Volatility and GrowthFinancial Dev elopmen t and Cyclical Composition of Investment∗Philippe Aghion George-Marios Angeletos Abhijit Banerjee Kalina ManovaHarvard and NBER MIT and NBER MIT and NBER Harvard and [email protected] [email protected] [email protected] [email protected][very preliminary draft]AbstractWe examine how financial development affects the cyclical behavior of the composition of invest-ment and thereby volatility and growth. We first develop a simple endogenous-growth model inwhich firms engage in two types of investment: a short-term one, and a long-term productivity-enhancing one. Because it tak es longer time to complete, long-term investment has a relativelyless procyclical return but also a higher probability to be hit by a liquidity shock. Under com-plete financial markets, only the opportunity-cost effect is present — long-term investment iscountercyclical, thus mitigating volatility. But when firms face tight borrowing constraints, theliquidity effect dominates — long-term investment turns procyclical, thus amplifying volatility.We next confront the model with a panel of coun tries over the period 1960-2000. We find thata lower degree of financial development predicts a higher sensitivity of both the composition ofinvestment and productivity growth to exogenous shocks.JEL codes: E22, E32, O16, O30, O41, O57.Keywords: Growth, fluctuations, business cycles, credit constraints, amplification, R&D.∗For helpful comments a nd discussions, we thank Daron Acemoglu, Philippe Bacchetta, Robert B arro, OlivierBlanchard, V .V. Chari, Bronwyn Hall, Peter Howitt, Olivier Jeanne, Patrick K ehoe, Ellen McG rattan, Klaus Walde,Ivan Werning, and seminar participants in Amsterdam, U C Berkeley, ECFIN, Harvard, IMF, MIT, and the FederalReserve Bank in M inneapolis. Special thanks to Do Q uoc-Anh for excellent research assistance.1IntroductionThis paper is motivated by three observations: the negative correlation between growth and volatil-ity in the cross-section of countries; the failure of investment/saving rates to account for thisnegative correlation; and the fact that this correlation is stronger the lower the level of financialdevelopment. In the light of these facts, we propose that a key missing element is the compositionof investment: how it varies over the business cycle; how it depends on financial development; andhow it affects gro wth and volatility.Motivating facts. The idea that there is a close connection between productivity growthand the business cycle goes back at least to Schumpeter, Hicks, and Kaldor in the 1940s-1950s.Usingcross-sectionaldatafrom92countries, Ramey and Ramey (1995) find a negative correlationbetween volatility and growth (the former measured by the standard deviation of annual per-capitaGDP gro wth rates and the latter by the corresponding mean). As shown in columns (1)-(4) of Table1, which repeat their exercise in an expanded cross-country data set that we use in this paper, thisrelation is robust to controlling for policy and demographics variables as in Levine et al. (2000).To the extent that this evidence reflects a causal effect of volatility on growth,1a straightforwardcandidate explanation is risk aversion: higher volatility means more investment risk, which tendsto discourage investment and slow down growth. This effect may be partly or totally offset by theprecautionary motive for savings: higher income risk may raise precautionary savings, reduce inter-est rates, and thereby boost investment. In an AK economy, for example, the general-equilibriumeffect of volatility on savings and growth is negative if and only if the elasticity of intertemporalsubstitution is higher than one (Jones, Manuelli and Stacchetti, 2000).2In any event, the simpleneoclassical paradigm can account for a negative correlation between volatility and growth to theextent that higher volatility is correlated with lower investment rates.As shown in columns (4)-(8) of Table 1, however, controlling for investment rates does notsubsume the impact of volatility. In the whole sample, for example, the change in the coefficientof volatility is insignificant, with the point estimate falling from −0.26 to −0.22.Primafacia,thisfinding suggests that the main channel through which volatility affects growth is not the rate ofinvestment.[insert Table 1 here]Earliers studies have documented the positive correlation between financial development andeconomic growth. (See Levine, 1997, for a review.) In Table 2, we repeat the growth regressions1The effect remains negative when Ramey and Ramey (1995) instrument volatility with (arguably) exogenousinnovations in government spending. See also Gavin and Hausmann (1996).2The results in Angeletos (2004) suggest that in a neo classical growth economy — where the income share ofcapital is less than one — productivity risk can have a negative impact on saving rates even when the elasticity ofintertemporal substitution is substantially be low one.1of Table 1 including the country’s level of financial development as one of the controls, but alsoadding its interaction with volatility.3The level effect is the familiar one; the news is the stronginteraction effect.The negative correlation between volatility and growth appears to be stronger in countries withlo wer financial development. In column (1), for example, a one standard deviation increase in thelevel of financial development w ould reduces the impact of a 1% rise in volatility by −0.32% (=0.011 · 29). This interaction effect is robust to controlling for policy and demographics variables, aswell as for the rate of investment rate.[insert Table 2 here]Model and theoretical results. Motivated by these facts, the first part of the paper developsa simple endogenous-growth model which focuses on the cyclical composition of investment as themain propagation channel. A large number of agent s (“entrepreneurs”) engage in two types ofinvestment activity. The one type, which we call “short-term investment”, takes relatively littletime to build and generates output relatively fast. The other, which we call “long-term investmen t”,takes more time to complete and yields a return further in the future, but contributes relativelymore to productivity growth.That borrowing constraints may amplify the cyclical variation in the demand for investmentis of course well known. Note, however, that this effect may be offset in general equilibrium bythe