Sectoral Volatility, Trade Partners, and Development

Transcription

1 Sectoral Volatility, Trade Partners, and Development Ridwan Karim Draft: 28 June, Introduction and Literature Review This paper investigates the transmission channels through which shocks in domestic demand and output of a country s trading partners may affect the volatility of that country s output growth at the sectoral level. By decomposing sectoral volatility of output growth in terms of volatility in domestic consumption growth, export growth and import growth, I investigate whether the export and import channels of a country acts as a mechanism through which shocks from other countries can be transmitted to a country. The baseline empirical specification is based on a statistical variance decomposition exercise. I find that volatility of demand of importing countries are carried over to an exporting country through trade, and volatility of output of exporting countries are also carried over to an importing country through trade. Both these channels are robust to alternate specifications, though the second channel has a larger magnitude. This finding provides a hitherto unexplored explanation as to why the average developing country become far more volatile due to the same increase in trade openness compared to the average developed country. Developing countries are characterized by higher demand and output volatility, and since the bulk of trade by developing countries involve other developing countries, I contend that increase in trade openness lead to higher levels of volatility for developing countries because they become exposed to shocks from their more volatile trade partners. Several influential articles in the relevant literature have identified economic volatility as having important consequences for a country s economic welfare. Using a sample of 92 countries and a separate sample for OECD countries, Ramey and Ramey (1995) found that countries with

2 higher volatility have lower growth. The addition of standard control variables strengthens the negative relationship between volatility and growth. Lucas (1988) observes that developed countries exhibit stable growth rates over long periods of time, whereas poorer countries are prone to sharp fluctuations in growth rates. These two results in conjunction seem to imply that macroeconomic volatility has graver implications for developing countries. Pallage and Robe (2003) report that the welfare cost of macroeconomic volatility is much higher in developing countries than it is in developed countries. They find that that the welfare cost of consumption volatility is far from trivial and averages a substantial multiple of the corresponding U.S. estimate. They also claim that in many poor countries, the welfare gain from eliminating volatility may in fact exceed the welfare gain from an additional percentage point of growth forever in many poor countries. By using data from 40 industrial and developing countries, Mendoza (1997) reports that terms of trade volatility has a large adverse effect on economic growth. He also finds that this effect is robust to the addition of the other key determinants of economic growth. Laursen and Mahajan (2005) point out that economic volatility can deepen poverty and inequality. The effect is particularly strong for developing countries, where good times are not able to reverse the downward impacts of bad times, leading to long-term problems. Guillaumont and Korachais (2006) claim that there is evidence, particularly in African countries, that macroeconomic instability makes growth less pro poor. They argue that argue that income instability may result in higher poverty for a given level of income or a lower reduction of poverty for a given growth of income. Furthermore, by investigating 97 developing countries over the period , Guillaumont, Korachais and Subervie (2006) find that macroeconomic instability may actually lead to lowering the probability of child survival in developing countries. They find that a given income growth can have very differing impacts on reducing child mortality, depending on the volatility associated with that growth. These papers have led to an extensive discussion in the literature regarding the main determinants of economic volatility, particularly the microeconomic underpinnings that lead to macroeconomic volatility.

3 One determinant of macroeconomic volatility that has received a lot of attention in the literature is the role played by trade openness. Rodrik (1997) argued that trade openness can play a role in affecting macroeconomic volatility. The International Labor Organization (ILO), in their 2004 report ILO, A Fair Globalization: Creating Opportunities for All, discussed several examples of how globalizing economies were becoming increasingly more exposed to global shocks due to a larger integration between countries. Of course, there is a huge volume of empirical evidence in the economic literature which has stressed the importance of trade as a major driver of economic growth and development (UNIDO, 2009). However, it appears that the relationship between trade and development is complex, and that it is not true that trade will automatically lead to economic growth for developing countries. One reason being the strong positive relationship found between trade openness and macroeconomic volatility. Giovanni and Levchenko (2009) empirically explored three channels through which trade openness may affect output growth volatility: sectoral volatility, specialization and comovement between domestic sectors. They report that, added together, the three channels imply that the relationship between trade openness and overall volatility is positive and economically significant. This result seems to be at odds with certain theoretical predictions made in the literature regarding the impact that trade may have on a country s output volatility. For example, using a calibrated version of the Eaton-Kortum and Alvarez-Lucas model, Caselli, Koren, Lisicky and Tenreyro (2011) argue that the impact of trade on volatility can be negative, because trade becomes a source of diversification if country-wide shocks are dominant. Haddad, Lim, Pancaro, Saborowski (2013) also claim that trade can reduce volatility if countries are well diversified in terms of their export baskets. It appears that currently there is no consensus, either empirically or theoretically, on the nature of the relationship between trade openness and macroeconomic volatility. A number of papers have investigated the role of sectoral fluctuations in causing aggregate volatility. By looking at the sectoral and national aggregate disturbances to industrial output in seven European countries, Stockman (1988) reports that shocks that are nation-specific and common to industries are important, and cast doubt on the hypothesis that most

