Natural disasters as creative destruction? Evidence from developing countries.

Cuaresma, Jesus Crespo ; Hlouskova, Jaroslava ; Obersteiner, Michael 等

I. INTRODUCTION

The literature on the economic effects of natural disasters has concentrated traditionally on the short-run response of economic variables to catastrophic events. Most of the research carried out on this topic, starting with Dacy and Kunreuther (1969), tends to find that gross domestic product (GDP) increases after the occurrence of a natural disaster. Albala-Bertrand (1993a, 1993b) showed that even for large disasters (as measured by the loss-to-GDP ratio), the reconstruction effort needed to keep the level of output from falling is relatively small. Tol and Leek (1999) also provided evidence of positive effects of natural disasters on macroeconomic variables in the short run.

While the predominant view usually stated in official statements of international organizations and governments is that natural disasters are an enormous barrier to economic development (see, e.g., United Nations Development Programme 2004), the quantification of long-run economic effects of natural disasters was an empty field of research until very recently. To our knowledge, the article by Skidmore and Toya (2002) is the only piece of empirical research that assesses directly the long-run economic impact of natural disasters. Using a cross-section of developed and developing countries, Skidmore and Toya (2002) showed that after conditioning on other determinants, the frequency of climatic disasters is positively correlated with human capital accumulation, total factor productivity (TFP) growth, and GDP per capita growth.

One of the explanations put forward in support of the existence of a positive partial correlation between the frequency of natural disasters and both TFP and GDP per capita growth is related to the absorption of new technologies. A country whose capital stock is reduced by a natural disaster may have an incentive to replace it with capital that embodies newer technology than that which was destroyed. (1) This would lead to higher rates of TFP and GDP per capita growth and would render natural disasters an example of Schumpeterian "creative destruction," a concept that, embedded in the theory of endogenous growth, has recently become a key explanation of long-run economic growth patterns (see, e.g., Aghion and Howitt 1998). The same idea was put forward in Okuyama (2003) and Okuyama, Hewings, and Sonis (2004), where it was argued that older equipment is more exposed to damage when a disaster hits the capital stock; thus, the replacement of these facilities would constitute a positive productivity shock, which may have permanent consequences in the growth rate of the whole economy. Skidmore and Toya (2002) found a positive partial correlation between the frequency of climatic disasters and TFP growth for a cross-section of 89 developed and developing countries. The results for geologic disasters indicate no significant effect of these on TFP growth. Although their conclusion is that "disasters provide opportunities to update the capital stock and adopt new technologies" (Skidmore and Toya 2002, p. 681), the measure of TFP used in order to arrive at this conclusion contains information on many other (observable and unobservable) institutional, political, and economic variables. Furthermore, given that the method used in computing TFP, based on Coe and Helpman (1995), does not account for human capital as a factor of production, the correlation found between TFP growth and climatic disasters may just be picking up the substitution of physical for human capital in disaster-prone countries.

It should be explicitly stated that there are basic differences between the Schumpeterian concept of creative destruction and the effect that natural disasters are hypothesized to have as mentioned in the studies given above and in the analysis performed in this article. Schumpeter's view on creative destruction emphasized competition dynamics as the engine behind technological progress (Schumpeter 1950), while the term in this article refers to a more literal interpretation with similar ex-post effects, namely, technology replacement after a catastrophic event. (2) The greater exposure of old vintage capital stock to catastrophic phenomena, together with the fact that natural disasters tend to be rare events, may give the affected economy the opportunity to upgrade obsolete equipment with the leading edge technology after the older stock is destroyed. Similar arguments have been raised in order to explain the economic performance of countries defeated in major wars (see, e.g., Koubi 2005; Organski and Kugler 1977).

This article contributes to the literature on the economic effects of natural disasters by assessing directly the relationship between foreign technology absorption and catastrophic events. We use an estimate of the R&D stock embodied in the imports of developing economies from the G-5 countries in order to investigate the relationship between foreign knowledge spillovers and catastrophic risk. In principle, concentrating on developing countries, which can be assumed to have relatively more obsolete capital stock, may give some insight as to whether the impact of natural disasters triggers technological upgrading. After conditioning upon the usual determinants of trade implied by gravity equations, we do not find systematic evidence of a positive partial correlation between the frequency of natural disasters and the R&D content of imports for our cross-section of developing countries in the period 19761990. If an interaction with the level of development of the receptor country is included in the regression, the results indicate that natural disasters tend to affect technology absorption positively only in countries with relatively high levels of GDP per capita. The results are reinforced if the time dimension of the data is exploited, and we present results for a panel obtained by pooling different subperiods of the available sample.

This article is structured as follows. Section II presents details on the computation of the foreign R&D stock variable, together with some preliminary results on its relationship with the different proxies for catastrophic risk. Section III reports the results of the estimation of different gravity equations augmented with variables accounting for the frequency and intensity of natural disasters. Section IV concludes.

