May 2000



May 2000

Exchange Rate Barometers

Giorgio Radaelli*

Bank One NA, London

1 Triton Square

London NW1 3FN

United Kingdom

(Phone: 0044 20 7903 4296, Fax: 0044 20 7309 9032)

(*) This paper was presented at the 21st International Symposium on Forecasting, Lisbon, June 2000. The author gratefully acknowledges helpful suggestions received by the Symposium’s participants. He is also indebted to Avi Asimov and Christine Foessel for comments on an earlier draft.

How to Rank Currencies’ Relative Value Through Econometrically Estimated Exchange Rate Barometers

Abstract

Exchange rates are difficult to model and forecast. Following some successes in the seventies, the Meese-Rogoff papers suggested the exchange rate is, basically, unpredictable. Fundamentals-based models even fail to beat a random-walk rule in forecasting. Come the 1990s, however, some studies purported to show that “this or that” model actually beats the random-walk rule. However, setting the latter as the yardstick is inappropriate since no serious fund manager would attach much value to a random walk, no-change kind of forecast. Moreover, most studies tend to test just one theory at a time without discriminating amongst competing models. In truth, no single theory does a good job at explaining/forecasting the exchange rate. A way forward may be using co-integration techniques in a “theoretically agnostic” way. The equilibrium models thus obtained can generate so-called “forex barometers” - indicators of “overbought/oversold”. These allow investors to rank currency risks and invest and hedge accordingly, without requiring detailed numerical forecasts.

Key Words: Exchange Rate Determination, Co-integration.

(1) Introduction and Conclusions

This paper is quite pragmatic in nature. It does not purport to offer a new, unified theory of exchange rate determination, partly reflecting my view that such a theory hardly exists. It also reflects my belief that it is nigh impossible to obtain models that produce worthwhile point-forecasts of exchange rates. That some researchers have recently been able to claim models that “beat a random walk rule” in forecasting is quite meaningless, in the sense that no respectable forex investor would rate a no-change forecast as a useful yardstick against which to judge competing theories/models. In sum, I think the old Meese-Rogoff wisdom that no model is really useful in forecasting exchange rates still stands.

Given the above, this paper suggests a simple methodology to rank currencies according to their deviations from equilibrium, or fair value, as defined by the fundamentals that matter. Although this may be reminiscent of the FEER methodology originally proposed by John Williamson (1985), the important departure is that my models do not reflect any particular theoretical priors. Whilst the models are fundamentals-based, the variables entering the equations are freely determined by a specific-to-general search based on co-integration tests.

The deviation of the observed market rate from the value “dictated” by the co-integrating model may be plotted against time. I name this the barometer. This can be seen as an indication of misalignment, leading to a currency ranking according to relative cheapness/dearness. Such models may be used by currency investors who need a view within which to frame hedging and financing decisions.

The next section is a brief excursus on the theoretical evolution of exchange rate models over the last three decades. Section 3 draws the parallel between the theoretical evolution and the econometric approaches that prevailed at different times. Section 4 details on the variables entering the models and the theoretical rationales/ambiguities involved. Finally, section 5 presents the econometric results on seven main exchange rates, while charting the market messages that the barometers convey.

(2) How We Got Here

Explaining and forecasting the behaviour of exchange rate is notoriously difficult. Unlike interest rates, exchange rates are less amenable to be controlled by governments (central banks) and this leaves them subject to the vagaries of numerous potential determining factors, both measurable and un-measurable. In this sense, exchange rates are more akin to share prices, another variable known for being hard to model. A difference between exchange rates and share prices is that the former has been more directly affected by the massive increase in economic openness, international trade volumes and globalisation in general. Also, significant shifts in exchange rates have more potential to affect the economic fortunes of this or that country than comparable changes in national stock market indices.

