Definitions

# Balassa-Samuelson effect

The Balassa-Samuelson effect (also known as Harrod-Balassa-Samuelson effect (Kravis and Lipsey 1983), the Ricardo-Viner-Harrod-Balassa-Samuelson-Penn-Bhagwati effect (Samuelson 1994, p. 201), productivity biased purchasing power parity (PPP) (Officer 1976) and the rule of five eights (David 1972)) is either of two related things:

1. The observation that consumer price levels in wealthier countries are systematically higher than in poorer ones (the "Penn effect").
2. An economic model predicting the above, based on the assumption that productivity or productivity growth-rates vary more by country in the traded goods' sectors than in other sectors (the Balassa-Samuelson hypothesis).

This article deals with point (2): Balassa and Samuelson's causal model. For a fuller description of the stylized fact it attempts to explain see: Penn effect.

## The theory

The Balassa-Samuelson effect (BS-effect) depends on inter-country differences in the relative productivity of the tradable and non-tradable sectors.

### The empirical "Penn Effect" effect

The exchange of tradable goods and services should lead prices to converge, but convergence is only partial, because some products are not tradable, and some products are generally produced locally (e.g. bread). (Software development is an example of tradable service, while a haircut is a non-tradable one.)

The Penn effect is that the RER (Real Exchange Rate) deviations usually occur in the same direction: where incomes are high, price levels, as for example measured by the Consumer Price Index are relatively high compared to an international average, and where they are low, they tend to be below the average.

### Basic form of the effect

The simplest model which generates a Balassa-Samuelson effect has two countries, two goods (one tradable, and a country specific nontradable) and one factor of production, labor. For simplicity assume that productivity, as measured by marginal product of labor, in the nontradable sector is equal between countries and normalized to one.

$MPL_\left\{nt,1\right\}=MPL_\left\{nt,2\right\}=1$

where "nt" denotes the nontradable sector and 1 and 2 indexes the two countries.

In each country, under the assumption of competition in the labor market the wage ends up being equal to the value of the marginal product, or the sector's price times MPL (note that this is not necessary, just sufficient. What is needed is that wages are at least related to productivity.):

$w_1=p_\left\{nt,1\right\}*MPL_\left\{nt,1\right\}=p_\left\{nt,1\right\}=p_\left\{t\right\}*MPL_\left\{t,1\right\}$

$w_2=p_\left\{nt,2\right\}*MPL_\left\{nt,2\right\}=p_\left\{nt,2\right\}=p_\left\{t\right\}*MPL_\left\{t,2\right\}$

Where the subscript "t" denotes the tradables sector. Note that the lack of a country specific subscript on the price of tradables means that tradable goods prices are equalized between the two countries.

Suppose that country 2 is the more productive, and hence, the wealthier one. This means that

which implies that

So with a same (world) price for tradable goods, the price of nontradable goods will be lower in the less productive country, resulting in an overall lower price level.

### The effect in more detail

A typical discussion of this argument (e.g. by Paul Krugman) would include the following features:

• Workers in some countries have higher productivity than in others. This is the ultimate source of the income differential. (Also expressed as productivity growth.)
• Certain labour-intensive jobs are less responsive to productivity innovations than others. For instance, a highly skilled Zurich burger flipper is no more productive than his Moscow counterpart (in burger/hour) but these jobs are services which must be performed locally.
• The fixed-productivity sectors are also the ones producing non-transportable goods (for instance haircuts) - this must be the case or the labour intensive work would have been off-shored.
• To equalize local wage levels with the (highly productive) Zurich engineers, McDonalds Zurich employees must be paid more than McDonalds Moscow employees, even though the burger production rate per employee is an international constant.
• The CPI is made up of:
• local goods (which are expensive relative to tradables in rich countries)
• Tradables, which have the same price everywhere
• The (real) exchange rate is pegged (by the law of one price) so that tradable goods follow PPP. The assumption that PPP holds only for tradable goods is testable.
• Since money exchange rates will vary fully with tradable goods productivity, but average productivity varies to a lesser extent, the (real goods) productivity differential is less than the productivity differential in money terms.
• Productivity becomes income, so the real income varies less than the money income does.
• This is equivalent to saying that the money exchange rate exaggerates the real income, or that the price level is higher in more productive, richer, economies.