II. R&D STOCKS, FOREIGN KNOWLEDGE ABSORPTION, AND NATURAL DISASTERS

Starting with the seminal contribution by Romer (1986), endogenous growth theory has provided a sound theoretical framework for the empirical examination of the effect of knowledge spillovers on economic growth (see, e.g., Aghion and Howitt 1992; Grossman and Helpman 1991). The main focus of this empirical literature has been the measurement of trade-related R&D spillovers. The underlying idea is that countries gain access to foreign technologies through trade; thus, those economies benefiting from imports from nations with a higher level of technological knowledge will experience higher growth rates of income per capita than those whose trade partners possess a lower technological level. Evidence on the existence and size of such knowledge spillovers is given in Coe and Helpman (1995): Coe, Helpman, and Hoffmaister (1997); and Eaton and Kortum (1996), to name some of the most relevant pieces of research in this field. (3)

Our aim was to measure the effect of catastrophic risk on the degree of absorption of foreign technology. The approach usually taken in empirical work to obtain a proxy for foreign knowledge spillovers is to obtain a measure of the R&D stock embodied in the imports of the country of interest using some weighted sum of the R&D stock of its trading partners. The measure of total foreign R&D stock proposed by Coe and Helpman (1995) is an import share-weighted average of the domestic R&D of country i's trade partners,

(1) [RD.sup.f.sub.i,t] = [summation over (j)] [[eta].sub.ij,t]/ [[eta].sub.i,t] [RD.sup.d.sub.j,t],

where [RD.sup.d.sub.j,t] is the domestic R&D stock of the exporting country j (country i's trade partner) at time t, [[eta].sub.ij,t] is the volume of imports of goods and services from country j to country i, and [[eta].sub.i,t] is the total volume of imports of country i from its trade partners at time t. The [RD.sup.f.sub.i,t] variable can thus be interpreted as the technological content of an average unit of imported good. Figure 1 presents scatterplots of the average [RD.sup.f.sub.i,t] (in logs) in the period 1976-1990 against the number of natural disasters per square kilometer for 49 developing countries listed in Table 1. (4) In the figure, we consider imports of manufactured goods from the G-5 (United States, Germany, United Kingdom, Japan, and France) for the computation of Equation (1). (5) In the scatterplot, the disaster variable is defined as log(1 + [dis.sub.i]), where [dis.sub.i] is the number of disasters per square kilometer in country i during the period 1960-1990. (6) Following Skidmore and Toya (2002), we also consider a disaggregation of total disasters into climatic (floods, cyclones, hurricanes, ice storms, snow storms, tornadoes, typhoons, storms, wild fire, drought, and cold wave) and geologic (volcanic eruptions, natural explosions, avalanches, land slides, earthquakes, and wave/surge) disasters. We present scatterplots for the aggregate number of disasters per square kilometer in order to account for the fact that bigger countries may be more subject to the occurrence of catastrophes. A linear regression line is also plotted. The unconditional correlation between the technological content of an average unit of imported good, [RD.sup.f], and our catastrophic risk variables is positive and significant in all cases, which could be taken as a first indication of technology upgrading, driven by natural disasters. There are, however, some other variables that have an effect on technology spillovers and should be taken into account when assessing the effect of catastrophic risk on R&D spillovers.

A usual critique to the use of Equation (1) is that this measure of knowledge spillovers does not take into account the intensity of trade. (7)

For a given set of countries with equal size and composition of imports (in terms of trading partners), one would expect larger spillovers taking place in the economy that imports more. Coe, Helpman, and Hoffmaister (1997), for instance, found that the higher the degree of openness of the receptor country, the higher the impact of foreign R&D spillovers will be. (8) In order to account for this, several other measures have been proposed that incorporate some measure of openness in the definition of the technology spillover. A measure for the total stock of foreign R&D proposed by Falvey, Foster, and Greenaway (2002), which is suited to the analysis of technological spillovers by means of gravity equations, is given by (9)

(2) [RD.sup.f*.sub.i,t] = [[eta].sub.i,t][RD.sup.f.sub.i,t] = [summation over (j)] [[eta].sub.ij,t][RD.sup.d.sub.j,t],

where the R&D spillover between country j and country i is thus given by [RD.sub.ij,t] = [[eta].sub.ij,t] [RD.sup.d.sub.j,t]. This measure accounts for the overall volume of imports of country i as a decisive factor for the size of the knowledge spillover, and thus, it is the basic measure used in our analysis. The study of the relationship between catastrophic risk and knowledge spillovers can be embedded in the usual framework of gravity equations, which have been widely used for addressing the empirics of international trade in the economic literature.

[FIGURE 1 OMITTED]

Admittedly, developing countries could develop new technologies themselves. Unfortunately, there are no reliable (and most of the times, no available) sources for R&D expenditures in most developing economies in our sample. The literature treating with the issue of technology transfer to developing countries systematically excludes domestic R&D activities for developing countries from the analysis or assumes that the R&D expenditure in developing countries is equal to zero. We therefore concentrate exclusively on the trade channel of technology absorption and abstract away from analyzing the effect of natural disasters on other forms of technology upgrading.

III. NATURAL DISASTERS AND TECHNOLOGY TRADE

In this section, we obtain estimates of the effect of catastrophic risk on technology transfer using gravity equations. We start by estimating gravity equations for the cross-section of countries with average values for the period studied (1976-1990) and then turn to disaggregated data in 5- and 3-yr intervals. Gravity models have become one of the most powerful empirical tools for analyzing trade relationships. Introduced originally by Linder (1961) and Linnemann (1966), the basic specification of a gravity model relates aggregate trade flows between two countries to the aggregate GDP in the respective countries and the geographical distance between the two economies, which is the usual variable for trade costs.