4 macroeconomic fluctuations can be ascribed to shocks to technology. Horvath (1998) argues against the traditional contention that positive shocks in some sectors are offset by negative shocks in other sectors as dictated by the law of large numbers. He hypothesized that the interactions among producing sectors play an important role, and used simple statistical analysis to demonstrate that 80% the volatility in U.S. gross domestic product growth rates could be the result of independent shocks to 2-digit Standard Industrial Code sectors. Carvalho (2008) employs a theoretical model to show how fluctuations in aggregate economic activity can be obtained from independent shocks to individual sectors. Gabaix (2011) also argued against the idea that individual firm shocks average out in the aggregate, by proposing that idiosyncratic firm-level shocks can explain an important part of aggregate movements and provide a microfoundation for aggregate shocks. He used a model of firm-size heterogeneity to argue his case. It therefore appears that there is a general consensus that volatility at the sector-level has important consequences for aggregate volatility. This paper explores certain mechanisms through which trade openness may affect growth volatility through its impact on sectoral volatility. As can be seen, there is a lack of consensus in the literature regarding the nature of the relationship between trade openness and macroeconomic volatility. Most of the empirical work has focused on broad variables like trade openness, or the extent of export portfolio diversification. One channel that has been unexplored in the literature is the effect that the characteristics of trading partners may have on sectoral volatility. I contend that, holding the level of income and extent of export diversification constant, shocks to demand and output in partner countries captured in volatility measures will be transmitted through trade to other countries. A corollary is that developing countries tend to trade with partners that are more volatile, which may explain a part of the story as to why trade openness leads to greater volatility in countries with lower level of percapita income.

5 2. Empirical Strategy, Issues and Data 2.1 Baseline Specification In Giovanni and Levchenko (2009), the authors postulate three channels through which trade openness can affect output volatility. These channels are: sector-level volatility, specialization and comovement. The paper decomposed variance of aggregate output growth into variances and covariances of output growth in individual industries in the following manner: Here, refers to aggregate volatility, while, and refer to sectoral volatility, comovement and sector shares respectively. Incidentally, these three variables correspond to the three channels mentioned previously: sector-level volatility, specialization and comovement, respectively. The authors then proceed to analyze the effect of trade openness on, and. My paper will focus only on the first channel, i.e. the sector-level volatility arising from trade. To compute the effect of trade openness on this variable, Giovanni and Levchenko run a regression with the following specification: ( ) Here, the subscript i, c, and t refer to industry, country and year respectively. The trade openness measure at the industry-level is computed by summing overall imports and exports for a given industry in a given country, and then dividing it by the overall output of that particular industry. This specification assumes that only the overall imports and exports matter for the variance of industry output. However, my proposition is that this variance may also depend on characteristics of trading partners. If countries have different levels of domestic supply and demand variances, then exporting to or importing from more volatile trading partners may have a stronger impact on. I carry out the following variance decomposition

6 exercise to pinpoint the channels through which demand and output volatility in partner countries can affect sectoral volatility of a home country. Using the market clearing condition,, domestic output growth,, can be written as a weighted-share of growth in demand, exports and imports, Here,, and represent the shares of domestic demand, export and import in output respectively. Thus, the variance of output growth for industry i (omitting country and industry indices for brevity) can be written as: Trade will affect sectoral output variance through the last five terms. Because equation (2) is an identity and not the result of a structural model, it cannot be used to infer causality. I am simply employing this identity to motivate the channels through which shocks from partner country can be carried over through trade. Notice that this equation implicitly captures the notion of trade openness. If a country is in autarky, then the last five terms will drop out, and only volatility of domestic demand will matter for output volatility. Conversely, trade will matter more if the shares of export and import in output are higher. Notice also that this specification suggests that trade might lead to decreased volatility if the last two terms dominate the other three terms affected by trade. For example, if the covariance between demand and import is high enough, then shocks to demand can be matched by corresponding changes to imports, mitigating the overall impact of the demand shock on sectoral volatility. To illustrate the specific channels through which trade may affect sectoral volatility, I will focus on the second term in the right-hand side of equation (2). The term can be expressed as a weighted-average as follows: ( ) ( )

