Menzie Chinn (Professor of Public Affairs and Economics at the University of Wisconsin, Madison) has a nice roundup of exchange rate valuation approaches, with the pros and cons of each. Worth a read if you’re interested in the subject (as I am).
Showing posts with label Macroeconomic Balance Model. Show all posts
Showing posts with label Macroeconomic Balance Model. Show all posts
Thursday, March 25, 2010
Tuesday, March 9, 2010
Currencies and Current Account Adjustments Part II
I did a post about six months ago on an article in VoxEU (link) that evaluated exchange rate imbalances between the USD and Asian currencies, which suggested that the MYR was as much as 1/3 too low against the USD. I criticised the paper on both methodological and procedural grounds.
Now another article on VoxEU is also basically challenging the findings of that article (excerpts):
On the renminbi and economic convergence
“Many economists agree that the build-up and maintenance of international imbalances, with their accompanying capital flows, contributed to the overleveraging of finance and underpricing of risk. How to rebalance then? Many observers are increasingly emphasising that China should let its exchange rate appreciate.
For example, Cline and Williamson (2009) have recently estimated “fundamental equilibrium exchange rates” compatible with moderating external imbalances. They estimate that the required renminbi appreciation is more than 20% in real effective terms and 40% relative to the dollar. Ferguson and Schularick (2009) point to the manufacturing wage unit-costs to estimate the degree of undervaluation of the renminbi relative to the dollar and come up with the figure of 30% and 50%. Finally, the Bank of China’s continuous intervention in the foreign exchange market also suggests that the renminbi would appreciate significantly if let loose; this intervention has accumulated $2.3 trillion of foreign exchange reserves.
…
To be sure, poor-country currencies are normally undervalued in terms of purchasing power parity with rich countries. In fact, poorer countries do have undervalued exchange rates (due to the Balassa-Samuelson effect), and convergence will imply considerable correction of that undervaluation. Services (and wages) are cheap in poor countries and expensive in rich countries, while prices for internationally traded goods are roughly equalised in a common currency. When the productivity in traded goods rises (while productivity growth for haircuts and other services are very limited), more income is generated and spent on services. The price ratio of non-traded to traded goods will rise. In other words, the real exchange rate will appreciate. Hence, part of the undervaluation ascribed to China’s and other currencies results from market forces that make non-traded goods relatively cheap in poor countries, rather than from deliberate currency manipulation by China’s authorities.
While growing and converging fast, China is still poor. Its per capita income in 2008 was 6.2% of the US’s at market rates and 12.8% at PPP-adjusted rates, according to World Development Indicator data. Figure 1 relates the log of real per capita GDP as a fraction of the US level and the deviations of current market exchange rates per US dollar from PPP rates for the year 2008. It shows strong support for the Balassa-Samuelson effect and suggests a well-determined elasticity (0.2) by which the undervaluation of the currency will be eroded during the catch-up toward the US per capita income level. Real exchange rates can thus be expected to appreciate as economies grow, approaching PPP exchange rates as economies converge with US living standards, as posited by the Balassa-Samuelson effect.
Figure 1. Income convergence and exchange rates appreciation
To gauge a converging country’s degree of undervaluation, the appropriate yardstick cannot be purchasing power parity; it should rather be the regression (over 145 countries) that provides the best fit for the Balassa-Samuelson effect. While the renminbi was undervalued by 60% in PPP terms, it was merely undervalued by 12%, if the regression fitted value for China’s per capita income level is compared to the current value in 2008. Note that India and South Africa (which had a current account deficit) were more undervalued than China by that Balassa-Samuelson benchmark, by 16% and 20%, respectively, in 2008. The currencies of Brazil and Russia were appropriately valued, i.e. close to the regression line.”
My kind of guy! Have a read through - there's some interesting policy conclusions as well.Friday, June 19, 2009
Currencies and Current Account Adjustments
A recent article by Cline & Williamson (2009)* on VoxEU.org investigates the state of play in terms of global currency misalignments, particularly against the USD. The results aren’t too surprising:
“The main counterpart to the overvalued dollar is the undervaluation of the Chinese renminbi, along with a few of the smaller Asian currencies…Our analysis is one more piece of evidence that the major macroeconomic imbalance in the world today stems from China’s exchange-rate policy.”
