Monday, June 22, 2009

Exchange Rates and Purchasing Power Parity: Why PPP Only (Sort Of) Works

I made a couple of assertions in my previous post that deserve some clarification.

First, I said taking some absolute level of the exchange rate as the equilibrium level, and thus making it the reference point at which you can judge whether a particular exchange rate is “weak” or “strong”, only works if you believe purchasing power parity (PPP) actually holds. Since PPP in its absolute form does not in fact empirically hold, this sort of analysis is inherently flawed. Second, I gave the opinion that since Malaysia appears to be shifting to a deliberate policy of promoting the services sector as the future engine of growth, the exchange rate must necessarily appreciate. These two statements are in fact intimately related, and actually come from the same analytical framework.

To begin, let’s examine the underlying principles of PPP. Take a world with two economic actors and two different currencies– a small open economy (A) and the Rest of the World (B). Only one good is produced and consumed (called W), and that good is globally traded with no friction (i.e. no transport costs, tariffs or taxes). The price of the good is then determined by global supply and demand, and applies to everyone equally (since A is small and thus a price-taker). There are only two factors of production, capital and labour, with constant returns to scale technology and where capital is freely mobile but labour is not. The exchange rate between currencies A and B will simply be the relative price of good W in A and B – in other words, the exchange rate equates the local price of W in both economy A and economy B. That in essence is what PPP says – prices of goods should be equal anywhere in the world after adjusting for the exchange rate. Extend that to the factors of production and in equilibrium, wages and interest rates should also be equal.

Adding complications to this model doesn’t change the underlying principle – trade friction and inflation only adjust the exchange rate required to equalize prices. You can even add stuff like foreign risk premiums or productivity differentials, but nothing fundamental will change – prices should still be equal after these additional costs are taken into account. You can see why PPP is both a powerful and seductive concept: it makes sense, and appeals to people’s sense of symmetry and fairness.

If structural and policy changes to economies happen simultaneously and symmetrically, then absolute PPP will hold – i.e. there is some level of the exchange rate than can be called the relative true value of each currency. If changes do not occur homogeneously, then weak-form or relative PPP will hold – i.e. changes in variables will cause relative changes to the exchange rate. For instance, higher inflation in A would require a weakening of the exchange rate relative to B, but the domestic prices would still equalize to their previous level.

Empirical testing for weak form PPP is thus fairly simple – all you have to do is prove that the real exchange rate (the nominal exchange rate adjusted for inflation) is stationary i.e. it reverts to its previous equilibrium level. Proof for strong form PPP is obtained if the equilibrium level is also equal to unity.*

*If anybody wants to find out how this is done, drop me a line in the comments.

Empirically however, proving PPP has been highly problematical – real exchange rates don’t revert; or if they do, the movement is so slow that it can’t be proven that PPP holds. In fact, most early alternative models of exchange rate determination found little evidence that any single factor actually determines exchange rate movements (e.g. interest rate parity and monetary models).

One possible explanation for the empirical failure of PPP to hold is what’s known as the Balassa-Sameulson hypothesis (henceforth BSH for short). To illustrate the concept, let’s extend our model by adding another good, X, which is produced in both A and B, but is not tradable. The domestic price of X will be determined by local supply and demand, and not at the global level. This implies that the price of X can only be equalized globally under very special circumstances. Why this matters is the impact on the returns to the factors of production – if the price of X is not equalized, then wages cannot also be equalized globally (under the assumption of capital mobility, real interest rates would).

If demand for good X is higher in economy B, then wages would be uniformly higher in economy B relative to A, which reduces the productivity of B in the tradable sector (the production of W) because production costs are now higher - B's supply curve for W shifts left, because of the competition for labour for production of X. Since the supply of W is now lower and wages are higher in economy B, B’s domestic price of W will rise. In response, since PPP still applies to good W, the exchange rate of currency B relative to A must necessarily rise to equate the domestic prices. A will ceterus paribus produce more of W which will be exported to B, producing a trade surplus and a relatively weaker currency for A.

The only point at which the exchange rate would be identical to our first PPP model is if the relative size of the non-tradable sector to the tradable sector in A and B are the same. If they are not, then the equilibrium exchange rate must necessarily be different from the PPP model, and PPP will not hold for all goods and wages.

