Tuesday, March 24, 2009

Exchange Rate Policy 1: Concepts

If it seems I’ve taken a break from blogging, it’s because I’ve been hard at work reconstructing my MYR exchange rate indexes. Since I’m still in the midst of refining them, this post should serve as a holdover until that’s done.

Nothing in economics is quite so confusing as exchange rates. The problem is that exchange rates are in essence relative prices, not absolute prices, and because in a world of multiple exchange rate regimes, making sense of exchange rate movements can be complicated. For instance the chart below shows the cumulative movements of the Ringgit against the G3 currencies since 2000 (2000=100):

Relative to the start date, Ringgit (MYR) is 6.9% stronger against Dollar (USD), 9.5% weaker against Yen (JPY), and 26.9% weaker against the Euro (EUR). Can we therefore say that the Ringgit is generally weaker or stronger? More to the point, what are the implications for policy and trade? At what stage should authorities decide that a currency is overvalued or undervalued enough to merit intervention? These and other questions will be the topic of this series of posts I’m embarking on.

The most intuitive and attractive concept of exchange rates is the theory of purchasing power parity, or PPP. In its strong form, PPP states that the same good should have the same price everywhere (aka The Law of One Price). Therefore the exchange rate of any two currencies should equate this price in local currency terms. For example if a Big Mac is RM8 in Malaysia, and US$2 in the USA, then the PPP exchange rate should be US$0.25 if we use MYR as the base and RM4.00 if we use USD as the base. If PPP is true, and the actual exchange rate is RM3.80, then we could consider the MYR as being 5% overvalued against the USD.

Unfortunately, empirical evidence for strong form PPP is patchy even over periods of hundreds of years. There is better evidence for weak-form PPP, where prices differ but the difference is constant. Even here, the proof is not absolute as exchange rates appear to deviate from the PPP predicted value for extended periods. The last 60 years has seen a lot of effort to decipher how this can be, given the importance of the exchange rate to international trade and the balance of payments.

One notion, called the Balassa-Samuelson effect, suggests that only tradable goods face price arbitrage in international markets and PPP should hold for these goods, but not for non-tradable goods and non-tradable components such as tax structures, property rentals, and so on. This effect should be particularly strong in services, which by definition are difficult to trade (imagine the relative value of the haircut for instance). The general effect of a strong non-tradable sector is an appreciation of the exchange rate.

Another school of thought focused purely on monetary effects and the impact of real interest rate differentials, called the concept of uncovered interest parity. The idea is that capital flows to where the real yield is highest, which in turn moves the exchange rate to equate the relative differences. Since investors view countries as having different risk profiles, a risk premium can also affect the exchange rate, which is the concept of covered interest rate parity. Inflation should also have an impact, as it is a measure of the relative supply of each currency – the higher the relative inflation rate, the more a currency is expected to depreciate.

The problem with these and other theories is that, taken in isolation, none do a good job of explaining exchange rate movements in the floating rate period (1972 onwards). The best statistical model of short term exchange rate movements is and remains the random walk, which in statistical notation is:

X(t) = X(t-1) + ε

Notice that this differs substantially from a simple regression model in that there is no intercept and no slope to the equation. What the random walk model states is that the best predictor of tomorrow’s market price is today’s price!

So what is a poor central banker to do? What finally made sense, at least in identifying medium to long term movements of the exchange rates, were structural models which put a lot of these concepts together. In real life therefore, it’s a combination of factors that influence the exchange rate, not one or two. But before models like these can be used, we have to figure out how to actually arrive at a single view of a currency rather than the multiplicity of rates that any currency is subject to, which will be the subject of the next post.


  1. Hi HishamH,

    Are the CPI of different countries composed of different baskets of goods and calculated based on different assumptions ?
    How does an economist handle these differences when they want to check on purchasing power parity (either weak form or strong form) ?

    Feel free to reply only when you are free.
    Don't wish to disturb your flow of writing.. :)

  2. 1. Yes;
    2. They don't

    Actually, the problem is worse than you think - CPI baskets within countries also change over time. Malaysia's has changed three times in the last decade. IPI is even worse.

    That said, there's really not much alternative. PPI measures are generally harder to get and suffer the same mesurement problems, and GDP deflators are at best estimates (while also including export inflation).

    This is one of the main issues with economic statistics - there is no way to ever get an apples to apples comparison with data, even between advanced economies. In emerging markets the problem is worse. China for instance calculates CPI on a year-to-year basis, not month-to-month like everyone else.

    On the subject of PPP, the actual methodology to prove it exists is ridiculously easy - all you have to do is test for a unit root. If the real exchange rate is a stationary series, then PPP holds. Takes about 1 sec once your exchange rate index is compiled.

  3. Thx HishamH.

    To prove PPP, using the real exchange rate and test for unit root,
    is it true that we are actually assuming price level is apple-to-apple comparison ?

  4. Yes it does. That's one reason I'm saying that general evidence for PPP is inconclusive - there's a degree of noise involved in measurement that can bias a sample. However, using different price deflators doesn't appear to have a significant impact on the results, which suggests that CPI measures are "good enough".

    Secondly, from a micro prespective, having an apple-to-apple comparison doesn't make sense as consumer preferences will vary according to country and across time. Since we're interested in the general price level rather than specific goods, different baskets of goods make sense even if they are not a strictly comparable.

    There is also "small sample bias" in all unit root tests, which means that there is always a degree of uncertainty involved. In other words, a small sample (say 50 years), can appear to be non-stationary while a much larger one (500 years) can appear to be stationary.

    Have a read of this paper for more detailed arguments:

    Kenneth A Froot, Kenneth Rogoff, Michael Kim, "The Law of One Price Over 700 Years", IMF Working Paper WP/01/174, (Nov 2001)

    Link is here.