Exchange Rate Policy 2: Measurement

In Malaysia, it’s common to take the bilateral rate of the Ringgit (MYR) against the US Dollar (USD) as the exchange rate. This is understandable because not only does Malaysia have substantial trade links with the USA, the USD is important as a reserve currency and as the currency of international trade contracts. However, the USD bilateral rate is not the sole influence on MYR movements – other countries like Singapore also have substantial trade and capital links. How then to capture the multiplicity of influences on a currency? What we need is a single composite measure of a currency, an index of movement.

Let’s take the example of Malaysian trade with Aruba. Since this trade is pretty much non-existent, the exchange rate between the Ringgit (MYR) and the Arubian guilder (AWG) shouldn’t matter to Malaysia very much. From this example we can take the basic principle for constructing a composite view of the Ringgit or any other currency – taking the trade share* as the basis of a weighting scheme. Obviously, the greater the trade share, the more important a particular currency should be. The general rule is that weights don’t need to be changed very often (the IMF changes weights every ten years or so), although if a country’s trade patterns change substantially you can get gross errors in the index without knowing it.

* A more complicated scheme also uses the concept of third-country export competition as part of the weights. The IMF and Federal Reserve incorporate this in their currency indexes, but most do not. New Zealand uses share of global GDP as a proxy.

There is also a choice to be made at this point – which currencies to put in the “basket”. The more currencies involved the more work is required in gathering data and calculations, but the more accurate the measurement. The Federal Reserve uses a 0.5% export or import share threshold for their broad index, while the BOJ uses 1% overall trade share. I favour using 1% of exports or imports as giving a broad enough selection of currencies, without losing too much information – 15 currencies altogether based on 2007 Malaysian trade data. This also nicely allows me to avoid tracking Middle East countries, where data holes are a serious problem.

Having gathered the required trade and exchange rate data, we now have enough to construct a nominal effective exchange rate index. This is what MYR has been doing for last eight years (2000=100):



Ideally of course, real prices (net of inflation) are more important, so inflation data has to be incorporated in our measurement. This is complicated by having multiple types of inflation – consumer inflation (Consumer Price Index), producer inflation (Producer Price Index or Wholesale Price Index), and whole economy inflation (GDP deflator). Since trade competitiveness is the basis of our exchange rate measurement, PPI** is probably the best series to use (CPI and the GDP Deflator include non-tradable goods inflation). Unfortunately, PPI is not often available particularly in emerging markets, so CPI is often used as a proxy. In practice, most researchers have found that the price series used doesn’t really matter.

** The OECD use changes in real manufacturing wages

Deflating each currency by their respective inflation measures, we arrive at the real effective exchange rate index, here contrasted with the nominal index:



There's little to choose from between the two, which suggests the MYR is tracking pretty well with the currencies of our trade partners.

So now we have a pretty accurate idea of MYR movements holistically. Unfortunately, this still doesn’t tell our poor central banker very much, because all it shows is the movements of the currency relative to any other specific period - it says nothing of where the currency should be. While it’s better than nothing, he still has no idea of the valuation of the currency relative to all the others. This is where econometric models come in, which I'll cover in my next post.

However, even using the nominal and real charts can tell us something. First, here's a graph of trade shares (exports and imports), relative to the trade within the currency basket I'm using:



Note the massive decline in trade share to and from the US and the corresponding rise of China. Next, the nominal and real exchange rates against the G3 currencies:







The general interpretation is that if there is a difference in inflation experience between two currencies, there will be pressure for the nominal rate to close on the real rate. By that standard, MYR is about 17% overvalued against the JPY, but close to even with USD and EUR. Not surprising since Bank of Japan intervention to prevent JPY appreciation is common. This second batch of currencies also shows relatively high misalignments relative to the real rate:







The MYR appears overvalued against the HKD and TWD, but undervalued against the KRW, on about the same order as the misalignment against the JPY.

This last batch is so ridiculous, I had to triple check my calculations (please note the respective scales):









The differences here range from 33% against the PHP and IDR, to 87% (!!!) against the INR. These are serious misalignments that can only be explained by highly restricted capital accounts or heavy intervention (definitely true for the Reserve Bank of India). Given that the MYR is relatively close to “fair value” against the G3 currencies, it follows that the four currencies above are seriously out of whack with the rest of the world too.

Bear in mind, however, that I’m using inflation as the only currency alignment metric here – other forces can be at work as well that may explain these gross divergences.

Either that or you’re looking at the next good candidates for speculative attacks. I suspect the Rupee especially, might be in for some rough times ahead.

Update:

I should point out that because these are index numbers, measuring misalignments can be problematical because the index year has an impact on the perceived divergence. If I had taken 2005 as the base year for example, the misalignments would appear considerably smaller. The underlying assumption behind the analysis I'm using here is that currencies were largely in alignment in 2000. This weakness doesn't apply to econometric models.

Also forgot the acknowledgements:

1. (Free) foreign exchange data from the PACIFIC Exchange Rate Service. Forex data used are monthly averages.
2. CPI data from IMF International Financial Statistics (IFS) and the International Labour Organization, supplemented by national sources where necessary.
3. Trade data is from DOS and various issues of BNM Monthly Statistical Bulletins.