Modern theories of exchange rate determination started with the break up of the Bretton Woods fixed rate system in 1971. The predominance, back then, of currency flows related with current account transactions meant that initial theories were “derivatives” of previous theories of the balance of payments. This meant an emphasis on flows rather than stocks, i.e. on determining factors like price competitiveness and relative GDP growth. Financial prices were brought in via the back door by the so-called monetary approach, which would posit the exchange rate as being determined by relative money demand functions – hence relative GDP and relative interest rates. An extension of that was the so-called portfolio-balance approach, which broadened the realm of tradable assets from monies to bonds, thus giving more of a role to (difficult-to-model) exchange rate expectations and bond yields. Relative equity yields or prices were almost always left out of the, typically, single-equation models on the grounds that equities are not outside assets – i.e. they net out within one nation’s private sector (unlike government bonds).

US And Germany: Capital Flows Relative To Overall Balance Of Payments

As time progressed, however, globalisation trends (i.e., the reduction in trade barriers and capital controls) meant that capital account flows started growing relatively faster than current account transactions – see the chart above. Fundamentals-based predictions of exchange rates became hazardous. All the more so in a market which started churning out hundreds of USD billion worth of forex transactions daily (today, this statistic is close to some USD 1.5 tr.). Perhaps, no more than 10% of these were related to international trade in goods/services. The greater part of forex flows reflected, as they do today, hot money chasing (quick) profits from purely financial transactions – a part from foreign direct investments. The rapidity of such movements made fundamentals-based predictions more difficult, chiefly because measurable economic variables move more slowly than financial prices like the exchange rate. This favoured a de facto abandonment of PPP as a forecasting model, at least for major currencies over meaningful trading time horizons. Also, models that narrowly focused on the money markets, such as the monetary approach, fell rightly out of favour.

Disillusion started to creep in about economists’ ability to explain, let alone forecast, the exchange rate. This was exemplified by the proliferation of papers along the lines of Meese and Rogoff (1983, 1984) which purported to show how forecasting with standard exchange rate determination models would fare miserably out of sample, to the point of forecasting even more poorly than a simple random-walk rule. This said, every now and then there would a study claiming some success in beating the random walk, on a particular exchange rate for a particular period. Nevertheless, the basic fact remains that standard econometric models of the exchange rate are not very good at forecasting – this being true even if, occasionally, they beat a random walk forecasting rule.

As a result, from the mid eighties, indeed with the publication of Williamson (1985)[1], there has been a shift of emphasis, away from exchange rate forecasting towards models that assess exchange rate misalignments. These policy-related efforts allow one to say where a given exchange rate ought to be, not necessarily where it will go.

Another turn in the research agenda is the most recent literature on the causes of currency attacks, such as Mexico in 1994, Asia in 1997 and Russia in 1998. The several papers on the subject are reviewed and critically tested in Berg and Pattillo (1999) which concludes that the explanatory power of currency-crisis models is quite low. More recently, Caramazza, Ricci and Salgado (2000) broadened the analysis beyond the consideration of domestic fundamentals. Their effort allowed trade spillovers and financial linkages (such as the common creditor problem) to have a potential role in explaining why countries fall under speculative attack even if their fundamentals “look good”. However, this effort produced models that are not markedly better at forecasting currency crises. Moreover, one is still left with the key question of how to systematically forecast the major exchange rates. To date, this problem seems largely unsolved.

The theoretical developments sketched above were paralleled by developments in the econometric approaches, or methodologies, used in empirical work. This process is examined in the next section.

(3) Evolving Modelling Approaches

To appreciate my fundamentals-based technique, it is useful to review the main modelling strategies on exchange rates, as they have evolved over the last 25 years.

Old-style Regression Analysis. Prevailing in the 1970s and early 1980s, this consists of relating the exchange rate to a set of variables, usually suggested by a preferred theory of exchange rate determination, and using econometric techniques to quantify the parameters that shape the relationship between each variable and the exchange rate. Basically, economists would go for the set of variables (model) that maximises the explanatory power of the regression. (A similar selection criterion would characterise also so-called Box Jenkins models.) The problem with this strategy is that it is actually too easy to obtain very high R2, say between 0.90 and 0.99; but it comes at the cost of spurious correlation. This is usually the result of putting on the right hand side of the equation variables that have a temporary and/or tenuous relationship with the exchange rate. A typical consequence is that even a model with a very high R2 would forecast poorly. This is the classic symptom of over-fitting - there being too much “garbage” on the right hand side of the equation.