For the MYR, Cline and Williamson suggest an undervaluation of 17.7% in the real effective exchange rate, and 33.2% against the USD, with a medium term target rate of RM2.63. That’s a substantial movement for MYR, and will have a pretty massive impact on Malaysia’s external demand as well as the current account.
The modeling framework takes its cue from Williamson’s earlier work, which substantially helped launch non-purchasing power parity (PPP) based assessments of exchange rate valuations more than twenty years ago.
Some background is in order here. I’ve posted on these alternative models before, but the model used in this article is a variant of what’s called the Fundamental Equilibrium Exchange Rate (FEER) approach (aka Macroeconomic Balance approach), which uses a medium term current account target as a measure of a currency’s misalignment. Using an elasticity model, the extent of under or over valuation can then be calculated as the movement in the exchange rate required to bring the current account from its forecast level to the model’s target level over the medium term.
As you can imagine, I’m going to pick a few holes in this argument:
1. FEER models have a normative component – the target rate is selected by the researcher (admittedly based on global historical norms), rather than that inherent to each economy. In other words, the extent of misalignment derives directly from what the researcher considers a “sustainable” current account surplus/deficit. For instance, Williamson’s early work used a 2% band rather than the 3% used in the article - a tighter band implies a greater required adjustment. Since the current account covers both trade in goods and services as well as income flows, applying one number to all countries (or even one country at different points of time) isn’t obviously logical. The article attempts to account for this by incorporating a net foreign assets measure where required (essentially replacing a flow variable with a long-term stock variable), but the same critique applies.
2. Secondly, FEER models assume that all adjustments are made solely through exchange rates, which of course isn’t necessarily true. Demand and supply shocks, terms of trade shocks, secular changes in consumer preferences, productivity improvements, changes in portfolio holdings – all can have an impact on the current account without necessarily impacting the exchange rate at all.
3. FEER models don’t incorporate dynamics, which describes the interrelationship between the model variables across time. In short, you know your destination but you have no idea how to get there.
Lastly, I have a procedural criticism – the REERs used as reference points for the article are taken from the IMF International Financial Statistics Database. To my knowledge, the trade weights on these were last changed in 2006 based on averaged trade data from 1999-2001 (and thus accounting for the introduction of the Euro).
Ordinarily, I’d have no problem with this as trade weights rarely evolve much in any given 5-10 year span.
In this case however, I think the IMF REER indexes from about 2003 onwards are flawed – the emergence of China has had such far reaching effects on trade patterns that substantial changes in trade weights are warranted. That in turn implies that movements in the IMF REER indexes are too biased towards the G3 currencies, and not enough to emerging markets.
The effective difference doesn’t amount to a lot in an absolute sense – a few points at most on the index scale – but those few points do matter in terms of determining the threshold for possible policy action, and would matter even more if emerging market currencies were more volatile against the USD. In this case, the difference tends to support the article’s primary thesis of USD overvaluation against many Asian currencies.
To illustrate, here’s how the trade weights for the top 5 currencies in my own calculated MYR indexes have evolved during the same period:
Note that there are three different groupings: the EUR has been relatively stable; JPY, SGD and especially USD have been declining; and CNY has been steadily rising, with a big jump in 2003. The IMF static weights for these currencies for MYR are 14%, 15%, 6% (lumping SGD with all other ASEAN currencies), 24% and 5%; the latest weights for mine are 11.2%, 13.5%, 13.2%, 12.8%, and 13.7%. So there have been some pretty big changes in terms of which currencies are more important within a multilateral trade framework.
To illustrate the difference, here’s a comparison of the different indexes:
So, do I believe that MYR is that much undervalued? No, for a few reasons, not least of which are the flaws in the modeling framework I’ve pointed out above. But since this post is getting over long, I’ll save that for later.
*"Equilibrium exchange rates", William R. Cline & John Williamson
“The main counterpart to the overvalued dollar is the undervaluation of the Chinese renminbi, along with a few of the smaller Asian currencies…Our analysis is one more piece of evidence that the major macroeconomic imbalance in the world today stems from China’s exchange-rate policy.”
For the MYR, Cline and Williamson suggest an undervaluation of 17.7% in the real effective exchange rate, and 33.2% against the USD, with a medium term target rate of RM2.63. That’s a substantial movement for MYR, and will have a pretty massive impact on Malaysia’s external demand as well as the current account.