Here are some of the main implications of BSH:

1. A country with the tradable sector growing faster than the non-tradable sector will have a depreciating real exchange rate, and vice versa;

2. A positive productivity shock in the tradable sector (thus lowering production costs and prices) will result in a weaker real exchange rate, and vice versa;

3. A positive productivity shock in the non-tradable sector will result in a stronger real exchange rate, and vice versa;

4. It is not possible for the absolute form of PPP to apply to all goods, which rules out some notional nominal value of the exchange rate as a permanent, time-invariant equilibrium exchange rate;

5. The equilibrium exchange rate is thus time-varying, depending on the relative strengths of the non-tradable sector in each economy and their growth paths. QED

Thus, when applied to Malaysia, saying for instance that RM2.70 to the USD is the true equilibrium rate for the MYR may have been true in the early 1990s, but is unlikely to be correct now – continuing trade liberalization (which reduces the external price of our tradables), a faster growing tradable sector, and the increasing service sector orientation of the US economy necessarily implies that the equilibrium exchange rate between the MYR and the USD has been falling. In reference to the MYR-USD peg from 1998-2005, BSH also suggests that far from having a “weak” currency policy, Malaysia’s USD exchange rate was actually closing on the equilibrium real rate despite being nominally fixed.

If we consider the multilateral rate (the real effective exchange rate), it’s thus not obvious whether Malaysia ever had a “weak” currency policy that lasted much longer than 1998-2001 – strength in the USD to that point also meant an appreciation of the MYR relative to other floating currencies, which effectively negated any price advantage in export markets and in fact put us at a disadvantage by 2002. To take another example of BSH in action, Singapore shifted to a more service-oriented economy in the wake of the 1997-98 crisis; it is thus no accident that in real terms SGD has appreciated against the MYR since then. BSH also explains why the Indian Rupee continues to have an elevated value, when other metrics suggest it ought to be considerably lower – a strong service orientation as well as relatively high tariff barriers support a more appreciated currency.

This neatly leads into the second assertion in my previous post that I wanted to discuss. The services sector is generally taken to be composed of non-tradables – it is usually difficult to trade services, particularly personal services (the example usually given in text books is that of barbers). Exceptions of course exist, like segments of the transportation sector (port services and airlines for example) or tourism, but generally speaking prices of services are determined mostly by local factors rather than global ones. Thus a future strategy that places the services sector as the engine for growth (as likely to be followed by Malaysia in the coming years), will ceterus paribus result in an appreciation of the true equilibrium exchange rate. Since BNM policy appears to be neutral relative to the level of the MYR (as opposed to its volatility) the nominal rate will also appreciate.

Therefore, a deliberate policy of strengthening the currency to raise domestic incomes is unnecessary – the currency will strengthen anyway with a more services-oriented economy, as local wages are bid up by labour demand from industries which do not suffer from international price arbitrage. Nor is targeting some mythical nominal value of the exchange rate (particularly against the USD) economically sensible or viable – the equilibrium exchange rate can and will vary, depending on the relative balance between the tradable and non-tradable sectors in Malaysia as well as in each of our trade partners.

Of course, evidence of BSH is as hard to find as that of PPP, hence the reason it remains labeled a hypothesis rather than a theory. The main hurdle is fairly simple to articulate but almost impossible to overcome – it is extremely hard to make a distinction between what constitutes a tradable good and a non-tradable good. Government consumption is typically used to proxy the effect of BSH in empirical models, but that is under the (sometimes dubious) assumption that most of government spending falls on non-tradables. Some services sectors are subject to international competition as mentioned before, while most tradable goods contain elements of non-tradable inputs.

Nevertheless, BSH does appear to explain much of why exchange rates tend to persistently resist reverting to PPP values, as well as why some currencies are “stronger” than others. And it also explains why I have no faith in PPP-based judgements of “weakness” or “strength” in any currency.


  1. You make good points here - a lot of people trust PPP inherently, when there really are some things wrong with using that system as a method of currency comparison. Here is another interesting article about why PPP is flawed:

  2. Only some things wrong? LOL.

    Thanks for the link, and for stopping by.