Fundamental Equilibrium Exchange Rate (FEER). This became fashionable from Williamson’s (1985) seminal work. The FEER approach would seek the exchange rate level consistent with a sustainable current account balance, i.e. one matched by “sustainable” capital (in)flows. In other words, the current account can be in the red to the extent allowed by structural capital inflows, which are meant to finance smoothly the current account deficit itself. FEER is a normative notion, a concept of more use to policy makers than forex investors, unless FEERs can be proven to be long-run “attractors” towards which the actual exchange rate reverses from time to time[2]. Although market economists have used FEERs in recent times, they were originally devised as an aid for policy makers wishing to manage their exchange rate around an appropriate level. The FEER methodology takes such a level as being the one that is competitive enough to deliver a “sustainable”, if not balanced, current account in the long run.

In practice, econometric estimation of FEERs involves starting with an equation for the current account, whose right hand side features the classic determinants: the real exchange rate (competitiveness effect) and relative GDP (income effect). Once the equation’s parameters are estimated, by regression methods, the current account is made to equal the “sustainable” capital account (to be judgementally determined). Such a final equation is then solved for the real exchange rate. This done, stripping the latter of its relative prices’ component leads to the FEER estimate: that nominal exchange rate consistent with a “well behaved” current account.

There are severe problems in using FEERs as guide to currency investments. The first is that it is very difficult to establish what is the “sustainable” level of capital (in)flows. In theory, this should be determined by structural factors like demographics, productivity levels and the fiscal position. In practice, estimating a sustainable capital account balance is a very subjective matter. This makes the FEER estimates themselves quite subjective, hence uncertain. Another problem is that, in practice, “unsustainable” current account balances can be observed for quite long periods. Eventually, the macroeconomic pressures may grow unbearable for the progressively indebted country and, hence, its exchange rate may jump back towards FEER-like levels. But the methodology says nothing about when, if at all, such a jump will take place. Finally, and this stems from the way FEERs are constructed, the final equations fail to include relative rates of interest (or return), something most investors would think of as pretty important to exchange rates, and not just in the short run.

Error-correction Models (ECMs). Once the operational limitations of FEERs were fully digested, applied economists latched onto that econometric development of the early eighties: error correction models. These represent a marked improvement from the old-fashioned regression analysis. The ECM tries to combine the appealing distinction between short and long run forces. For instance, with reference to exchange rate determination, it makes intuitive sense to think of a shift in the yield curve (proxy for interest rate expectations) as a short-term determinant, whereas the slow-moving current account would be a long-run determinant. ECMs are driven by a long-run equilibrium, e.g. the relationship between the current account (net foreign assets), or prices, and the exchange rate. Deviations from this equilibrium, thought as being short lived, are modelled separately by the short-run segment of the model. Once the two segments, short and long run, are combined, we obtain the full ECM. “Error correction” implies that the full model allows for temporary exchange rate deviations from long-run equilibrium (the error) which tend to be corrected (i.e. eliminated) in time.

The ECM is an improvement on older-fashioned models because it is consistent with the observation that exchange rates tend to be reverting, if not towards their mean, towards trends that may be driven by slow-moving economic variables. Moreover, since the ECM exploits superior model specification strategies, the old problem of having lots of “garbage” on the right hand side was reduced. In fact, the appropriate techniques would allow economists to progressively eliminate those ephemeral exchange rate determinants that simply boost R2s without introducing information that improves the model’s forecasts. However, the old problem remained that, with many variables entering the model, many onerous assumptions on the future values of such variables are required for the computer to generate exchange rate forecasts. Thus, not only can the forecast be wrong because the model is badly specified, but also because mistakes occur in making assumptions for the right hand side variables.