The "technology exporting economies" in our setting is given by the G-5 countries, (10) and the basic specification to be estimated for the cross-section of countries for the period under study is:

(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [RD.sub.ij] is the level of technology spillovers between country i and country j, defined as [RD.sub.ij,t] = [[eta].sub.ij,t][RD.sup.d.sub.j,t], in which i refers to the developing (importing) country and j to each one of the G-5 (exporting) countries; [Y.sub.i] ([Y.sub.j]) is the level of GDP of country i (j); and [d.sub.ij] is the distance between the capital city of country i and country j. [I.sub.k] and [E.sub.k] refer to a set of importer- and exporter-specific dummies that will be used to control for unobserved characteristics of the countries in the sample, and [C.sub.h,ij] are dummy variables for former French and English colonies, which are expected to trade more intensively (in relative terms) with France and England than with the rest of the G-5 economies. [I.sub.k] contains regional dummies (Latin America, Africa, Asia), while [E.sub.k] contains individual exporter dummies for the G-5 countries. (11) The variable [n.sub.i] will be a measure of disaster incidence. A measure of catastrophic risk used in previous studies, the number of disasters per square kilometer, and a measure of the average loss caused by disasters as a percentage of GDP are used as disaster variables in Equation (3). (12) All variables are averages for the period 1976-1990. (13) Apart from Equation (3), we also estimate an alternative specification where an interaction term between catastrophic risk and the level of GDP per capita of the importing country is added to the specification since the overall level of development of a country is usually cited as the most relevant determinant of the vulnerability of an economy to the impact of natural disasters. On the other hand, countries with a higher level of GDP per capita are expected to possess more effective and sophisticated prevention and response programs. The interaction term is thus aimed at modeling this potential heterogeneity in the elasticity of technology trade to catastrophic risk caused by the relative differences in the level of development of the countries in our sample. This specification is given by:

(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [y.sub.i] refers to GDP per capita of country i. The source of the real GDP and real GDP per capita data is the World Bank's Worm Development Indicators database, disaster intensity data are from EM-DAT, and the data on distance between capital cities are computed using the "great circle" formula, as is usual when estimating gravity models.

Tables 2 and 3 present the cross-country results for our sample of developing countries. The simple gravity model augmented with the disaster variable explains almost 90% of the variation of technology transfer across developing countries for the period considered. The basic parameters of the gravity equation (those corresponding to log [Y.sub.i], log [Y.sub.j], and log [d.sub.ij]) are highly significant and present the expected sign (positive for the demand potential proxies and negative for the distance variable) in all cases. (14) When included in the gravity equation without an interaction term, the parameter estimates attached to the disaster variables are negative in all cases and highly significant for total catastrophic risk and climatic catastrophic risk. The parameter estimate for geologic disasters is negative but not significant. While the lack of significance of the geologic disasters variable can be traced back to the relatively low variability of this variable compared to climatic disasters, Skidmore and Toya (2002) argued that climatic disasters are a more reasonable proxy for physical capital-related catastrophic risk than geologic disasters since they tend to impact larger economic areas and occur periodically. The results for the disaggregation of natural disasters into climatic and geologic disasters resemble those in Skidmore and Toya (2002) in terms of significance, although the implications for the effect of catastrophic risk on technological upgrading are completely different. The estimates of Equation (3) imply that after controlling for the usual determinants of trade, relatively more disaster-prone countries tend to benefit less from R&D spillovers; thus, the creative destruction side of natural disasters referred to above does not seem to receive much empirical support. It should be noticed that the empirical evidence (see, e.g., Albala-Bertrand 1993a, 1993b) indicates that trade deficits tend to rise after a natural disaster, which implies that the results concerning the effects of natural disasters on R&D spillovers from the estimation of Equation (3) are probably picking up the differences in import composition more than the pure quantitative effect of increases in imports. The results are qualitatively similar if the measure of catastrophic risk used is the loss caused by natural catastrophes (Table 3). Using this measure of natural disaster intensity, the results imply that after conditioning on other determinants of trade, an increase of disaster loss over GDP by 1% decreases R&D spillovers by approximately 0.3%.

In order to control for the different level of vulnerability to natural catastrophes among the countries in our sample, we included an interaction term of the natural disaster variable with the overall level of development of the developing country (as proxied by its log-level of GDP per capita). The results for the model where the elasticity of the R&D spillover to catastrophic risk depends on the level of development of the developing country are included in Table 2 for the case of disaster frequency and in Table 3 for disaster loss. The results imply that the parameter estimate associated with the disaster variable turns positive for the countries in the highest two deciles of the distribution of GDP per capita but is only significant for those in the very end of the distribution. Figure 2 presents the implied elasticities for the sample used in the estimation together with the standard deviation of the estimates, using the frequency of disasters as an explanatory variable. The results are qualitatively similar for the case of disaster intensity presented in Table 3. (15)

The results from the cross-country regressions can be interpreted as measuring the overall effect of catastrophic risk on technological transfer in the long run. If we are interested in analyzing medium-term responses to natural disasters in terms of knowledge spillovers, we can make use of the time dimension of the data and estimate the model using a pooled sample by averaging subperiods. The time variation in catastrophic risk (as measured by the actual occurrence of disasters in the corresponding subperiods) can then be exploited, and short-run effects of natural disasters on R&D spillovers can be analyzed. Tables 4 and 5 present the results for the pooled sample with 5-yr subsamples, and Tables 6 and 7 present the results for the pooled sample with 3-yr subsamples. All variables in the estimation refer to averages in the respective subsamples, including in this case also the catastrophic risk proxies. The estimated models include common subperiod dummies, exporter dummies, and a finer set of importer regional dummies (with each broad regional group--Latin America, Asia, Africa--further divided into geographical subcategories). (16) The parameter estimates for the basic variables of the model do not change significantly when considering more disaggregated data in the time domain, and the gain in degrees of freedom results in more precise estimates. The results for the intensity of disasters (Tables 5 and 7) mirror those obtained with the cross-country sample, with lower elasticities in the response of R&D spillovers and no effect of geologic disasters. The same type of interaction with the level of development as in the cross-section estimations appears in all estimated models using the panel structure. The heterogeneity of parameters across countries seems to play a more determinant role for the medium-run response of technology transfer to natural disasters since the estimates of the panels including exclusively the frequency of disasters as an explanatory variable render insignificant results for total and climatic disasters, but when including the interaction with GDP per capita, the results are qualitatively similar to those found for the cross-section of countries.