The modeling framework takes its cue from Williamson’s earlier work, which substantially helped launch non-purchasing power parity (PPP) based assessments of exchange rate valuations more than twenty years ago.
Some background is in order here. I’ve posted on these alternative models before, but the model used in this article is a variant of what’s called the Fundamental Equilibrium Exchange Rate (FEER) approach (aka Macroeconomic Balance approach), which uses a medium term current account target as a measure of a currency’s misalignment. Using an elasticity model, the extent of under or over valuation can then be calculated as the movement in the exchange rate required to bring the current account from its forecast level to the model’s target level over the medium term.
As you can imagine, I’m going to pick a few holes in this argument:
1. FEER models have a normative component – the target rate is selected by the researcher (admittedly based on global historical norms), rather than that inherent to each economy. In other words, the extent of misalignment derives directly from what the researcher considers a “sustainable” current account surplus/deficit. For instance, Williamson’s early work used a 2% band rather than the 3% used in the article - a tighter band implies a greater required adjustment. Since the current account covers both trade in goods and services as well as income flows, applying one number to all countries (or even one country at different points of time) isn’t obviously logical. The article attempts to account for this by incorporating a net foreign assets measure where required (essentially replacing a flow variable with a long-term stock variable), but the same critique applies.
2. Secondly, FEER models assume that all adjustments are made solely through exchange rates, which of course isn’t necessarily true. Demand and supply shocks, terms of trade shocks, secular changes in consumer preferences, productivity improvements, changes in portfolio holdings – all can have an impact on the current account without necessarily impacting the exchange rate at all.
3. FEER models don’t incorporate dynamics, which describes the interrelationship between the model variables across time. In short, you know your destination but you have no idea how to get there.
Lastly, I have a procedural criticism – the REERs used as reference points for the article are taken from the IMF International Financial Statistics Database. To my knowledge, the trade weights on these were last changed in 2006 based on averaged trade data from 1999-2001 (and thus accounting for the introduction of the Euro).
Ordinarily, I’d have no problem with this as trade weights rarely evolve much in any given 5-10 year span.
In this case however, I think the IMF REER indexes from about 2003 onwards are flawed – the emergence of China has had such far reaching effects on trade patterns that substantial changes in trade weights are warranted. That in turn implies that movements in the IMF REER indexes are too biased towards the G3 currencies, and not enough to emerging markets.
The effective difference doesn’t amount to a lot in an absolute sense – a few points at most on the index scale – but those few points do matter in terms of determining the threshold for possible policy action, and would matter even more if emerging market currencies were more volatile against the USD. In this case, the difference tends to support the article’s primary thesis of USD overvaluation against many Asian currencies.
To illustrate, here’s how the trade weights for the top 5 currencies in my own calculated MYR indexes have evolved during the same period:
Note that there are three different groupings: the EUR has been relatively stable; JPY, SGD and especially USD have been declining; and CNY has been steadily rising, with a big jump in 2003. The IMF static weights for these currencies for MYR are 14%, 15%, 6% (lumping SGD with all other ASEAN currencies), 24% and 5%; the latest weights for mine are 11.2%, 13.5%, 13.2%, 12.8%, and 13.7%. So there have been some pretty big changes in terms of which currencies are more important within a multilateral trade framework.
To illustrate the difference, here’s a comparison of the different indexes:
So, do I believe that MYR is that much undervalued? No, for a few reasons, not least of which are the flaws in the modeling framework I’ve pointed out above. But since this post is getting over long, I’ll save that for later.
*"Equilibrium exchange rates", William R. Cline & John Williamson
Thursday, March 26, 2009
Exchange Rate Policy 3: Equilibrium exchange rates
How does one tell whether an exchange rate is over- or under-valued? Driver and Westway (2004)* lists 11 different methodologies on how to do this. If you recall from my first exchange rate policy post, structural models have the potential for making this determination. The insight into using these types of models is that exchange rates are affected by many factors, not just one or two, i.e. currencies should move in relation to the underlying fundamentals of an economy. If this sounds familiar, it’s because this is a common refrain of Tan Sri Zeti when talking about the level of the MYR. By implication, the fundamental equilibrium view rejects the concept of purchasing power parity or PPP, or alternatively that the PPP changes across time (depending on which model you use).