### Equivalent 'Balassa-Samuelson effect' within a country

The average asking price for a house in a prosperous city can be ten times that of an identical house in a depressed area of the same country. Therefore, the RER-deviation exists independent of what happens to the nominal exchange rate (which is always 1 for areas sharing the same currency). Looking at the price level distribution within a country gives a clearer picture of the effect, because this removes three complicating factors:

1. The econometrics of purchasing power parity (PPP) tests are complicated by nominal exchange rate noise. (This noise would be an econometric problem, even assuming that the exchange rate volatility is a pure error term).
2. There may be some real economy border effects between countries which limit the flow of tradables or people.
3. Monetary effects, and exchange rate movements can affect the real economy and complicate the picture, a problem eliminated if comparing regions that use the same currency unit.

A pint of pub beer is famously more expensive in the south of England than the North, but supermarket beer prices are very similar. This may be treated as anecdotal evidence in favour of the Balassa-Samuelson hypothesis, since supermarket beer is an easily transportable, traded good. (Although pub beer is transportable, the pub itself is not.) The BS-hypothesis explanation for the varying price differentials is that publican's 'productivity' in serving customers is more uniform (in pints per hour) than is the 'productivity' (in foreign earnings per year) of people working in the export sector in either half of the country. (Reputedly Financial services in the South of England, heavy industry in the North.) The implication that one region is less 'productive' than another is politically controversial.

## Empirical evidence on the Balassa-Samuelson effect hypothesis

Evidence for the Penn effect is well established in today's world (and is readily observable when traveling internationally). However, the Balassa-Samuelson (BS) hypothesis implies that countries with rapidly expanding economies should tend to have more rapidly appreciating exchange rates (for instance the Four Asian Tigers); conventional econometric tests have resulted with mixed findings for the predictions of the BS effect.

In total, since it was (re)discovered in 1964, according to Tica and Druzic (2006) the HBS theory "has been tested 60 times in 98 countries in time series or panel analyses and in 142 countries in cross-country analyses. In these analyzed estimates, country specific HBS coefficients have been estimated 166 times in total, and at least once for 65 different countries". Also, one should have in mind that a lot of papers have been published since then. Bahmani-Oskooee and Abm (2005) and Egert, Halpern and McDonald (2006) also provide quite interesting surveys of empirical evidence on BS effect.

Over time, the testing of the HBS model has evolved quite dramatically. Panel data and time series techniques have crowded out old cross-section tests, demand side and terms of trade variables have emerged as explanatory variables, new econometric methodologies have replaced old ones, and recent improvements with endogenous tradability have provided direction for future researchers.

The sector approach combined with panel data analysis and/or cointegration has become a benchmark for empirical tests. Consensus has been reached on the testing of internal and external HBS effects (vis a vis a numeraire country) with a strong reservation against the purchasing power parity assumption in the tradable sector.

Analysis of empirical evidence shows that the vast majority of the evidence supports the HBS model. A deeper analysis of the empirical evidence shows that the strength of the results is strongly influenced by the nature of the tests and set of countries analyzed. Almost all cross-section tests confirmed the model, while panel data results confirmed the model for the majority of countries included in the tests. Although some negative results were returned, there has been strong support for the predictions of the cointegration between relative productivity and relative prices within a country and between countries, while evidence for cointegration between real exchange rate and relative productivity were much more controversial.

Therefore, most of the contemporary authors (see for example: Egert, Halpern and McDonald (2006) or Drine & Rault (2002) ) analyze main BS assumptions separately:

1. The differential of productivities between traded and non-traded sector and relative prices are positively correlated.
2. The purchasing power parity assumption is verified for tradable goods.
3. The RER and relative prices of non-tradable goods are positively correlated.
4. As a consequence of 1, 2, & 3, there is a long-run relationship between productivity differentials and the RER.

Refinements to the econometric techniques and debate about alternative models are continuing in the International economics community. For instance:

"A possible explanation of the BS empirical rejection may simply be that there are additional long-run real exchange determinants that have to be considered." Drine & Rault conclude.