[FIGURE 2 OMITTED]

The effect of geologic disasters on technological transfer appears now highly significant when using frequency of disasters as an explanatory variable in the panel, as opposed to the pure cross-country results. While geologic disaster risk did not have significant explanatory power in discriminating cross-country patterns of technological transfer in long-term horizons, the incidence of geologic disasters seems to lead to very sizable decreases in knowledge spillovers in the shorter run. This result is in line with the nature of geologic catastrophes, which tend to be less frequent and systematically more difficult to predict than climatic catastrophes.

The results indicate that there is empirical evidence of technological upgrading of equipment following catastrophic events in our sample of developing countries only for relatively developed countries (as measured by GDP per capita). Contrary to the view put forward in the recent literature on natural disasters and growth, catastrophic risk tends to affect technology absorption negatively, and the effect is stronger the less developed the country is. Furthermore, catastrophic risk variables tend to be significant determinants of cross-country differences in the long-run patterns of knowledge spillovers to developing countries.

IV. CONCLUSIONS

In this piece of research, we provide an empirical analysis aimed at measuring the impact of catastrophic risk on technology transfer to developing countries. Recent studies argue that natural disasters may serve as a means of creative destruction by providing an opportunity to upgrade capital equipment in disaster-prone countries, thus enabling higher long-run growth rates of GDP per capita. We test directly this hypothesis by means of gravity equations aimed at modeling the flow of knowledge embodied in imports from the G-5 countries to a sample including 49 developing countries. Knowledge transfer is proxied by constructing estimates of the R&D stock embodied in the imports of the developing countries in the sample, following the methodology developed for the analysis of R&D spillovers in the recent empirical literature on endogenous growth and technological transfer (Coe and Helpman 1995; Coe, Helpman, and Hoffmaister 1997). The results indicate that natural catastrophic risk is negatively related to the extent of technological transfer taking place between developed and developing countries. Interactions with income variables indicate that the level of development of the country has an effect on the elasticity of R&D spillovers to catastrophic risk, with richer countries eventually experiencing creative destruction after a disaster. We also provide results for the intensity of climatic and geologic disasters. While the intensity of climatic disasters is a significant determinant of medium- and long-run patterns of technological transfer and is negatively related to the size of the spillover, the results for geologic disasters are only significant and very sizeable in the medium-run recovery following the occurrence of a disaster.

Further paths of research in this topic could include the analysis of the effect of catastrophic risk on the absorption of foreign technology, defined as the actual implementation of the new technological knowledge, which is acquired through imports in the production process. Several recent contributions have stressed the importance of institutions in a nation's absorptive capacity for foreign technologies (Parente and Prescott 1994, 1999, 2003). Given the results of this article, the study of the influence of catastrophic risk on the process of absorption of technology may be viewed as an important point in the research agenda on the economics of natural disasters.

ABBREVIATIONS

GDP: Gross Domestic Product

TFP: Total Factor Productivity

APPENDIX: DATA SOURCES

The data on trade flows are taken from the Organisation for Economic Co-operation and Development (OECD) International Trade by Commodity Statistic; the domestic R&D stocks are computed out of R&D expenditures (taken from OECD's Research and Development Expenditure in Industry, Basic Science and Technology Statistics, and Main Science and Technology Indicators) using the perpetual inventory method with 5% depreciation (Coe and Helpman 1995). Disaster data are obtained from EM-DAT: The OFDA/CRED International Disaster Database (www.em-dat.net), Universite Catholique de Louvain, Brussels (Belgium), which contains data on disaster occurrence in the world from 1900 to the present. GDP and GDP per capita data are sourced from the World Bank's Worm Development Indicators data set.

JESUS CRESPO CUARESMA, JAROSLAVA HLOUSKOVA and MICHAEL OBERSTEINER *

* The authors would like to thank Jarko Fidrmuc, Neil Foster, Gordon Hanson, Landis MacKellar, Reinhard Mechler, Dennis Mueller, an anonymous referee, and participants at the Spring Meeting of Young Economists 2004 in Warsaw, at the Econometric Research seminar at the Institute for Advanced Studies, Vienna, and the Economic Research seminar at the University of Vienna for very helpful comments and discussion on earlier drafts of this article. The authors acknowledge financial support by the Oesterreichische Nationalbank's Jubilaumsfonds under Grant 10803.

REFERENCES

Aghion, P., and P. Howitt. "A Model of Growth through Creative Destruction." Econometrica, 60, 1992, 323 51.

--. Endogenous Growth Theory. Cambridge, MA: MIT Press, 1998.

Albala-Bertrand, J. "Natural Disaster Situations and Growth: A Macroeconomic Model for Sudden Disaster Impacts." Worm Development, 21, 1993a, 1417-34.

--. Political Economy of Large Natural Disasters. Oxford: Clarendon Press, 1993b.

Coe, D. T., and E. Helpman. "International R&D Spillovers." European Economic Review, 39, 1995, 859-87.

Coe, D. T., E. Helpman, and A. W. Hoffmaister. "North-South R&D Spillovers." Economic Journal, 107, 1997, 134-50.

Coe, D. T., and A. W. Hoffmaister. "Are there International R&D Spillovers among Randomly Matched Trade Partners? A Response to Keller." International Monetary Fund Working Paper WP/99/18, 1999.

