While profits, dividends and hence share prices might all be expected to rise when economies perform well, experience indicates that the links between stock markets and the real economy are much more complex. For example, the London Stock Exchange's market capitalization fell by around $140 billion between early February and mid-July of 1994, a time of increasing optimism as the growth rate of UK output rose above its long-term trend; stock markets in the US and Continental Europe and prices on world bond markets also fell sharply. A joint conference with the Weiss Center for International Financial Research at the Wharton School of the University of Pennsylvania on `International Stock Returns and Business Cycles', held at CEPR on 3/4 June, brought together specialists in international macroeconomics and finance to investigate the complex links between stock market behaviour and the real economy. The conference was organized by Bernard Dumas, Professor of Finance at the HEC School of Management, Jouy-en-Josas, and Research Fellow in CEPR's International Macroeconomics and Financial Economics programmes, and Richard Marston, Director of the Weiss Center and Professor of Finance and Economics at the Wharton School.

International Portfolio Diversification and Home Bias

Economists have long been puzzled by the tendency of even sophisticated international investors to hold portfolios dominated by domestic equities when cross-country differences in performance suggest that a wider spread would better enable them to maximize returns and reduce exposure to risk. Many have attributed this `home bias' to the roles of non-traded goods or factors of production. In `Nontraded Goods, Nontraded Factors and International Non-Diversification', written with Urban Jermann and Robert King, Marianne Baxter (University of Virginia) developed a multi-country general equilibrium model to show that it is never optimal to exhibit home bias with respect to the equities of producers of traded goods, but it may be optimal under certain conditions for investors to devote substantial shares of their portfolios to the equities of domestic producers of non-traded goods. If the capitalized value of labour income is taken into account in portfolio decisions, moreover, then it may be optimal to hold substantial short positions in the equities of firms producing traded goods. Home bias is therefore difficult to justify theoretically, especially with respect to the equities of producers of traded goods.

Enrique Mendoza (IMF) reported general agreement among the IMF officials who talk regularly to financial market participants that home bias is substantial, although it has fallen in the past decade and varies across countries and asset types. These officials attribute home bias to exchange rate risk, differences in regulatory regimes and fear of the imposition of capital controls. Bernard Dumas questioned whether the threat of real exchange rate volatility could account for home bias, since nominal exchange rate changes were both easily hedged and well matched with real exchange rate changes.

In his paper, `Does Industrial Structure Explain the Benefits of International Diversification?', written with Steven Heston, K Geert Rouwenhorst (Yale University) challenged the view that cross-country differences in industrial composition account for differences in stock market behaviour and hence that international diversification simply provides a means of achieving industrial diversification. The performance of 829 shares from twelve European countries and seven industries during 1978-92 indicated that the national stock markets which achieved above-average performance did so because their shares performed well across the board rather than because their market compositions were skewed towards particularly high-performing industries. Country-specific influences may include national monetary and fiscal policies, regional shocks and institutional and legal regimes, which seem more important than exchange rate changes. Rouwenhorst found that the value of a randomly chosen portfolio of shares from different industries in one country varied almost twice as much on average as one of shares from different countries in a particular industry.

In discussion, Gordon Bodnar (Wharton School) mentioned that other studies had found the importance of country- and industry-specific influences in determining changes in industrial production to be roughly equal, which is puzzling since output is known to be closely linked to stock market performance. Sushil Wadhwani (Goldman Sachs International and CEPR) argued that Rouwenhorst had underestimated the importance of industrial composition: four of the ten best-performing US global investment funds employed industry analysts but no country analysts. His industrial classifications were also far too broad, with Glaxo and BMW implausibly in the same industry; the 1978 starting-point was too early, since investors had become globalized only recently; and the strong recent stock market performance of countries that had devalued out of the ERM clearly demonstrated that exchange rate changes do play an important role.

