5,27 6,61 7,41 7,69 5,59 6,48 5,90 5,72 6,29 7,96 7,94 4,60 4,65 0,00 p1 0,35** 0,26 0,49** 0,44** 0,58*** 0,27 0,77*** 0,45** 0,48** 0,14 0,24 0,25 0,63*** 0,00 Mean -1,18 -0,01 -0,24 -0,25 1,62 -2,65 2,31 1,23 -0,49 -3,09 -0,83 1,93 -1,05 0,00 Δ(P-P*) Stdev 2,73 1,54 1,77 2,12 1,93 3,47 1,97 2,19 2,43 3,84 2,44 2,28 2,65 0,00 p1 0,47*** 0,40** 0,59*** 0,78*** 0,75*** 0,78*** 0,40*** 0,65*** 0,59*** 0,75*** 0,48*** 0,72*** 0,50*** 0,00 Mean -0,02 -0,01 -0,02 -0,01 0,01 -0,04 0,03 0,01 -0,02 -0,03 -0,01 0,03 -0,02 0,00 Δ(I-I*) Stdev 0,03 0,02 0,03 0,02 0,03 0,04 0,03 0,02 0,03 0,04 0,03 0,03 0,02 0,00 p1 0,12 -0,09 -0,06 0,30 0,43** 0,09 -0,05 0,12 0,11 -0,22 0,02 0,22 -0,26 0,00 Mean -0,76 1,22 -0,52 2,42 0,84 -0,64 1,73 1,47 3,39 -3,10 3,12 2,02 2,63 0,00 Δ(M-M*)-Δ(Y-Y*) Stdev 14,10 12,14 14,14 13,56 12,94 12,50 13,88 14,05 13,44 22,69 12,14 14,84 11,95 0,00 p1 -0,06 0,24 -0,04 0,25 0,07 0,50* 0,49 0,24 -0,34 -0,07 -0,54 0,20 0,14 0,00 Mean -5,03 -1,33 -2,66 -0,80 1,93 -2,39 3,28 -1,63 -3,10 -5,77 -2,73 2,19 -3,43 0,00 Δ(C-C*)-Δ(Y-Y*) Stdev 5,10 6,84 7,32 6,99 5,94 6,41 5,40 5,88 6,51 7,61 8,29 5,12 4,96 0,00 p1 0,39** 0,33* 0,43** 0,53** 0,34* 0,31 0,64*** 0,38** 0,50*** 0,19 0,21 0,20 0,66*** 0,00 Mean 4,27 2,54 2,13 3,21 -1,08 1,75 -1,55 3,10 6,48 2,67 5,85 -0,16 6,06 0,00 Δ(M-M*)-Δ(C-C*) Stdev 14,50 12,51 14,98 14,65 13,52 13,52 16,04 13,97 13,74 23,53 13,76 14,93 11,82 0,00 p1 -0,01 0,09 0,32 0,16 0,21 0,39* 0,52* 0,11 -0,18 -0,10 0,05 0,15 0,17 0,00

a Variable defenitions: Δs = percentage change in US Dollar exchange rate (higher value indicates depreciation). Non US variables are denominated with an “*”.ΔM =percentage change in M1; ΔY is percentage change in real GDP; ΔC = percentage change in MFI credit; ΔP = percentage change in consumer prices; I = short term interest rate. p1 = the first order autocorrelation coefficient.

B. Cointegration and Granger causality tests

I test if the first level difference of all variables is stationary using an augmented Dickey-Fuller test with a time trend included. I test at level with automatic lag selection using the Schwarz info criterion with a maximum of eight lags for non-euro countries and six for countries that converted to the euro in 2001. I fail to reject unit root for most fundamentals. For most variables unit root is at most rejected 4 out of 28 times. Prices unit root is rejected nine times and interest levels unit root is rejected twelve out of twenty eight times. Table AI in data appendix A presents a more detailed overview of all unit root rejections. Therefore I will from here on use these fundamentals in first difference form with the exception of interest rates which I also test in level form.

I test all measures of fundamentals for cointegration with the Johansen cointegration test with trace and maximum eigenvalue and critical values from Osterwald-Lenum. Every bivariate VAR has one lag. Out of the 221 tests are 56 rejections of the null of no cointegration using the trace statistic and 21 using the maximum eigenvalue statistic. There is especially evidence for cointegration with the interest rate as the trace statistic was rejected a total of 15 times and the eigenvalue statistic 10 in 28 tests. For all other variables there is no strong evidence for cointegration, this can be explained if the unobservable variables are very persistent or have a permanent component.