*Rebecca L Driver & Peter F Westaway, "Concepts of equilibrium exchange rates", Working Paper no. 248, Bank of England (2004). Link is here - warning! PDF link.
Which fundamentals are important to which currency can be radically different – a lot of the time spent in specifying these models is determining which variables actually matter. For instance, for countries like Malaysia with strong natural resources, commodity prices are an obvious starting point, although there are many more factors to take into account. A non-exhaustive list of fundamentals would include:
1. Government consumption, which is presumed to fall more on non-tradables. Higher government consumption in that case should in theory (through the Balassa-Samuelson Hypothesis) cause an appreciation of the exchange rate;
2. Relative real interest rates, where higher rates cause appreciation;
3. Openness to trade, where higher openness allows for greater tradable goods arbitration. This will equate to a depreciation of the exchange rate;
4. Net foreign assets, which measures capital stocks. A typical approach is to take changes in stocks as the variable i.e. a flow approach using changes to the current account, or the international investment position. The impact depends on the policy stance and structure of the economy – where economic growth is slower, a higher net foreign asset position implies greater income flows which result in an appreciation. However, empirical evidence suggests in high growth periods or for fast growing developing economies, capital inflows can appreciate the currency despite decreasing the net foreign asset position;
5. The terms of trade, which measures the price of exports in terms of imports. The result on the exchange rate here depends on the relative strength of income and substitution effects, although the empirical evidence suggests the former. This implies an appreciation of the exchange rate;
6. Relative productivity, both internal (between the non-tradable and tradable sectors) and external (in tradables). Higher international productivity in the tradable sector suggests a depreciation, while higher productivity in the tradable sector relative to the non-tradable sector implies an appreciation;
7. Demographic structure, such as the dependency ratio.
I have seen a lot of variations and proxy alternates in the variables used, some due to specific country effects and others due to data limitations. Some of the variables, such as trade openness, are subjective. Government consumption sometimes is not significant, but the government deficit might if borrowing is primarily external.
The different modeling approaches also yield different estimates of the equilibrium value of the exchange rate. There is in fact no single, correct way to go about this. The consensus is to always apply two or three different modeling approaches, which should give a good ballpark figure as to how far a currency is off its “true” equilibrium value. If on the other hand all the models are pointing in one direction, then that is something our fictional central banker has to take seriously. In terms of actual use, I would take the following three models as the most prevalent:
1. Macroeconomic Balance model (MBM) – measures the difference between projected medium term current account balance with an estimated equilibrium current account balance.
2. External sustainability model (ESM) – a variant of the MBM model, but measures the difference between actual current account balances with the balance that would stabilise the net foreign asset position at some benchmark.
3. The Reduced Form Structural Model – equates the medium term equilibrium exchange rate as a direct function of medium term fundamentals. The term "reduced form" indicates that variables that don't impact the exchange rate are dropped from the specification i.e. these aren't "full" structural economic models.
These are the models in general use by the IMF* in assessing currency misalignments, and where possible all three are calculated to get a balanced view of a currency’s equilibrium position. I’m not about to lay out in any length of how the models are calculated – even within the broad categories above, you can use different statistical approaches in the estimation, and results may vary according to the sample period chosen as well as the presence of any structural breaks.
* Press statement here, PDF document here
Neither is data gathering a trivial exercise. If you take the 15 currencies I’m using in my short term broad MYR index, that means you have to have the required data for all the variables for all the currencies involved, and in the correct frequencies for the full sample period selected. Secondly, some of the theoretical concepts don’t translate well to real world data – such as for example trade openness – which requires using some form of proxy. Third, some data is just not collected, or available for limited periods or not accurately measurable, such as net foreign assets. There’s thus a lot of room for specification error and measurement error.
Even with those caveats, there’s still no alternative to making the attempt at measuring currency misalignments. Any currency that is not following a full free float with open capital account regime can get into serious misalignment problems, with potentially expensive adjustments required to regain equilibrium either through currency adjustments, real economy adjustments or both. The worse case of course, is when these adjustments are imposed by the market, as we saw in 1997-98.