Behavioural Equilibrium Exchange Rates (BEERs). These models drop the pretence of generating precise forecasts for the spot rate. They retain something of the old FEERs, i.e. the notion that there is an equilibrium exchange rate that we can estimate, while borrowing something else from the ECMs - that there is a bounded gap between the spot and equilibrium exchange rate. However, instead of trying to “fill” that gap through modelling efforts, the gap is retained and appreciated for what it is: a dis-equilibrium. Our operational addition to BEERs[3] is simply the notion that, if the dis-equilibrium, or gap, between spot and equilibrium exchange rates is “significant”, it can be translated into buy/sell signals – hence the barometers.

The crucial difference between FEER and BEER is that, while the former sees equilibrium as inevitably tied to a well-behaved current account, the BEER does not impose this. The term “behavioural”, part of the acronym, suggests that equilibrium in the BEER relates to a relationship between the exchange rate and other variables that has proven to hold in the long run. It is relatively unimportant whether such a relationship is consistent, or not, with a balanced current account.

Calculating BEERs, and the barometers that we derive from them, crucially relies on the concept of co-integration and the techniques that allow us to verify when this condition exists[4]. Briefly, two or more variables are co-integrated when: (a) they share the same stochastic characteristics (typically, their levels are not stationary, but the differences are) and, (b) they share long-term “attractors”. Simply put, two co-integrated variables may diverge for short-lived periods, but not indefinitely, as there will be some economic force that eventually brings them closer together. Examples of co-integrated variables are the price of potatoes in neighbouring regions, income and consumption, short and long-term interest rates. It is intuitive that these couplets can diverge temporarily, but not indefinitely. There are, in fact, well defined economic forces (arbitrage and budgetary constraints) that tie them together in the long run.

In essence, what I do is the following. (i) To search for those variables that co-integrate with the exchange rate; (ii) to estimate the co-integrating relationship, and (iii) take the difference between today’s spot rate and what the co-integrating relationship suggests. This is the “behaviourally determined” equilibrium, or long-run, exchange rate. As long as co-integration is “strong”, any statistically significant deviation between the equilibrium level and the spot rate can be converted into a buy/sell signal.

(4) Equilibrium Exchange Rates and Barometers

In order to arrive at the exchange rate barometers, we have first to estimate our BEERs. Having already detailed on the econometric methodology that we adopted, let us now take a look at the fundamental variables deemed to have a statistically significant, long-run equilibrium relationship with the exchange rate. What we did was to consider as candidates all the variables suggested by the main theoretical models of exchange rate determination. Given the practical nature of this paper, we do not delve in reviewing the various theories on offer[5]. Instead, we present below the main variables that, at various times, theorists and practitioners have seen as likely determinants of the exchange rate. Our next step is then singling out which, amongst these several candidates, are the variables that co-integrate with the exchange rate.

Short-term interest rate. An increase in money rates at home (relative to abroad) can be expected to strengthen the home currency. This is because, following the home rate rise, uncovered interest parity implies an expected depreciation of the currency. For given expectations on the equilibrium exchange rate level, this requires that the spot rate should appreciate today.

Long-term interest rate. An increase in domestic bond yields, relative to abroad, has an ambiguous effect on the exchange rate. While some point to a similar positive effect to that from money rates, another theory suggests that, insofar as a rise in bond yields may reflect growing inflation expectations, it can lead to currency depreciation, via expected PPP. The opposite would likely hold if the rise in bond yields reflected an expansionary fiscal policy, as opposed to higher inflation expectations per se.

Prices. Purchasing Power parity (PPP) suggests that accelerating domestic prices, relative to abroad, should cause a currency depreciation because otherwise the home country tends to become less competitive in international trade. However, note that there is often a shorter run effect by which accelerating prices cause expectations of higher policy interest rates - the latter tending to boost the home currency. In my experience, it is hard to find empirical support for PPP as an investment rule, especially on the exchange rates of mature economies.

Real GDP. This has often been a key variable in many models of exchange rate determination, particularly in the monetary mould. A rise in domestic real GDP, relative to abroad, would cause an increase in demand for domestic money, thus favouring an appreciation of the home currency. This is true even in wider (portfolio-balance) models focusing on the supply of securities, not just money.