The next section lists some of the alternative proposals to an explanation of the Penn effect, but there are significant econometric problems with testing the BS-hypothesis, and the lack of strong evidence for it between modern economies may not refute it, or imply that it produces a small effect. For instance, other effects of exchange rate movements might mask the long-term BS-hypothesis mechanism (making it harder to detect if it exists). Exchange rate movements are believed by some to have an impact on productivity; if this is true then regressing RER movements on differential productivity growth will be 'polluted' by a totally different relationship between the variables1.

## Alternative, and additional causes of the Penn effect

Most professional economists accept that the Balassa-Samuelson effect model has some merit. However other sources of the Penn effect RER/GDP relationship have been proposed:

### The distribution sector

In a 2001 International Monetary Fund working paper Macdonald & Ricci accept that relative productivity changes produce PPP-deviations, but argue that this is not confined to tradables versus non-tradable sectors. Quoting the abstract: "an increase in the productivity and competitiveness of the distribution sector with respect to foreign countries leads to an appreciation of the real exchange rate, similarly to what a relative increase in the domestic productivity of tradables does".

### The Dutch Disease

Capital inflows (say to the Netherlands) may stimulate currency appreciation through demand for money. As the RER appreciates, the competitiveness of the traded-goods sectors falls (in terms of the international price of traded goods).

In this model, there has been no change in real economy productivities, but money price productivity in traded goods has been exogenously lowered through currency appreciation. Since capital inflow is associated with high-income states (e.g Monaco) this could explain part of the RER/Income correlation.

Yves Bourdet and Hans Falck have studied the effect of Cape Verde remittances on the traded-goods sector. They find that, as local incomes have risen with a doubling of remittances from abroad, the Cape Verde RER has appreciated 14% (during the 1990s). The export sector of the Cape Verde economy suffered a similar fall in productivity during the same period, which was caused entirely by capital flows and not by the BS-effect..

### A demand side explanation

The Penn effect PPP-deviation can be derived from the demand side of the economy, rather than the Balassa-Samuelson supply side model, in a similar way to the Dutch Disease explanation above.

When any non-tradable comes up for sale, its price will be determined by the relative preference between it and money by the average market consumer. By definition, high income consumers have more money, and are indifferent at a higher sale prices between buying an item and not doing so, relative to consumers in a low income area. In tradable goods, supply could shift from poor regions to rich to take advantage of this, forcing price convergence. However, non-tradable supply cannot do this, by definition. Therefore, price differences are caused (in this model) by nothing but relative differences in the abundance of money.

In this demand-side model, the initial sources of income difference are treated as given. (Income is either exogenous or evolves based on the ability to sell non-tradables at higher prices where incomes are higher.) This model leads to random walk RER behaviour, as the exogenous rich trickle their wealth down to nearby workers without requiring them to improve productivity (the rich simply bid up local service prices). Charging what the market will bear creates the PPP-deviation in a similar way to the Balassa-Samuelson effect, but doesn't explicitly rely on productivity differentials or the changes in them.

### Services are a 'superior good'

Rudi Dornbusch (1998) and others say that income rises can change the ratio of demand for goods and services (tradable and non-tradable sectors). This is because services tend to be superior goods, which are consumed proportionately more heavily at higher incomes.

A shift in preferences at the microeconomic level, caused by an income effect can change the make-up of the consumer price index to include proportionately more expenditure on services. This alone may shift the consumer prince index, and might make the non-trade sector look relatively less productive than it had been when demand was lower; if service quality (rather than quantity) follows diminishing returns to labour input, a general demand for a higher service quality automatically produces a reduction in per-capita productivity.

A typical labour market pattern is that high-GDP countries have a higher ratio of service-sector to traded-goods-sector employment than low-GDP countries. If the traded/non-traded consumption ratio is also correlated with the price level the Penn effect would still be observed with labour productivity rising equally fast (in identical technologies) between countries.

### The protectionism explanation

Lipsey and Swedenborg (1996) show a strong correlation between the barriers to Free trade and the domestic price level. If wealthy countries feel more able to protect their native producers than developing nations (e.g. with tarrifs on agricultural imports) we should expect to see a correlation between rising GDP and rising prices (for goods in protected industries - especially food).