Table IV Cointegration tests

Rejections at 5%: Trace statistic (T), Eigenvalue (E) or both (ET) H0 Trace: cointegration vectors H0 Eigenvalue:

Fundamental AUS CAN DEN FRA GER ITA JPN NET NOR SPE SWE CHE GBP USA

(M-M*) UK USA ET T T ET T E T ET T (Y-Y*) UK USA T T T ET T (C-C*) UK USA ET T ET E T T (P-P*) UK USA T T T T T ET T T T (I-I*) UK USA T T E T ET ET ET ET T ET ET T ET ET T ET (M-M*)-(Y-Y*) UK USA T T T T ET ET T T ET ET (C-C*)-(Y-Y*) UK USA T T T T ET ET T T T (M-M*)-(C-C*) UK USA

a Variable definitions: Foreign variables are denominated with an “*”.M = M1 money supply; Y = real GDP; C = MFI credit; P = consumer prices; I = short term interest rate.

b Johansen test for cointegration with Ostwald Lenum critical values and 1 lag.

c Data is annual from 1975-2008 or 1975-2001 for countries that converted to Euro (FRA, GER, ITA, NET and SPE).

I focus on the bivariate relationship between the exchange rate and the fundamentals: , , , ,

, , and , because I try to see how credit variables perform next to monetary variables in a relatively unstructured investigation. The main goal of the empirical analysis is to find Granger causality where either the fundamental variable predicts the exchange rate or the other way

around.

Table V

Bivariate Granger causality tests different measures of Δ

A: Rejections at 1% (***), at 5% (**) and at 10% (*) of H0: Δ fails to cause Δ

Fundamental AUS CAN DEN FRA GER ITA JPN NET NOR SPE SWE CHE GBP USA

B: Rejections at 1% (***), at 5% (**) and at 10% (*) of H0: Δ fails to cause Δ

Fundamental AUS CAN DEN FRA GER ITA JPN NET NOR SPE SWE CHE GBP USA

∆(M-M*) UK USA ** ∆(Y-Y*) UK USA ** ** ** ∆(C-C*) UK USA ∆(P-P*) UK USA ** ** (I-I*) UK USA * ∆(I-I*) UK USA *** ** ∆(M-M*)-(Y-Y*) UK USA ** ∆(C-C*)-(Y-Y*) UK USA * * ∆(M-M*)-(C-C*) UK USA **

a Variable definitions: Foreign variables are denominated with an “*”.ΔM =percentage change in M1; ΔY is percentage change in real GDP; ΔC = percentage change in MFI credit; ΔP = percentage change in consumer prices; I = short term interest rate. b Granger causality is tested with 1 lag (using Akaike information criterion)

c Data is annual and usually 2008. The exceptions are France, Germany, Netherlands and Spain where the sample is 1976-2001. And for the no- differenced interest rate the sample starts at 1975 instead of 1976.

Table V A shows that the null that Δ fails to cause Δ can be rejected in 2 cases at the 1% level, 13 times at the 5% level and 14 times at the 10% level. Of these rejections the added credit variables account for 4 rejections at the 5% level and 3 rejections at the 10% level.

Table V B presents the null that Δ fails to cause Δ . This null is rejected 3 times at the 1% level, 7

times at the 5% level and 4 times at the 10% level. The credit variables account for 2 rejections at the 5% level and 2 rejections at the 10% level.

The credit variable on itself performs comparable to the monetary variable and can help predict the exchange rate only in one case. The exchange rate can help predict the credit fundamental in two cases compared to three for output and one for the money supply.

To find out if multiple variables interacting can help predict the exchange rate I use a multivariate VAR causality test. The different groupings are used based on the model in section three and the comparison with Engel and West (2005). There are the original five groupings; (Δ , ,

, )`, (Δ , , ) `, (Δ , , )`, (Δ , , , ))`, (Δ , , , ))` and three additional groupings: (Δ , , ) `, (Δ , , ) `, (Δ , , , ))`.

The multivariate VAR causality results are presented in table VI. There are rejections in 1 case at the 1% level, 11 cases at the 5% level and 15 cases at the 10% level. Of these rejections the added credit variables account for 2 rejections at the 5% level and 4 rejections at the 10% level.

Table VI

VAR causality tests

A: Rejections at 1% (***), at 5% (**) and at 10% (*) of H0: Δ fails to cause Δ

Fundamental AUS CAN DEN FRA GER ITA JPN NET NOR SPE SWE CHE GBP USA

∆(Y-Y*), ∆(P-P*), I-I* UK USA * ** * * ∆(M-M*), ∆(Y-Y*) UK USA ** * ** ∆(P-P*), ∆(Y-Y*) UK USA ** ** ** ** * ∆(M-M*), ∆(Y-Y*), ∆(P-P*) UK USA * * * * * ∆(Y-Y*), ∆(P-P*), ∆(I-I*) UK USA ** *** ** * ∆(M-M*), ∆(C-C*) UK USA * * ∆(C-C*), ∆ (Y-Y*) UK USA ** * ** ∆(M-M*), ∆(Y-Y*), ∆(C-C*) UK USA *

a Variable definitions: Foreign variables are denominated with an “*”.ΔM =percentage change in M1; ΔY is percentage change in real GDP; ΔC = percentage change in MFI credit; ΔP = percentage change in consumer prices; I = short term interest rate.

b Granger causality is tested with 1 lag (using Akaike information criterion)

The fundamentals with credit perform comparable or slightly worse than the original money and output fundamental.