Money supply. The monetary model suggests that, the higher the domestic money supply relative to abroad, the weaker the home currency will tend to be. This is because the exchange rate should be the relative price of “one money against the other”. Note, however, that this model ceased to be effective from the early eighties, i.e. since the development of financial markets has led investors to trade not just bank deposits or CDs, but increasingly securities in different currencies. This leads us to the two main variables featuring the portfolio-balance model: budget and current account balances.

Budget deficit / Public debt. A bond-financed budget deficit should lead, at least in the shorter run, to currency appreciation via the concomitant increase in bond yields. However, applied research suggests this mechanism is applicable only to countries without an excessive level of public debt – say, below 70% of GDP. In fact, in high debt countries, a fiscal expansion leads to currency depreciation because of heightened (default) risk premia. (Note that the budget balance in a given period tends to equal the difference in public debt levels at the end and start of said period.)

Current account / Net foreign assets. Quite understandably, the stronger one country’s external balance, the stronger its exchange rate will tend to be. This is because the current account surplus should boost the supply of foreign currency relative to that of home currency. (Note that the current account balance in a given period tends to equal the difference in net foreign asset levels at the end and start of said period.)

Commodity prices. For countries richly endowed with natural resources, a rise in commodity prices translates into better terms of trade. Other things being equal, this causes a permanent improvement in the current account balance which, in turn, is likely to induce exchange rate appreciation. This was the experience of the UK with the sharp rise in GBP after the discovery of North Sea oil. It is a feature thought of being relevant for countries like Australia, Canada, Norway and South Africa, amongst others.

Unemployment rate. It has been suggested, e.g. by Whadhwani (1999), that a significant decline in the domestic rate of unemployment, relative to abroad, may boost the home currency. This would be on the premise that the labour market improvement sends signals about efficiency in the country’s supply side, thus favouring inward foreign direct investment. This seems plausible, although often a decline in unemployment would reflect a cyclical upswing, whose effect would already be captured by the real GDP variable mentioned above.

In my strategy of model specification, I took an agnostic view, without bias in favour, or against, this or that theory. Given that the statistical power of co-integration tests is not too high, I adopted a specific-to-general modelling strategy on all variables that proved to be I(1). This means that, for each currency considered, the first step was to find that one variable that co-integrates most strongly with the exchange rate. Having thus found a (near) co-integrating couplet, I then seek a third variable that, once added to the couplet, boosts the level of co-integration. Hence, the triplet is possibly expanded to a quadruplet, as long as doing so adds to the overall degree of co-integration in the exchange rate model. In practice, the peak in co-integration tends to be achieved by considering three or four variables (a higher number would be inadvisable anyway, because co-integration tests are less reliable in larger models).

This strategy led me to the BEERs, whose estimates are summarised in the table below, alongside the resulting forex barometers (see charts at the end). As already said, the barometer is simply the difference between the actual spot rate and the equilibrium one dictated by the BEER at each point in time. (I used quarterly data going back as far as possible, at least to the early 1980s. The data source was usually Datastream. Full data set and programs available on request.)

(5) The Results – Currency Misalignments

The charts at the end of this paper present the barometers on seven exchange rates. The barometers’ “dear/cheap” signals are compared with the actual historical exchange rate data, so the reader can assess how accurate these instruments have been in the last decade or so. Note that the barometers’ value tends to be stationary around the zero line. Deviations from such a line are deemed to be “significant”, in a statistical sense, whenever they approach, or break through, the two horizontal lines that limit a confidence interval. Thus, whenever the barometer’s value lies outside this corridor, the buy/sell signal should be taken seriously. Note that my notion of equilibrium is consistent with what the data actually tell, as opposed to normative views of how and why exchange rates ought to move (as in the FEER’s case).