Dacy, D. C., and H. C. Kunreuther. The Economics of Natural Disasters. New York: Free Press, 1969.

Eaton, J., and S. Kortum. "Measuring Technology Diffusion and the International Sources of Growth." Eastern Economic Journal, 22, 1996, 401-10.

Falvey, R., N. Foster, and D. Greenaway. "North-South Trade, Knowledge Spillovers and Growth." Journal of Economic Integration, 17, 2002, 650-70.

Fidrmuc, J. "The Core and Periphery of the World Economy." Journal of International Trade and Economic Development, 13, 2004, 89-107.

Grossman, G., and E. Helpman. Innovation and Growth in the Global Economy. Cambridge, MA: MIT Press, 1991.

Keller, W. "Are International R&D Spillovers Trade-Related? Analyzing Spillovers among Randomly Matched Trade Partners." European Economic Review, 42, 1998, 1469 81.

--. "International Technology Diffusion." Journal of Economic Literature, XLII, 2004, 752 82.

Koubi, V. "War and Economic Performance." Journal of Peace Research, 42, 2005, 67-82.

Lichtenberg, F. R., and B. van Pottelsberghe de la Potterie. "International R&D Spillovers: A Comment." European Economic Review, 82, 1998, 942 63.

Linder, S. An Essay on Trade and Transformation. Uppsala, Sweden: Almqvist and Wiksells, 1961.

Linnemann, H. An Econometric Study of International Trade Flows. Amsterdam, The Netherlands: North-Holland, 1966.

Okuyama, Y. "Economics of Natural Disasters: A Critical Review." Research Paper 2003-12, Regional Research Institute, West Virginia University, 2003.

Okuyama, Y., G. J. D. Hewings, and M. Sonis. "Measuring Economic Impacts of Disasters: Interregional Input-Output Analysis using Sequential Interindustry Model, " in Modeling the Spatial and Economic Effects of Disasters, edited by Y. Okuyama and S. E. Chang. New York: Springer, 2004, 77-101.

Organski, A. F. K., and J. Kugler. "The Cost of Major Wars: The Phoenix Factor." American Political Science Review, 71, 1977, 1347-66.

Parente, S. L., and E. Prescott. "Barriers to Technology Adoption and Development." Journal of Political Economy, 102, 1994, 298-321.

--. "Monopoly Rights: A Barrier to Riches." American Economic Review, 89, 1999, 1216-33.

--. Barriers to Riches. Cambridge, MA: MIT Press, 2003.

Romer, P. "Increasing Returns and Long-Run Growth." Journal of Political Economy, 94, 1986, 1002-37.

Schumpeter, J. A. Capitalism, Socialism and Democracy. New York: Harper, 1950.

Skidmore, M. "Risk, Natural Disasters, and Household Saving in a Life Cycle Model." Japan and the Worm Economy, 13, 2001, 15-34.

Skidmore, M., and H. Toya. "Do Natural Disasters Promote Long-Run Growth?" Economic Inquiry, 40, 2002, 664-87.

Tol, R., and F. Leek. "Economic Analysis of Natural Disasters," in Climate, Change and Risk, edited by T. E. Downing, A. J. Olsthoorn, and R. S. T. Tol. London: Routledge, 1999, 308-27.

United Nations Development Programme. Reducing Disaster Risk: A Challenge/or Development. New York: Bureau for Crisis Prevention and Recovery, United Nations, 2004.

Crespo Cuaresma: Professor of Economics, Department of Economics, University of Innsbruck, Universitatstrasse 15, 6020 Innsbruck, Austria. Phone +43 (0) 512 507 7357, Fax +43 (0) 512 507 2980, E-mail jesus.crespo-cuaresma@uibk.ac.at

Hlouskova: Department of Economics and Finance, Institute for Advanced Studies, Stumpergasse 56, 1060 Vienna, Austria. Phone +43 1 59991-142, Fax +43 1 59991-555, E-mail hlouskov@ihs.ac.at

Obersteiner: International Institute for Applied Systems Analysis (IIASA), Schlossplatz 1, 2361 Laxenburg, Austria. Phone +43 2236 807-460, Fax +43 2236 71313, E-mail obestei@iiasa.ac.at

(1.) Skidmore and Toya (2002) argued that natural disasters may also affect long-run economic growth by altering the investment decisions concerning both physical and human capital. The relatively lower return to physical capital investment caused by an increased probability of natural disasters may be responsible for a shift toward human capital investment. At the same time, if this shift takes place, one may expect a subsequent increase in the return of physical capital. For a model of saving decisions under catastrophic risk, see Skidmore (2001).

(2.) One may argue that pure Schumpeterian creative destruction takes place after a natural disaster if the damage done to a productive sector leads to a change at the organizational level, thus changing the competitive environment faced by firms in an industry.

(3.) An excellent survey of the literature on trade-related technology spillovers can be found in Keller (2004).

(4.) The choice of developing countries in the sample is determined by data availability.

(5.) See the Appendix for the sources of the data used in the analysis.

(6.) Catastrophic events are reported, which fulfill at least one of the following criteria: (a) ten or more people reported killed, (b) 100 people reported affected, (c) a call for international assistance was issued, or (d) a state of emergency was declared.

(7.) This is not the only criticism that has been raised to the measure by Coe and Helpman (1995). Lichtenberg and van Pottelsberghe de la Potterie (1998) and Keller (1998) criticize the choice of weights by Coe and Helpman (1995). See also Coe and Hoffmaister (1999) for a response to criticism by Keller (1998).