The next paper, `Why Do Consumption and Stock Returns Suggest Such Different Costs of Imperfect Risk Sharing?', by Karen Lewis (Wharton School), examined the costs of home bias and was presented in her absence by Maurice Obstfeld. If these were small, the bias might reflect investors' avoidance of the time and expense of researching and maintaining broader international portfolios. Previous studies applying a general equilibrium framework to consumer spending data had found such costs to be less than 0.5% of long-term consumer spending, while those applying a partial equilibrium framework to data on stock market returns had yielded estimates of at least 2%. Lewis attributed these differences to the greater variability of stock market returns relative to consumption rather than any relationship among stock market returns across countries. Consumer spending is not sufficiently variable to justify holding diversified portfolios in order to smooth consumption.

Linda Tesar (University of California, Santa Barbara) stressed that shares were not the only financial assets available and that individuals' portfolios might also include bonds, real estate and human capital (skills and education). Maurice Obstfeld (University of California, Berkeley, and CEPR) noted that dividends accounted for only 7% of US national income and almost no-one lived solely on income from investments; future research should focus on the distribution of non-dividend income and differences in individuals' access to financial markets.

Inflation and Wealth Redistribution

Policy-makers have long pursued low inflation to minimize arbitrary redistributions from lenders to borrowers that arise from any general rise in the price level. Holdings of nominal bonds provide a connection between the real economy and financial markets, as unanticipated inflation encourages debtors (creditors) to reduce (increase) their spending. In `Debt and Disinflation', Mario Crucini (Ohio State University) noted that the pessimistic ten-year forecasts of US inflation made in 1978 had led to large wealth transfers from long-term borrowers to long-term lenders. If bonds with a three-month maturity were held by a tenth of the population, they would have gained $2,769 per capita and raised the tax liability of the remaining nine-tenths by $287 per capita. Ten-year bonds would have provided wealth transfers about ten times greater, so investors in long-term bonds made substantial gains in the 1980s. A similar calculation for the wealth transfer from Latin America to the US because international debt is denominated in dollars and inflation was unexpectedly low revealed that transfers of $576 from each Latin American citizen would have benefited each US citizen to the tune of $64 with three-month debt, while transfers arising from ten-year bonds issued in the late 1970s and not renegotiated would have been some ten times greater. In practice, however, most of Latin America's debt was renegotiated and paid off at variable interest rates.

Ingrid Werner (Stanford University) questioned whether US investors and foreigners borrowing in dollars forecast US inflation in different ways and wondered why the US had not encouraged inflation in the 1980s, when it was a net borrower on international financial markets. Gregor Smith (Queen's University, Ontario) agreed that Crucini's estimated transfers for Latin America might be too large because of debt rescheduling, but they might also be too small because of possible bankruptcy, because debt might remain outstanding after the period of the inflation forecast, or because of differences in debtors' and creditors' spending behaviour. The nature of the wealth redistributions depended critically on who held the debt, its maturity and expected future tax liabilities. Bernard Dumas suggested that measuring the wealth transfers in the currency of the debtor country might indicate that the total values of the gains and losses no longer cancelled out.

Booms and Crashes

Violent booms and crashes in asset prices involving fluctuations substantially larger than any changes in the determinants of their supply or demand have been common in market economies for centuries. Popular explanations of these swings emphasizing the role of irrational speculation are unsatisfactory, and more recent accounts allow demand and supply to balance at more than one equilibrium and prices to change suddenly as the system swings from one to the other. In `Informational Overshooting, Booms and Crashes', Joseph Zeira (Hebrew University of Jerusalem and CEPR) provided an explanation of booms and crashes that relied on neither irrational behaviour nor multiple equilibria. Booms and crashes in share prices may arise as investors update their expectations in a process of `informational overshooting'. When the dividend on a particular share begins to rise, agents know that it cannot rise for ever but do not know at which point it will stop. As the dividend rises, they raise their estimates of its final level and their expectations of future dividends, which raises the share price further. When the dividend finally stops rising, however, expected future dividends suddenly fall to their current level and the share price crashes to its long-run level. Booms in asset markets may often be triggered by financial deregulation or liberalization: these reduce the costs of market entry for new investors, which increases its size, efficiency and productivity, which in turn attracts more entrants until this process also ends with a crash. Such a process may account for the US stock market crashes of the 1920s and the 1980s.