This naturally takes us to the choice of exchange rate regime, which will be covered by the next post.
*Rebecca L Driver & Peter F Westaway, "Concepts of equilibrium exchange rates", Working Paper no. 248, Bank of England (2004). Link is here - warning! PDF link.
Which fundamentals are important to which currency can be radically different – a lot of the time spent in specifying these models is determining which variables actually matter. For instance, for countries like Malaysia with strong natural resources, commodity prices are an obvious starting point, although there are many more factors to take into account. A non-exhaustive list of fundamentals would include:
1. Government consumption, which is presumed to fall more on non-tradables. Higher government consumption in that case should in theory (through the Balassa-Samuelson Hypothesis) cause an appreciation of the exchange rate;
2. Relative real interest rates, where higher rates cause appreciation;
3. Openness to trade, where higher openness allows for greater tradable goods arbitration. This will equate to a depreciation of the exchange rate;
4. Net foreign assets, which measures capital stocks. A typical approach is to take changes in stocks as the variable i.e. a flow approach using changes to the current account, or the international investment position. The impact depends on the policy stance and structure of the economy – where economic growth is slower, a higher net foreign asset position implies greater income flows which result in an appreciation. However, empirical evidence suggests in high growth periods or for fast growing developing economies, capital inflows can appreciate the currency despite decreasing the net foreign asset position;
5. The terms of trade, which measures the price of exports in terms of imports. The result on the exchange rate here depends on the relative strength of income and substitution effects, although the empirical evidence suggests the former. This implies an appreciation of the exchange rate;
6. Relative productivity, both internal (between the non-tradable and tradable sectors) and external (in tradables). Higher international productivity in the tradable sector suggests a depreciation, while higher productivity in the tradable sector relative to the non-tradable sector implies an appreciation;
7. Demographic structure, such as the dependency ratio.
I have seen a lot of variations and proxy alternates in the variables used, some due to specific country effects and others due to data limitations. Some of the variables, such as trade openness, are subjective. Government consumption sometimes is not significant, but the government deficit might if borrowing is primarily external.
The different modeling approaches also yield different estimates of the equilibrium value of the exchange rate. There is in fact no single, correct way to go about this. The consensus is to always apply two or three different modeling approaches, which should give a good ballpark figure as to how far a currency is off its “true” equilibrium value. If on the other hand all the models are pointing in one direction, then that is something our fictional central banker has to take seriously. In terms of actual use, I would take the following three models as the most prevalent:
1. Macroeconomic Balance model (MBM) – measures the difference between projected medium term current account balance with an estimated equilibrium current account balance.
2. External sustainability model (ESM) – a variant of the MBM model, but measures the difference between actual current account balances with the balance that would stabilise the net foreign asset position at some benchmark.
3. The Reduced Form Structural Model – equates the medium term equilibrium exchange rate as a direct function of medium term fundamentals. The term "reduced form" indicates that variables that don't impact the exchange rate are dropped from the specification i.e. these aren't "full" structural economic models.
These are the models in general use by the IMF* in assessing currency misalignments, and where possible all three are calculated to get a balanced view of a currency’s equilibrium position. I’m not about to lay out in any length of how the models are calculated – even within the broad categories above, you can use different statistical approaches in the estimation, and results may vary according to the sample period chosen as well as the presence of any structural breaks.
* Press statement here, PDF document here
Neither is data gathering a trivial exercise. If you take the 15 currencies I’m using in my short term broad MYR index, that means you have to have the required data for all the variables for all the currencies involved, and in the correct frequencies for the full sample period selected. Secondly, some of the theoretical concepts don’t translate well to real world data – such as for example trade openness – which requires using some form of proxy. Third, some data is just not collected, or available for limited periods or not accurately measurable, such as net foreign assets. There’s thus a lot of room for specification error and measurement error.
Even with those caveats, there’s still no alternative to making the attempt at measuring currency misalignments. Any currency that is not following a full free float with open capital account regime can get into serious misalignment problems, with potentially expensive adjustments required to regain equilibrium either through currency adjustments, real economy adjustments or both. The worse case of course, is when these adjustments are imposed by the market, as we saw in 1997-98.
This naturally takes us to the choice of exchange rate regime, which will be covered by the next post.
Subscribe to:
Posts (Atom)
Posts