(8.) Falvey, Foster, and Greenaway (2002) proposed an alternative explanation of the importance of the intensity of trade in explaining R&D spillovers based on whether the knowledge embodied in imports is a public or a private good.

(9.) The proposed measure depends obviously on the size of the receptor country. The size of the importing and exporting countries will be conditioned upon in the framework of gravity equations.

(10.) It is well known that the world distribution of R&D spending is extremely skewed (see, e.g., Keller 2004). In 1997, the most developed economies accounted for 84% of total world R&D expenditures, and just two countries (the United States and Japan) accounted for 61% of that amount. This justifies concentrating on a relatively small set of countries, which report relatively accurate figures on R&D expenditures as "exporters of technology."

(11.) In the final specification, only those dummy variables with significant parameters are actually included.

(12.) For the case of disaster frequency, the log transformation introduced above was used. We also performed estimations using the number of affected persons as a measure of disaster intensity. The results are similar to those for disaster loss as percentage of GDP and are thus not reported. They are available from the authors upon request.

(13.) The results are not significantly affected if catastrophic risk is measured for the period 1960-1990.

(14.) The estimation of the basic specification rendered residuals with significant non-Gaussian features, as measured by the Jarque-Bera test statistic. The deviation from normality was caused by three observations, which were dummied out for the results presented in Table 2. The parameter estimates are not affected by the exclusion of these observations, and the residuals of the resulting estimation without these three observations could not reject normality of the residuals at the usual significance levels. The problem of non-normality of residuals in estimated gravity models has recently been pointed out by Fidrmuc (2004).

(15.) Further interactions including polynomials of GDP per capita were tried in order to assess the existence of nonlinearities in the relationship between development, technology transfer, and catastrophic risk. No evidence of such nonlinearities was found.

(16.) In the panel specification, the following dummies were tried: South America, South Asia, Central Africa, Central America, Caribbean, North Africa, West Africa, South-East Asia, West Asia, East Africa, East Asia, and South Africa. Only those which appeared significant are included in the final specifications.

TABLE 1
Developing Countries in the Sample

Argentina
Bangladesh
Bolivia
Brazil
Central African Republic
Chile
Cameroon
Colombia
Costa Rica
Dominican Republic
Algeria
Ecuador
Ghana
Guatemala
Guyana
Honduras
Haiti
Indonesia
India
Israel
Jamaica
Kenya
Korea
Kuwait
Mexico
Malawi
Malaysia
Niger
Nicaragua
Pakistan
Panama
Peru
Philippines
Paraguay
Sudan
Senegal
Sierra Leone
El Salvador
Sri Lanka
Togo
Thailand
Trinidad and Tobago
Tunisia
Uruguay
Venezuela
South Africa
Zaire
Zambia
Zimbabwe

TABLE 2
Cross-Country Gravity Equations with Natural Disaster Frequency

GDP import 0.82 *** (0.04) 0.81 *** (0.04)

GDP export 0.35 ** (0.16) 0.35 ** (0.16)

Distance -0.91 *** (0.09) -0.89 *** (0.09)

Total natural disaster -0.69 ** (0.29) -5.81 *** (1.72)
per [km.sup.2]

Total natural disaster per -- 0.73 *** (0.24)
[km.sup.2] x GDP per capita

Total climatic disaster -- --
per [km.sup.2]

Total climatic disaster per -- --
[km.sup.2] x GDP per capita

Total geological disaster -- --
per [km.sup.2]

Total geological disaster -- --
per [km.sup.2] x GDP per
capita

Adjusted [R.sup.2] 0.88 0.89

Observations 245 245

GDP import 0.82 *** (0.03) 0.81 *** (0.03)

GDP export 0.35 ** (0.18) 0.35 ** (0.17)

Distance -0.91 *** (0.09) -0.89 *** (0.09)

Total natural disaster -- --
per [km.sup.2]

Total natural disaster per --
[km.sup.2] x GDP per capita

Total climatic disaster -0.71 ** (0.34) -5.82 *** (2.23)
per [km.sup.2]

Total climatic disaster per -- 0.73 ** (0.32)
[km.sup.2] x GDP per capita

Total geological disaster -- --
per [km.sup.2]

Total geological disaster -- --
per [km.sup.2] x GDP per
capita

Adjusted [R.sup.2] 0.88 0.89

Observations 245 245

GDP import 0.84 *** (0.03) 0.83 *** (0.03

GDP export 0.34 ** (0.17) 0.34 ** (0.17)

Distance -0.89 *** (0.09) -0.88 *** (0.09)

Total natural disaster
per [km.sup.2]

Total natural disaster per
[km.sup.2] x GDP per capita

Total climatic disaster --
per [km.sup.2]

Total climatic disaster per --
[km.sup.2] x GDP per capita

Total geological disaster -0.82 (1.67) -62.74 *** (19.94)
per [km.sup.2]

Total geological disaster -- 8.41 *** (2.67)
per [km.sup.2] x GDP per
capita

Adjusted [R.sup.2] 0.88 0.89

Observations 245 245

Notes: Robust standard errors are given in parenthesis. The symbols
"**" and "***" stand for 5% and 1% significant. The specifications
include importer region dummies, exporter dummies, and colonial
dummies.