7 Where j is a subscript for partner country j, and is variance of home country s exports of product i to partner country j, and is the share of exports to country j out of home country s total exports in industry i. X is total exports of product i and Y is total output of product I for home country. I am interested in transmissions of supply/demand shocks across countries through trade. For example, I postulate that the variance of export of product i to country j will be affected by the variance of demand of product i in country j. In other words, my hypothesis is that sectoral demand volatility in partner countries will be transmitted to home country through the variance of home country s exports in that sector. Therefore, to examine these transmission channels, I assume a linear relationship between and the variance of country j s demand for product i in the following manner: While this assumption is an intuitive one, it is necessary to examine to what extent the intuition bears out in the data. I therefore construct import and demand volatilities at the sectoral level for every country in my sample (explained in more detail in the data section below). I require a positive association between the sectoral import and demand volatilities for my assumption to be valid, since home s export volatility to a particular partner is the importing country s import volatility, which is being substituted by the importing country s demand volatility. The scatter plot for the import and demand volatilities can be seen in Figure 1. The positive association is very robust and statistically significant at the 1% level. Thus, the data provides support for the assumption in equation 4.

8 Figure 1: Scatter Plot of Sectoral Demand and Import Volatilities Plugging the expression from equation (4) back into equation (3), I obtain, ( ) ( ) Where ( ) ( ). Equation (5) demonstrates how demand volatility in partner countries can affect home/origin country s sectoral volatility by affecting that country s export volatility. Demand volatility in partner countries ( and the square of export shares to partner countries ( ) can be constructed from the data, allowing the estimation of. If is estimated to be positive, then that would imply that demand volatility of an importing country is carried over to an exporting country through trade. Notice also that this formulation captures the notion of diversification. If a country is exporting to only one other country within a given industry, then the partner country's demand volatility will be assigned a higher weight, and thus will matter more. Conversely, trading with more partners will diminish the impact of demand volatility of any one partner country.

9 Similarly, I postulate that output volatility in import partner will affect sectoral volatility of home country through its effect on import volatility, captured by the third term in the righthand side of equation (2). Thus, my second assumption is that variance of import of product i from country j is related to variance of output of product i in country j, which can be written as: As before, I empirically verify this assumption by plotting the corresponding sectoral export and output volatilities, as shown in Figure 2. The assumption holds in the data: the positive association is very robust and statistically significant at the 1% level. Figure 2: Scatter Plot of Sectoral Output and Export Volatilities Therefore, by following an argument similar to the one above, the third term can be expressed as follows: ( ) ( ) Equation (6) demonstrates how output volatility in partner countries can affect home country s sectoral volatility by affecting that country s import volatility. If is estimated to be positive, then that would imply that output volatility of an exporting country is carried over to an importing country through trade.

10 The last three terms in equation (2) can be rewritten as follows: ( ) ( ) ( ) ( ) And, ( ) ( ) Therefore, I can use the following equation to estimate equation (2), This is the equation I wish to estimate. Based on the identity in equation (2), my expectations are that and. 2.2 Potential Endogeneity of Trade Shares The explanatory variables in equation (10) are basically interaction terms of various shares and variances/covariances. One concern is the potential endogeneity between the shares and volatility variables, e.g. between and, or between and. It may be argued that shares of exports are influenced by future expectations of demand volatility in potential destinations, or shares of imports are influenced by expectations of output volatility of destination countries. If the share variables adjust endogenously as a response to expected volatilities, then there may be a downward bias in the coefficients to be estimated from equation (10).

