International and intranational business cycles.
Two important measures of business cycles are the comovement of economic variables and their volatilities. While volatility issues, such as the smoothness of consumption and the magnitude of investment fluctuations, have been continuously subject to examination within business cycle research, comovement issues have not always been emphasized. For example, time-series studies often concentrate on a single variable, such as total output, to characterize the business cycle -- e.g. Hamilton (1989) and Hess and Iwata (1997). An early definition of the business cycle provided by Burns and Mitchell (1946) highlighted that expansions occurred simultaneously across many economic activities, followed by similarly general recessions, contractions, and revivals, which then merged into the expansion phase of the next cycle. Fortunately, the early tradition of emphasizing the comovement of economic variables at business cycle frequencies has been revived in real business cycle (RBC, hereafter) theory.
Beginning with Kydland and Prescott (1982), RBC theory initially concentrated on the closed economy and tried to construct a dynamic general equilibrium model which mimics the business cycle features of the real economy. These models generate artificial data, such as output, consumption, investment, and employment, from a model economy whose fluctuations are driven solely by technological shocks. Researchers then compare the comovement and volatility of these simulated series with those present in actual data. Many proponents of RBC models tend to judge the success or failure of a model solely based on how close these two measures calculated from artificial data are to their real data counterparts.
RBC models for a closed economy have been generalized to examine international business cycles. The most comprehensive examination has been conducted by Backus, Kehoe, and Kydland (1992, 1993, 1994, 1995) -- hereafter, BKK. As the international trade in goods and capital flows have increased in volume and importance throughout history, researchers have begun to consider how the business cycle in one country can influence the business cycle in other countries. This comovement of the same variables across countries is termed `the international business cycle'.
Since theory develops as it tries to explain phenomena which existing models cannot, studies of international business cycles have concentrated on stylized facts which are inconsistent with the available theory. Importantly, BKK (1995) point out two anomalies between theory and data for international business cycle research: the quantity anomaly and the price variability anomaly. The first anomaly is as follows. In a two-country RBC model, where fluctuations of output are driven solely by exogenous technological shocks, since individuals in each country can offset their nation-specific output uncertainty by holding diversified international portfolios, the cross-country correlation of consumption should be much higher than the cross-country correlation of output or technological shocks. In the data, however, they find that the opposite is true. The second discrepancy between theory and data that they point to is that the terms of trade are much more volatile in the data than RBC models suggest.
To identify the causes of these anomalies, and especially to see whether borders between countries are responsible for them, we examine data generated across regions within a single country. This approach is appealing since two countries in the international model are introduced in the same way as two regions or sectors are modelled within a single country. Hence the model has exactly the same implications for the comovement of variables across regions. This idea of using data for regions within a country as a natural experiment for interpreting results for data across countries has generated a research agenda termed `the intranational business cycle'. Different from the research on the closed-economy RBC approach, which analyses the comovement of different variables in an economy, research on the intranational business cycle emphasizes the comovement of a variable across different regions.
In a broader context, intranational business cycles provide more than just a `proving ground' for international business cycle theories. In fact, in the 21st century, international business cycles are likely to look increasingly more like intranational business cycles. For instance, throughout the European Union, both the barriers to trade in goods as well as the migration of capital and labour have fallen substantially as these policy-induced distortions between countries have vanished. Moreover, these countries are also considering taking the coordination of monetary policy to its ultimate extreme -- the adoption of a single currency. But this blurring of national economic boundaries is not just a European phenomenon. Recently, the North America Free Trade Agreement (NAFTA) has also begun to integrate the economies of Canada, Mexico, and the United States, with perhaps Chile to follow. It seems, therefore, that an understanding of the empirics of intranational business cycles may prove useful for predicting the features of international business cycles in the future.
In section II we explore more carefully the findings from the international business cycle literature. In section III we present evidence for intranational business cycles within the USA, with some results for other countries. Finally, in section IV, we discuss the implications of integrating the two research agendas for policy issues.
II. INTERNATIONAL BUSINESS CYCLES
Below we discuss the important aspects of international business cycles, especially as they have been highlighted by RBC theory. Accordingly, as the single impulse in a traditional business cycle is productivity fluctuations, we begin with a discussion of this. We then proceed to examine more carefully the evidence for the quantity and price anomalies identified by BKK.
(i) World Business Cycles?
Recently, many studies document business cycle similarities and international comovement of output among developed countries. Stockman (1988) decomposes the quarterly and annual growth rate of industrial production in ten manufacturing industries for eight countries -- Germany, France, Italy, Belgium, The Netherlands, the United Kingdom, Switzerland, and the United States -- into industry-specific shocks and country-specific shocks. Industry-specific shocks are defined as changes in industrial output growth that are unique to an industry but common to all countries in the sample; country-specific shocks are defined as changes in output that are specific to a country but shared by all industries in that country. For example, we can think of shocks which are common to the textile industry in all countries in the world (industry-specific shocks to the textile industry) and shocks which are common to all industries in the UK (country-specific shocks to the UK). Stockman finds that both industry-specific shocks and country-specific shocks are statistically important as well as economically meaningful in explaining national aggregate industrial production growth rates. More recently Gregory et al. (1995) also decompose fluctuations in aggregate output for the G7 countries into a worldwide component and country-specific component, and find that both components are significant both statistically and economically.