For each of the seven exchange rates considered, the table above summarizes the determining variables and their impact as featured in the co-integrating, equilibrium models. There is one variable that quite consistently pops up as an equilibrium determinant of exchange rates. It is the (relative) bond yield - more often than not in nominal as opposed to real terms. Note that in all cases there is a positive relationship between the bond yield and the currency’s value, i.e. higher bond yields tend to cause, at equilibrium, stronger exchange rates. This validates the portfolio balance channel, by which assets are in greater demand when they are cheaper (yield higher), as opposed to the expected-PPP channel, by which higher bond yields reflect higher expected inflation (the latter being bad for the exchange rate). The only one equilibrium exchange rate that does not co-integrate with the relative bond yield is the USD/CAD, although it does respond to short-term interest rates. On average, there is a 0.041 semi-elasticity between the bond yield and the currency’s value. This means that a yield rise of one percentage point at home tends to cause, at equilibrium, a 4.1% rise in the home currency’s value. The qualification “at equilibrium” matters because it leaves open the possibility that short-term rises in bond yields be negative for the currency – as its bonds are seen to lose value. However, once the yield is seen to have reached a higher and stable level, this should induce an appreciation in the currency through higher demand for its (cheapened) bonds.

Relative GDP appears to be an equilibrium determinant only in two out of seven exchange rates. The causal relationship is in line with the monetary model, according to which a rise in home (relative) GDP is good for the home currency because it increases demand for home money. In truth, many market participants would likely say that GDP growth is more important than what our results seem to imply. At the end of the day, popular wisdom would have it that traders like buying into “growth currencies”. I generally agree with this, but can see how our results can be reconciled with popular wisdom. First, note that, as GDP growth perceptions improve, there is a tendency for interest rates (short and long term) to rise, and my models do give a role to interest rates in boosting the exchange rate. Secondly, perception of growth swings, as opposed to steady GDP growth, is probably what boosts currencies. Were this true, relative GDP (growth) would be more of a short-term determinant of changes in the exchange rate rather than a long-term determinant of the currency’s equilibrium level. Thus, one can see how my econometric findings do not downplay the role of economic growth in determining exchange rates. (For instance, note how relevant relative economic growth is for the EUR, corroborating anecdotal evidence that, between 1999 and 2000, the currency was sold because of Euroland’s deficient growth.)

Goods’ prices quite seldom turn out to be a determinant of equilibrium exchange rates. The CAD and SEK are the only two clearly affected by relative prices. This means there is little support for PPP-related factors in affecting exchange rates. This is hardly surprising, in my view, given that the largest part of the main currencies’ transactions are unrelated with trades in goods/services[6].

Commodity prices are, as we expected, determinants of the equilibrium level for the CAD and AUD exchange rates. This squares with popular wisdom that dubs the pair “commodity currencies”. Note, however, that the AUD is nearly twice as exposed to commodity prices as the CAD.

Outstanding public debt seems to have an impact on the JPY only. This may have to do with the recent dramatic rise in Japan’s gross debt but also with the fact, as the Japanese are the fastest ageing population in the G7, local economic agents are aware that higher debt today means less reliable pensions and welfare tomorrow (and/or higher taxes). This might induce the Japanese to shift capital abroad in an attempt to preserve their (expected) living standards. Generally, one would expect higher supply of public bonds to depress the exchange rate via a portfolio-balance channel. The extra bonds have to be sold, and one way to cheapen them (for multi-currency investors) is to lower their relative value via an exchange rate depreciation. Relative net foreign assets have a significant effect on the EUR, JPY, CHF and SEK exchange rates – the impact being positive as theory would suggest. Similarly, an improving current account balance tends to boost the equilibrium exchange rate of the AUD.

Following Whadhwani (1999), I tested for an unemployment effect on the UK exchange rate. In this case, I identified a positive link from the relative unemployment rate to the GBP exchange rate. This means there may be a signalling effect by which lower unemployment hints at supply-side flexibility and hence contributes to attract long-term capital.