11 To overcome this problem, I plan to instrument the various export and import shares using the gravity model. I want to use the approach of Do and Levchenko (2007) who extended the methodology of Frankel and Romer (1999) to industry-level data. The main idea is to use a gravity model to predict the observed trade flows between every pair of countries using predetermined geographic variables such as distance, population and other standard covariates of trade costs used in gravity models. By allowing all the coefficients to vary across industries, it is possible to construct a predicted value of trade between every country pair for every industry and use it to calculate the natural, i.e.exogenous, export/import shares. Since most of the geographic variables do not vary a lot over time, the predicted import shares will have no time dimension. This approach will bring me closer to interpreting the results as causal. I reproduce the standard gravity model below. The coefficients to be estimated are indexed by i to highlight the fact that they will be allowed to vary across industries: [ ] [ ] [ ] [ ] [ ] [ ] Here, is exports of product i by country c to country d in year t, is the distance between trading partners, is a dummy variable which takes the value of one if countries c and d share the same border and zero otherwise, is the land area of country j, is population of country j in year t, and is a dummy variable for country j s landlocked status. Equation (11) predicts trade flows with geographic and demographic factors only, and the predicted is thus independent of any economic influences. With industry-specific coefficients, model (11) can predict export shares, which I plan to use for estimating equation (10): For this instrument to be valid, it has to be fairly strongly correlated with the observed shares in the data, something that can only be verified once I have constructed the instruments. Given

12 that the shares are only determined by geographic and demographic variables, and not by economic factors, I believe that it is reasonable to claim that the exclusion restriction will hold in this case. In fact, instrumenting export/import shares using this method is a standard practice in empirical trade papers, in order to overcome the endogeneity of observed import / export shares. As a robustness test, I would also like to estimate predicted trade shares from the standard gravity model in the spirit of Anderson and Van Wincoop (2001) where multilateral resistance terms are proxied with two sets of country-year fixed effects ( one each for importing and exporting countries). Country-year fixed effects introduce economic factors into the model since they capture the effect of aggregate prices, aggregate demand and supply shocks, and other macroeconomic variables. For this reason, this specification may make the explanatory variables more likely to violate the exclusion restriction. However, it might be a worthwhile exercise, in case the estimated shares based on equation (11) do not turn out to be strong instruments. I discuss briefly the issue of simultaneity in the next section. 2.3 Concerns Regarding Simultaneity Another concern that I have relates to simultaneity bias. I claim in equation (10) that output volatility at home is driven by the output volatility of home s import partners. However, as is often the case, the same countries trade both ways within a given sector, implying that home is also likely to export to its import partners within the same sectors. Thus, as an exporter, home s output volatility will also affect the output volatility of its import partners, which in turn will affect home s output volatility. This potentially leads to the problem of a simultaneity bias, leading to the coefficients estimated from a simple OLS regression being biased. One potential solution is to look at small countries trading with big ones where reverse causality is unlikely to be a major issue. Ideally, I would like to instrument for the output volatility of home s import partners. One instrument that I plan to use is the sector-specific trade openness of home s import partners, controlling for home s own trade openness within

13 that sector. As Giovanni and Levchenko (2009) reports, and I verify in the next section, trade openness is strongly positively associated with sectoral volatility, and is a plausibly exogenous instrument if home s trade openness is controlled for. This is a step that I have not carried out yet in my empirical analysis. 2.4 Data For this paper, I use three separate databases. The Production data comes from he Trade, Production and Protection Database (Nicita and Olarreaga, 2006), which is available online and can be freely accessed through the World Bank trade website. The Trade, Production and Protection Database merges trade flows, production and trade protection data available from different sources into a common classification: the International Standard Industrial Classification (ISIC). The database covers 40 industries in 173 developing and developed countries over the period The database includes information on the value of annual production of each country in each of these sectors, the annual value of exports to other countries for each of these countries within each sector/industry, and the value of annual imports from other countries for each of these countries within each of these sectors. Domestic demand within each sector can be constructed by subtracting overall exports and adding overall imports to the total domestic output in that sector. Data on bilateral trade flows is obtained from the NBER-United Nations Trade Data constructed by Robert Feenstra and Robert Lipsey. The data spans over the years across 40 manufacturing sectors with unique 3-digit ISIC identifier (SITC converted to ISIC based on Marc- Andreas Muendler's concordance tables). The database provides values of trade flows between 201 countries in each of these sectors during this period. Data on geographic and demographic variables (for the purpose of instrumenting observed trade shares) comes from the CEPII Gravity Dataset. This dataset covers all world pairs of countries for the period The GDP and populations come from the World Bank Development Indicators (WDI), while bilateral distances and common (official) language come