This explanation is similar to the BS-effect, since an industry needing protection must be measurably less productive in the world market of the commodity it produces. However, this reasoning is slightly different from the pure BS-hypothesis, because the goods being produced are 'traded-goods', even though protectionist measures mean that they are more expensive on the domestic market than the international market, so they will not be "traded" internationally

## Trade theory implications

The supply-side economists (and others) have argued that raising International competitiveness through policies that promote traded goods sectors' productivity (at the expense of other sectors) will increase a nation's GDP, and increase its standard of living, when compared with treating the sectors equally. The Balassa-Samuelson effect might be one reason to oppose this trade theory, because it predicts that: a GDP gain in traded goods does not lead to as much of an improvement in the living standard as an equal GDP increase in the non-traded sector. (This is due to the effect's prediction that the CPI will increase by more in the former case.)

## History

The Balassa-Samuelson effect model was developed in 1964 by both Balassa Béla & Paul Samuelson, working independently.

It is surprising that both of these economists should have completed their models separately & simultaneously (submitting them to different economic journals) because the outlines of the explanation had been described twenty-five years earlier by Roy Forbes Harrod in "International Economics".

Partly because empirical findings have been mixed, and partly to differentiate the model from its conclusion, modern papers tend to refer to the "Balassa-Samuelson hypothesis", rather than the "Balassa-Samuelson effect". (See for instance: " A panel data analysis of the Balassa-Samuelson hypothesis", referred to above.)

## Bibliography

• Bahmani-Oskooee, Mohsen and Abm Nasir. 2005. Productivity Bias Hypothesis and the Purchasing Power Parity: A Review Article Journal of Economics Surveys, 19:4, pp. 671-696.
• Bahmani-Oskooe & Rhee (1996) Time-series Support for Balassa's Productivity-bias Hypothesis: Evidence from Korea, Review of International Economics 4
• Balassa, B. (1964),"The Purchasing Power Parity Doctrine: A Reappraisal", Journal of Political Economy, vol. 72, December, pp. 584-596
• David, Paul. A. 1972. Just How Misleading are Official Exchange Rate Conversions? The Economic Journal, 82:327, pp. 979-990.
• Dornbusch, R, 1988. Purchasing Power Parity. In The New Palgrave Dictionary of Economics. Palgrave Macmillan; Reprint edition (September, 1998) ISBN 1-56159-197-1
• Harrod, R. F. 1933 International Economics, Cambridge University Press (Reset edition published by Pitman, 1949 ASIN: B0007E5ZX6)
• Tica, J. and Druzic, I. 2006 The Harrod-Balassa-Samuelson Effect: A Survey of Empirical Evidence, EFZG Working Paper Series 0607.
• Drine, I., Rault, C., Panel data analysis of the Balassa-Samuelson hypothesis, 2002 Sorbonne University ()
• Egert, Balazs; Laszlo Halpern and Ronald Macdonald. (2006.) Equilibrium Exchange Rates in Transition Economies: Taking Stock of the Issues, Journal of Economic Surveys, 20:2, pp. 257-324.
• Kravis & Lipsey (1991), The International Comparison Program: Current Status and Problems, from International Economic Transactions : Issues in Measurement and Empirical Research (National Bureau of Economic Research Studies in Income and Wealth) , University of Chicago Press (1992) ISBN 0-226-35135-1
• Lipsey, Robert E. and Birgitta Swedenborg (1996), "The High Cost of Eating: Causes of International Differences in Consumer Food Prices", Review of Income and Wealth June 1996
• Lu, D. and Q.Yu. 1999, HongKong's Exchange Rate Regime: Lessons from Singapore, China Economic Review
• MacDonald, R. and Ricci, L. 2001 PPP and the Balassa Samuelson Effect: The Role of the Distribution Sector, IMF Working Paper, WP/01/38
• Officer, Lawrence H. 1976. The Productivity Bias in Purchasing Power Parity: An Econometric Investigation IMF Staff Paper 23, pp. 545-579.
• Porter, M.E., 1998. The Competitive Advantage of Nations, Free Press (publisher) ISBN 0-684-84147-9 (Discusses national comparative advantage as well as the productivity—exchange rate link.)
• Paul Samuelson (1964), Theoretical Notes on Trade Problems, Review of Economics and Statistics, volume 23