Finally, I decided to add the EUR exchange rate as a possible determinant of the CHF and SEK’s equilibrium exchange rates. This is in recognition that, for a few years now, the authorities managed their currency vis a vis Euroland’s currency (DEM before, EUR now). This twist in the policy rule seems most applicable to the Swiss National Bank. Indeed, the model suggests that any given equilibrium change in the EUR acts as a constraint to the variability of the USD/CHF equilibrium rate, whilst I find no significant effect on the USD/SEK. The estimates suggest that some 57% of any given change in the EUR’s equilibrium value, against the USD, would be matched by a move of the CHF in the same direction – other things equal. This said, the remaining “free variability” of the CHF is influenced by Switzerland’s relative net foreign assets and bond yields.

At the time of writing (April 2000), the barometers were sending important misalignment signals that could inform medium-term currency exposures. First, the EUR was significantly undervalued against the USD, by no less than 15%. The EUR/USD barometer chart shows the oversold signal as being significant. Euroland’s currency was seen as cheap because it had lagged to react to improving relative GDP and relative bond yields. However, whilst this would argue for an imminent rise in the EUR, one should be opened to the possibility that a risk premium has developed on Euroland’s financial prices. This may have been due to inefficient management of the EUR by the too many bodies that apparently are in charge of it (the European Central Bank, Ecofin, national governments, the EC Commission). At the opposite end of the spectrum, the JPY seemed overvalued by a good 10% against the USD, hence ripe for a medium-term decline. One important reason why the JPY looked expensive was the ongoing significant rise in public debt relative to GDP - Japan’s ratio having grown from 76% in 1995 to some 115% in 2000.

EUR/USD Barometer And Actual Exchange Rate

USD/JPY Barometer And Actual Exchange Rate

USD/GBP Barometer And Actual Exchange Rate

USD/CHF Barometer And Actual Exchange Rate

USD/SEK Barometer And Actual Exchange Rate

USD/CAD Barometer And Actual Exchange Rate

AUD/USD Barometer And Actual Exchange Rate

References

Berg, Andrew and Catherine Pattillo, 1999, Predicting Currency Crises: The Indicators Approach and an Alternative, Journal of International Money and Finance, Vol.18, pp. 561-586.

Caramazza, Francesco, Ricci, Luca and Ranil Salgado, 2000, Trade and Financial Contagion in Currency Crises, IMF, Working Paper, March.

Clark, Peter b., and Ronald Mac Donald, 1998, Exchange Rates and Economic Fundamentals: A Methodological Comparison of BEERs and FEERs, IMF, Working Paper, May.

Engle, R.F. and Clive W.J. Granger, 1991, Long-run Economic Relationships: Readings in Co-integration, Oxford University Press.

Mac Donald, Ronald and Marc P. Taylor, 1992, Exchange Rate Economics: A Survey, IMF Staff Papers, Vol. 39, pp. 1-57.

Meese, Richard and Kenneth Rogoff, 1983, Empirical Exchange Rate Models of the Seventies: Do They Fit Out of Sample?, Journal of International Economics, Vol. 14, pp. 3-24.

---------, 1984, The Out-of-Sample Failure of Empirical Exchange Rate Models: Sampling Error or Mispecification, in “Exchange Rates and International Macroeconomics”, ed. by Jacob Frenkel (Chicago University Press), pp. 67-109.

Whadhwani, Sushil B., 1999, Currency Puzzles, unpublished, Bank of England.

Williamson, John, 1985, The Exchange Rate System, Policy Analyses in International Economics, Vol.5 (Washington: Institute for International Economics).

--------------, 1994, Estimating Equilibrium Exchange Rates, Washington, Institute for International Economics.

-----------------------

[1] See Williamson (1994) for a refinement.

[2] Empirical analyses, including our own, suggest this is not the case. The long-run fundamentals exchange rates tend to be anchored to are not quite those featured in the FEER approach.

[3] The acronym was coined by Clark and MacDonald (1998).

[4] Out of the many references on co-integration, we found handy Engle and Granger (1991).

[5] There are many surveys on offer. One we recommend, though not too recent, is McDonald and Taylor (1992).

[6] Note, however, that PPP tends to hold better for emerging markets’ exchange rates, or whenever inflation differentials are significant.

-----------------------

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

[pic]

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download