TABLE 3
Cross-Country Gravity Equations with
Natural Disaster Intensity (Loss as % GDP)

GDP import 0.83 *** (0.03) 0.80 *** (0.03)

GDP export 0.34 ** (0.16) 0.34 ** (0.16

Distance -0.90 *** (0.08) -0.87 *** (0.08)

Natural disaster -0.28 *** (0.06) -2.79 *** (0.51)
loss (% GDP)

Natural disaster loss -- 0.37 *** (0.07)
(% GDP) x GDP per capita

Climatic disaster -- --
loss (% GDP)

Climatic disaster loss -- --
(% GDP) x GDP per capita

Geological disaster -- --
loss (% GDP)

Geological disaster loss -- --
(% GDP) x GDP per capita

Adjusted [R.sup.2] 0.89 0.90

Observations 245 245

GDP import 0.82 *** (0.03) 0.81 *** (0.03

GDP export 0.35 ** (0.16) 0.35 ** (0.16)

Distance -0.87 *** (0.08) -0.87 *** (0.08)

Natural disaster -- --
loss (% GDP)

Natural disaster loss -- --
(% GDP) x GDP per capita

Climatic disaster -0.31 *** (0.07) -2.66 *** (0.54)
loss (% GDP)

Climatic disaster loss -- 0.35 *** (0.08)
(% GDP) x GDP per capita

Geological disaster -- --
loss (% GDP)

Geological disaster loss -- --
(% GDP) x GDP per capita

Adjusted [R.sup.2] 0.89 0.90

Observations 245 245

GDP import 0.85 *** (0.03) 0.83 *** (0.03)

GDP export 0.34 ** (0.16) 0.34 ** (0.16)

Distance -0.90 *** (0.09) -0.89 *** (0.09)

Natural disaster -- --
loss (% GDP)

Natural disaster loss -- --
(% GDP) x GDP per capita

Climatic disaster -- --
loss (% GDP)

Climatic disaster loss -- --
(% GDP) x GDP per capita

Geological disaster -0.06 (0.08) -8.79 ** (4.36)
loss (% GDP)

Geological disaster loss -- 1.19 ** (0.60)
(% GDP) x GDP per capita

Adjusted [R.sup.2] 0.88 0.88

Observations 245 245

Notes: Robust standard errors are given in parenthesis. The symbols
"**" and "***" stand for 5% and 1% significant. The specifications
include importer region dummies, exporter dummies, and colonial
dummies.

TABLE 4
Gravity Equations: 5-Yr Pooled Sample with Natural Disaster Frequency

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.42 *** (0.10) 0.43 *** (0.10)

Distance -1.19 *** (0.07) -1.18 *** (0.07)

Total natural disaster -0.39 (0.40) -7.02 ** (2.88)
per [km.sup.2]

Total natural disaster -- 0.95 ** (0.40)
per [km.sup.2] x GDP
per capita

Total climatic disaster -- --
per [km.sup.2]

Total climatic disaster -- --
per [km.sup.2] x GDP
per capita

Total geological -- --
disaster per [km.sup.2]

Total geological -- --
disaster per [km.sup.2]
x GDP per capita

Observations 735 735

Adjusted [R.sup.2] 0.84 0.84

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.42 *** (0.10) 0.42 *** (0.10)

Distance -1.19 *** (0.07) -1.18 *** (0.07)

Total natural disaster -- --
per [km.sup.2]

Total natural disaster -- --
per [km.sup.2] x GDP
per capita

Total climatic disaster -0.29 (0.42) -6.98 ** (2.96)
per [km.sup.2]

Total climatic disaster -- 0.96 ** (0.41)
per [km.sup.2] x GDP
per capita

Total geological -- --
disaster per [km.sup.2]

Total geological -- --
disaster per [km.sup.2]
x GDP per capita

Observations 735 735

Adjusted [R.sup.2] 0.84 0.84

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.43 *** (0.10) 0.43 *** (0.10)

Distance -1.19 *** (0.07) -1.18 *** (0.07)

Total natural disaster -- --
per [km.sup.2]

Total natural disaster -- --
per [km.sup.2] x GDP
per capita

Total climatic disaster -- --
per [km.sup.2]

Total climatic disaster -- --
per [km.sup.2] x GDP
per capita

Total geological -5.23 * (2.77) -82.11 *** (30.44)
disaster per [km.sup.2]

Total geological -- 10.72 ** (4.28)
disaster per [km.sup.2]
x GDP per capita

Observations 735 735

Adjusted [R.sup.2] 0.84 0.84

Notes: Robust standard errors are given in parenthesis. The symbols
and "***" stand for 10%, 5%, and 1% significant. The specifications
include importer region dummies, exporter dummies, and colonial
dummies.

TABLE 5
Gravity Equations: 5-Yr Pooled Sample with
Natural Disaster Intensity (Loss as % GDP)

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.42 *** (0.10) 0.42 *** (0.10)

Distance -1.19 *** (0.07) -1.18 *** (0.07)

Natural disaster loss -0.06 *** (0.02) -0.64 ** (0.27)
(% GDP)

Natural disaster loss -- 0.08 ** (0.04)
(% GDP) x GDP per capita

Climatic disaster loss -- --
(% GDP)

Climatic disaster loss -- --
(% GDP) x GDP per capita

Geological disaster loss -- --
(% GDP)

Geological disaster loss -- --
(% GDP) x GDP per capita

Observations 735 735

Adjusted [R.sup.2] 0.84 0.84

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.42 *** (0.10) 0.42 *** (0.10)

Distance -1.18 *** (0.07) -1.18 *** (0.07)

Natural disaster loss -- --
(% GDP)

Natural disaster loss -- --
(% GDP) x GDP per capita

Climatic disaster loss -0.07 *** (0.02) -0.58 ** (0.27)
(% GDP)

Climatic disaster loss -- 0.08 * (0.04)
(% GDP) x GDP per capita

Geological disaster loss -- --
(% GDP)