14 from the CEPII distance database. It also contains data on geographic areas of countries and which Regional Trade Agreements they are part of. 3. Preliminary Results I first run the Giovanni and Levchenko (2009) specification for the regression of sectoral volatility on trade openness to see if my results are consistent with their findings. I split samples into , and , to allow for time variation. I construct the volatility of output growth for each sector and each country. I only keep those volatilities which have been calculated by using at least 5 observations. I run the following regression: ( ) This is the specification that Giovanni and Levchenko used to assess the impact of trade openness on sectoral volatility. However, if my hypothesis is correct, the coefficient in the regression above should be different for each country based on the specific characteristics of a country s trading partner. To see whether this may indeed hold true, I split the sample in countries with higher than the average productivity of all countries in my dataset, and countries with lower than the average productivity of all countries. I run the above regression for the full sample, for the higher productivity sample and for the lower productivity sample separately. The results can be seen in table 1. Table 1: Sectoral Volatility on Trade Openness Dependent Variable: Sectoral Volatility (1) (2) (3) Full Sample Higher Productivity Sample Lower Productivity Sample Trade Openness (Trade/Output) 0.065*** 0.025*** 0.075*** (0.005) (0.010) (0.005) Observations R-Squared Fixed Effects Country Fixed Effects Country Fixed Effects Country Fixed Effects Notes: Robust Standard Errors in parentheses. *significant at 10%;**significant at 5%; ***significant at 1%

15 As expected, trade openness has a statistically significant and positive impact on sectoral volatility. This supports the Giovanni and Levchenko assertion that sectors more open to international trade are more vulnerable to shocks in global supply and demand. However, my coefficients are not as large as the ones that Giovanni and Levchenko report. It should be mentioned here that my data set does not cover all the time periods that the dataset employed by Giovanni and Levchenko does, and perhaps that explains the differences in the coefficients. However, the effect of trade openness on sectoral volatility is markedly different for the high productivity and the low productivity samples. It appears that trade openness has a greater impact on sectoral volatility for countries with less than average productivity in my dataset. This is consistent with my proposition that trade openness can have variable impacts on countries. It also demonstrates that developing countries are likely to be more adversely impacted by trade openness in terms of sectoral volatility, compared to developing countries. In my next step, I construct partner-specific trade openness measures to see whether trade openness has variable impact on sectoral volatility depending on the partner characteristics. The results are reported below: Table 2: Trade Openness Regressions Dependent Variable: Sectoral Output Volatility Trade Openness (Trade/Output) Openness to Trade with Developed Countries Openness to Trade with Developing Countries (1) (2) *** ( ) *** ( ) *** ( ) Observations Notes: Robust standard errors in parentheses. *significant at 10%;**significant at 5%; ***significant at 1% As can be seen, trade openness with respect to developing countries has a higher impact on sectoral volatility, compared to trade openness with developed countries. This is consistent with my hypothesis that trade openness will have heterogenous impacts depending on the

16 volatility of a country s trade partners. I conducted the Hausman test for equality of the two coeffcieints reported in the second column in Table 2, and the null hypothesis was rejected at the 5% level. Thus, the impact of trade openness is statistically different, depending on which countries home is trading with. I then constructed the five main explanatory variables for my model, based on equation (5)-(9). For the shares of exports and imports, I took the average of export/import shares over the time-frame considered to construct these variables. I only keep those variances and covariances which have been calculated by using at least 5 observations. This means that I have to drop a significant number of observations in my dataset before I can estimate equation (10). I run the following regression as implied by equation (10) on a sample of 973 observations comprising of 61 countries and 28 sectors / industries. I include time trends in the regression. The results are given in Table 3. Table 3: Baseline Specification Dependent Variable: Sectoral Output Volatility Demand Volatility of Export Partners Output Volatility of Import Partners Covariance between Domestic Demand and Demand of Export Partner Covariance between Domestic Demand and Output of Import Partner Cavoariance between Demand of Export Partner and Output of Import Partner (1) (2) (3) (4) (5) * * ** *** *** ( ) ( ) ( ) ( ) ( ) *** *** *** *** ** ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) Country Fixed Effects No Yes No Yes Yes Decade Fixed Effects No No No No Yes Sector Fixed Effects No No No No No Clustering Level Country Country Country-Industry Country-Industry Country-Industry Observations Notes: Robust Clustered standard errors in parentheses. *significant at 10%;**significant at 5%; ***significant at 1%