Kenneth Rogoff (Princeton University) questioned whether this model could be applied successfully to exchange rate changes, since a boom in one currency implies a crash in another. Also the 1929 fall in share prices was not a conventional crash, since share prices initially fell by only 15% but finally fell by 90% in a `slow melt'. Robert King (University of Virginia) said that the booms in Zeira's model were implausibly abrupt, but smoother price rises might be explicable if investors did not observe the beginning of the rise directly but rather formed a belief that dividends will begin to rise after a couple of years.

Explaining Stock Market ReturnsIn his joint paper with John Donaldson, `Asset Pricing Implications of Real Market Frictions', Jean-Pierre Danthine (Université de Lausanne and CEPR) investigated whether frictions in the real economy might help explain some of the puzzles arising from financial market behaviour. Simple theoretical models fail to explain why shares yield much higher returns than risk-free assets, why share prices vary so much more than dividends or why stock market capitalizations vary so much more than the total values of goods and services produced. They also suggest that consumer spending should vary less over time than it does in practice. Such models assume that returns to investments in capital equipment and in financial assets are equal, and introducing `frictions' into the real economy breaks this link. He showed that allowing the installation of capital equipment to be costly in itself goes some way to improving such models' performance, while assuming that companies raise money from debt as well as equity also improves their results slightly in part because dividends can only be paid after servicing the debt. Differentiating between shareholders and workers improves the results substantially. Workers cannot hedge against risk in financial markets and therefore negotiate contracts in which they pay their employers a premium to smooth wages, which takes the form of a higher return on equities. Dividends and returns on shares vary more as the economy fluctuates because employees must be supported when times are bad.

In discussion, Andrew Abel (Wharton School) said that the structure of Danthine's model implied that his finding that labour contracts contributed more than company debt to explaining the high return on equities rested critically on the fact that the share of wages in national income was much greater than the share of interest payments in total output. David Backus (New York University) suggested investigating the implications of possible differences between agents in different countries. Robert King reported that many models which generated realistic premiums for the return on equities relative to short-term, risk-free bonds also tended to raise the risk premium on long-term assets to unrealistic levels. He suggested testing whether this applied to Danthine's model.

In his paper with Gianni De Nicolo, `Stock Returns and Real Activity: A Structural Approach', Fabio Canova (Universitat Pompeu Fabra, Barcelona, and CEPR) developed two versions of a general equilibrium model to investigate the effects of factors that change both share prices and the indicators of real economic performance that are often used to account for them. A model focusing on technological advances produced results that corresponded more closely to US experience than a model of government spending changes, although the latter provided stronger econometric results. A persistent technological advance boosts future expected cash flows by raising expected output above its long-run trend but encourages people to postpone consumption, so that share prices are only weakly associated with the ups and downs of the business cycle. In contrast, government spending boosts future expected cash flows as people work harder and raise output, but they also encourage people to bring forward their consumption; this produces an unrealistically strong relationship between the business cycle and share prices.

Alberto Giovannini (Columbia University and CEPR) suggested investigating whether stock market behaviour provides any information with which to predict business cycle changes beyond that already available from changes in interest rates or the money supply. Evidence from the US suggests that it does.

In his paper with Warren Bailey, `Business Cycles and the Predictability of International Stock Returns: Domestic versus Global Effects', Eric Jacquier (Cornell University) investigated whether Japanese stock market returns were determined by information on the Japanese business cycle or US economic fundamentals, which might proxy global business cycle conditions. He found that neither US nor Japanese economic fundamentals could explain Japanese stock market returns during 1975-81, but the Japanese fundamentals were effective predictors during 1981-91 when the US fundamentals provided little additional explanatory power. Monthly data on the fundamentals using only the information available at the time had no more predictive power than past average stock market returns, however, although using past fundamentals provides a slight edge over this naïve forecasting technique when the data on returns is filtered to remove short-term fluctuations.

Craig MacKinlay (Wharton School) noted that the Japanese dividend yield had displayed a strong downward trend for much of the period, and the US and Japanese dividend yields tended to move in opposite directions, so the latters' poor performance as predictors of market returns was hardly surprising. Correcting for the Japanese yields' long-term trend revealed, however, that both yields then tended to move together; indeed, the detrended Japanese yield performed better than the US yield as a proxy for global trends and proved an equally effective predictor of US stock market returns.