Geological disaster loss -- --
(% GDP) x GDP per capita

Observations 735 735

Adjusted [R.sup.2] 0.84 0.84

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.42 *** (0.10) 0.43 *** (0.10)

Distance -1.19 *** (0.07) -1.19 *** (0.07)

Natural disaster loss -- --
(% GDP)

Natural disaster loss -- --
(% GDP) x GDP per capita

Climatic disaster loss -- --
(% GDP)

Climatic disaster loss -- --
(% GDP) x GDP per capita

Geological disaster loss -0.02 (0.04) -2.47 * (1.27)
(% GDP)

Geological disaster loss -- 0.34 * (0.18)
(% GDP) x GDP per capita

Observations 735 735

Adjusted [R.sup.2] 0.84 0.84

Notes: Robust standard errors are given in parenthesis. The symbols
"**" and "***" stand for 5% and 1% significant. The specifications
include importer region dummies, exporter dummies, and colonial
dummies.

TABLE 6
Gravity Equations: 3-Yr Pooled Sample with Natural Disaster Frequency

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.44 *** (0.08) 0.44 *** (0.08)

Distance -1.19 *** (0.05) -1.18 *** (0.05)

Total natural disaster -0.67 (0.44) -9.91 *** (3.53)
per [km.sup.2]

Total natural disaster -- 1.30 *** (0.49)
per [km.sup.2] x GDP
per capita

Total climatic disaster -- --
per [km.sup.2]

Total climatic disaster -- --
per [km.sup.2] x GDP
per capita

Total geological -- --
disaster per [km.sup.2]

Total geological -- --
disaster per [km.sup.2]
x GDP per capita

Observations 1,225 1,225

Adjusted [R.sup.2] 0.83 0.83

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.43 *** (0.08) 0.43 *** (0.08)

Distance -1.19 *** (0.05) -1.18 *** (0.05)

Total natural disaster -- --
per [km.sup.2]

Total natural disaster -- --
per [km.sup.2] x GDP
per capita

Total climatic disaster -0.56 (0.45) -9.90 *** (3.66)
per [km.sup.2]

Total climatic disaster -- 1.31 *** (0.50)
per [km.sup.2] x GDP
per capita

Total geological -- --
disaster per [km.sup.2]

Total geological -- --
disaster per [km.sup.2]
x GDP per capita

Observations 1,225 1,225

Adjusted [R.sup.2] 0.83 0.83

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.44 *** (0.08) 0.44 *** (0.08)

Distance -1.19 *** (0.05) -1.19 *** (0.05)

Total natural disaster -- --
per [km.sup.2]

Total natural disaster -- --
per [km.sup.2] x GDP
per capita

Total climatic disaster -- --
per [km.sup.2]

Total climatic disaster -- --
per [km.sup.2] x GDP
per capita

Total geological -6.58 ** (3.05) -83.59 *** (31.97)
disaster per [km.sup.2]

Total geological -- 10.72 ** (4.42)
disaster per [km.sup.2]
x GDP per capita

Observations 1,225 1,225

Adjusted [R.sup.2] 0.83 0.83

Notes: Robust standard errors are given in parenthesis. The symbols
"**" and "***" stand for 5% and 1% significant. The specifications
include importer region dummies, exporter dummies, and colonial
dummies.

TABLE 7
Gravity Equations: 3-Yr Pooled Sample with
Natural Disaster Intensity (Loss as % GDP)

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.43 *** (0.08) 0.43 *** (0.08)

Distance -1.19 *** (0.05) -1.18 *** (0.05)

Natural disaster -0.04 *** (0.01) -0.49 *** (0.18)
loss (% GDP)

Natural disaster loss -- 0.07 ** (0.03)
(% GDP) x GDP per capita

Climatic disaster loss -- --
(% GDP)

Climatic disaster loss -- --
(% GDP) x GDP per capita

Geological disaster loss -- --
(% GDP)

Geological disaster loss -- --
(% GDP) x GDP per capita

Observations 1,225 1,225

Adjusted [R.sup.2] 0.83 0.83

GDP import 0.84 ***(0.02) 0.84 *** (0.02)

GDP export 0.43 ***(0.08) 0.43 *** (0.08)

Distance -1.19 *** (0.05) -1.18 *** (0.05)

Natural disaster -- --
loss (% GDP)

Natural disaster loss -- --
(% GDP) x GDP per capita

Climatic disaster loss -0.05*** (0.01) -0.44 ** (0.20)
(% GDP)

Climatic disaster loss -- 0.06 ** (0.03)
(% GDP) x GDP per capita

Geological disaster loss -- --
(% GDP)

Geological disaster loss -- --
(% GDP) x GDP per capita

Observations 1,225 1,225

Adjusted [R.sup.2] 0.83 0.83

GDP import 0.84 *** (0.02) 0.84 *** (0.02)

GDP export 0.43 *** (0.08) 0.43 *** (0.08)

Distance -1.19 *** (0.05) -1.19 *** (0.05)

Natural disaster -- --
loss (% GDP)

Natural disaster loss -- --
(% GDP) x GDP per capita

Climatic disaster loss -- --
(% GDP)

Climatic disaster loss -- --
(% GDP) x GDP per capita

Geological disaster loss -0.05 (0.02) -1.04 (0.74)
(% GDP)

Geological disaster loss -- 0.14 (0.10)
(% GDP) x GDP per capita

Observations 1,225 1,225

Adjusted [R.sup.2] 0.83 0.83

Notes: Robust standard errors are given in parenthesis. The symbols
and "***" stand for 10% 5%, and 1% significant. The specifications
include importer region dummies, exporter dummies, and colonial
dummies.