17 As can be seen from Table 3, the index of demand volatility of export partners, and output volatility of import partners are consistently significant, and the magnitude of the coefficients are fairly stable across different specifications. These are the two variables that I consider to be of first-order importance, given that they indicate the direct channels through which shocks can be transmitted to a country through its trade partners. In particular, the demand volatility of export partners seem to have a greater effect on sectoral volatility. Given that developing countries are characterized by higher demand volatility, and that they primarily trade with other developing countries, this previously unexplored channel appears to be particularly important. It should be kept in mind that this is simply the OLS regression based on variables as they appear in the naïve regression model. Thus, the estimates are likely to be biased, as the potential endogeneity have not been taken into account. However, these results give some indication of the validity of the idea that characteristics of the partner countries affect sectoral volatility. 4. Next Steps My immediate next step would be to construct the instrument variables for the export and import shares using the gravity model, and verifying whether the robustness condition holds. As mentioned previously, I expect the exclusion restriction to hold as these shares are going to be constructed based only on geographic and demographic variables, and not economic factors. I then plan to tackle the issue of simultaneity by constructing another set of instruments for the output volatility of import partners. Another potential problem is the measurement error associated with the data. Particularly, there are grounds for concern about the reliability of the data reported by developing countries. One solution would be to trim the data at the top and bottom 10% to observe whether outliers are influencing the results.

18 References Anderson, James E., and Eric Van Wincoop. Gravity with gravitas: a solution to the border puzzle. No. w8079. National bureau of economic research, Carvalho, Vasco, Aggregate fluctuations and the network structure of intersectoral trade, 2008 Meeting Papers 1062, Society for Economic Dynamics. Caselli, Francesco, et al. "Diversification through trade." Society for Economic Dynamics Meeting Papers. No Do, Quy-Toan, and Andrei A. Levchenko. "Comparative advantage, demand for external finance, and financial development." Journal of Financial Economics86.3 (2007): Frankel, Jeffrey A., and David Romer. "Does trade cause growth?." American economic review (1999): Gabaix, Xavier, The Granular Origins of Aggregate Fluctuations," Econometrica, May 2011, 79 (3), Giovanni, Julian di, and Andrei A. Levchenko. "Trade openness and volatility."the Review of Economics and Statistics 91.3 (2009): Haddad, Mona, et al. "Trade openness reduces growth volatility when countries are well diversified." Canadian Journal of Economics/Revue canadienne d'économique 46.2 (2013): Horvath, Michael, Cyclicality and Sectoral Linkages: Aggregate Fluctuations from Independent Sectoral Shocks," Review of Economic Dynamics, October 1998, 1 (4), ILO, A Fair Globalization: Creating Opportunities for All, World Commission on the Social Dimension of Globalization Report, Geneva, Switzerland (2004). Koren, Miklos, and Silvana Tenreyro. "Volatility and development." The Quarterly Journal of Economics (2007):

19 Laursen, Thomas, and Sandeep Mahajan, Volatility, Income Distribution, and Poverty (pp ), in Joshua Aizenman and Brian Pinto (Eds.), Managing Economic Volatility and Crises: A Practitioner s Guide (New York: Cambridge University Press, 2005). Lucas, R. E. J., On the mechanics of economic development, Journal of Monetary Economics, XXII (1988), Mendoza, E.G Terms-of-trade Uncertainty and Economic Growth, Journal of Development Economics, 54: Nicita, A. and M. Olarreaga, "Trade, Production and Protection ", World Bank Economic Review 21(1) Pallage, S. and Robe,. M. A. (2003), On the Welfare Cost of Economic Fluctuations in Developing Countries. International Economic Review, 44: doi: / t Patrick Guillaumont, Catherine Korachais, Julie Subervie. How Macroeconomic Instability Lowers Child Survival Ramey, Garey, and Valerie A. Ramey, Cross-Country Evidence on the Link Between Volatility and Growth, American Economic Review 85:5 (1995), Rodrik, Dani, Has Globalization Gone Too Far? (Washington, DC: Institute for International Economics, 1997). Stockman, Alan C., Sectoral and National Aggregate Disturbances to Industrial Output in Seven European Countries," Journal of Monetary Economics, 1988, 21, UNIDO Growth, Exports and Technological Change in Developing Countries: Contributions from Young Scholars. Research and Statistics Branch, Working Paper No26/2009