IV. Conclusion

The above results are comparable to Engel and West (2005) and the literature that followed. If the present value model holds it can be supported by causation from either the fundamentals to the exchange rate or from the exchange rate to the fundamentals. However failure to find such direct causation could either mean there is no relationship or the discount rate is close to one.

Overall there is only weak evidence that fundamentals help predict the exchange rate. There is weak evidence for cointegration and weaker evidence for Granger causality. It seems that taking into account non-GDP transactions by including credit, the unobservable fundamentals still account for much of the exchange rate movements.

So the addition of a credit variable does not change the overall conclusions of the monetary model. But does it add explanatory power to the overall model? It does not with the cointegration tests. However it does in some cases with the Granger causality tests. By including credit I improved explanatory power slightly for the Japan, the Netherlands and Spain with bivariate VAR and only for Spain with multivariate VAR. This is only very weak evidence.

On the other hand I could not find evidence that replacing either the money or output variable with credit is useless. The credit model performs comparable to the original. I have used an existing dataset for this analysis, however it does have its limitations in that it only has annual observations for a limited time period. Also in the definition of credit, total credit to the non-MFI financial sector is left out while this is the credit that has mostly funded transactions in financial asset markets.

Appendix A

Table AI

Unit root tests augmented Dickey Fuller

A: Rejections at 5% (***), at 5% (**) and at 10% (*) of H0: no unit root

Fundamental AUS CAN DEN FRA GER ITA JPN NET NOR SPE SWE CHE GBP USA

(M-M*) UK USA * ** *** ** (Y-Y*) UK USA * * * *** (C-C*) UK USA * * (P-P*) UK USA * * *** *** * ** * ** ** (I-I*) UK USA * *** *** * * * ** *** *** *** * * (M-M*)-(Y-Y*) UK USA ** *** * (C-C*)-(Y-Y*) UK USA *** * *** (M-M*)-(C-C*) UK USA ** *

a Variable definitions: Foreign variables are denominated with an “*”.M = M1 money supply; Y = real GDP; C = MFI credit; P = consumer prices; I = short term interest rate.

References

Balke, Nathan S., Ma, Jun and Wohar, Mark E., 2012, The contributions of Economic Fundamentals to Movements in the Exchange Rate, Journal of International Economics,

http://dx.doi.org/10.1016/j.jinteco.2012.10.003.

Bezemer, Dirk J., 2009, Banks as Social Accountants: Credit and Crisis through an Accounting Lens,

MPRA Paper, No. 15766.

Cagan, Philip, 1956, The monetary dynamics of hyperinflation, Friedman M (ed.), Studies in the quantity theory of money, University of Chicago press, 25-117.

Cerra, Valerie, Saxena, Sweta C., 2010, The Monetary Model Strikes Back: Evidence from the World,

Journal of International Economics, 81, 184-196.

Cheung, Yin-Wong, Chinn, Menzie D. and Pascual, Antonio G., 2005, Empirical Exchange Rate Models of the Nineties: Are Any Fit to Survive? Journal of International Money and Finance, 24(7), 1150-1175.

Dornbusch, Rudiger, 1976, Expectations and Exchange Rate Dynamics, Journal of Political Economy, 84, 1161-1176.

Engel, Charles, Mark, Nelson C., and West, Kenneth D., 2007, Exchange Rate Models Are Not as Bad as You Think, NBER Macroeconomics Annual, University of Chicago Press, pp.381-441.

Engel, Charles and West, Kenneth D., 2005, Exchange Rate and Fundamentals. Journal of Political

Economy, 113, 485-517.

Frankel, Jeffrey, 1976, A Monetary Approach to the Exchange Rate: Doctrinal Aspects and Empirical Evidence, Scandinavian Journal of Economics, 78, 200-224.

Meese, Richard A., and Rogoff, Kenneth, 1983, Empirical Exchange Rate Models of the Seventies,

Journal of International Economics, 14, 3-24.

Rogoff, Kenneth and Obstfeld, Maurice, 2001, The Six Major Puzzles in International Macroeconomics: Is There a Common Cause? NBER Macroeconomics Annual 2000, 15, 200-224.

we Blame the Discount Factor? Journal of Money, Credit and Banking, 41, 437-442.

Schularick, Moritz and Taylor, Alan M., 2012, Credit Booms Gone Bust: Monetary Policy, Leverage Cycles, and Financial Crises, 1870-2008, American Economic Review,102(2): 1029-61. Werner, Richard A., 2012, Towards a New Research Programme on ‘Banking and the Economy’-