If we interpret fluctuations in output to be primarily due to technological innovation, as RBC models suggest, then the above studies present two possibilities. First, the importance of country-specific shocks implies that technological innovation in one industry spreads across different industries in the same country rapidly. On the other hand, the importance of industry-shocks and worldwide shocks shows the possibility that technological innovation spreads rapidly within the same industries in different countries and induces a worldwide business cycle. In an attempt to further clarify these issues, Costello (1993) measures technological growth directly, using Solow residuals which are calculated as the percentage change in output minus the percentage change in inputs, where the inputs are weighted by their factor shares.(2) She studies Solow residuals from 1960 to 1985 in five manufacturing industries (food, textiles, chemicals, basic metals, and metal manufacturing) for six countries (Canada, Germany, Italy, Japan, the UK, and the USA). She finds that short-run technological growth is more similar across industries within one county than across countries for a particular industry. Even though the studies by Stockman (1988) and Gregory et al. (1995) find that both country-specific and worldwide industry-specific shocks are important, Costello's (1993) finding suggests that technological innovation spreads more rapidly across different industries in the same country. Hence the world business cycle based solely on technological shocks does not seem to be characterized by a single technological innovation governing the world economy; rather, it is more likely to be due to the interactions between countries through trade and capital flows, where each country is subject to continuous technological shocks which have some limited influence on one another. Further research is needed, therefore, to establish the empirical relationships of productivity fluctuations between and across industries and countries, and to make models and the data more consistent.
(II) The International Comovement of Output and Consumption
In this section we summarize the features of international comovement of business cycles as well as within-country business cycles. Tables 1 and 2 summarize stylized facts of the business cycle for 17 OECD countries using annual data from 1963 to 1992.(3) While Table 1 reports average statistics for the 17 OECD countries, Table 2 reports some essential statistics for each individual country.
Table 1 Mean Statistics for 17 OECD Countries, 1963-92
Variable Correlation Volatility Correlation across countries with output Output GNP 0.311 2.777 1.000 (0.320) (0.766) GDP 0.288 2.609 0.987 (0.330) (0.686) (0.013) Technology (GNP) 0.229 2.203 0.891 (0.321) (0.677) (0.085) Technology (GDP) 0.200 (2.096) 0.866 (0.399) (0.627) (0.098) Consumption Total 0.277 2.298 0.808 (0.283) (0.736) (0.098) Private 0.283 2.723 0.852 (0.275) (0.842) (0.071) Prices Price deflator 0.762 5.907 -0.323 (0.146) (2.347) (0.235) Terms of trade 0.229 5.472 -0.222 (0.322) (2.692) (0.263) Variable Persistence Output GNP 0.651 (0.113) GDP 0.649 (0.111) Technology (GNP) 0.608 (0.137 Technology (GDP) 0.605 (0.127) Consumption Total 0.687 (0.109) Private 0.671 (0.118) Prices Price deflator 0.886 (0.036) Terms of trade 0.597 (0.117)
Notes: Each series has been detrended using the Hodrick-Prescott (1997) filter, with the smoothing parameter set to 400. Volatility is measured by the time-series standard deviation. Column one reports the mean correlation across 17 OECD countries. Column two reports the mean volatility of state variables. Column three reports the mean correlation of the country variables with GNP of the country. Column four reports the volatility of the variables. Estimated standard deviations of the statistics are in parentheses.
Table 2 Statistics for Comovement and Volatility for 17 OECD Countries, 1963-92
Cross-correlation of Variable Output Consumption Australia 0.378 0.245 Austria 0.323 0.195 Belgium 0.388 0.350 Canada 0.348 0.392 Denmark 0.343 0.219 Finland 0.255 0.262 France 0.435 0.386 Germany 0.210 0.122 Ireland 0.123 0.289 Italy 0.334 0.297 Japan 0.355 0.333 The Netherlands 0.417 0.349 New Zealand 0.070 0.208 Norway -0.049 -0.032 Sweden 0.384 0.402 UK 0.361 0.255 USA 0.296 0.169 Volatility of Variable Output Terms of trade Consumption Australia 2.119 8.191 1.001 Austria 2.188 2.904 1.570 Belgium 2.503 2.841 2.401 Canada 2.518 4.323 2.120 Denmark 2.372 3.817 2.087 Finland 4.030 4.292 2.858 France 1.971 4.285 1.138 Germany 3.128 5.352 3.754 Ireland 4.047 5.279 3.432 Italy 2.226 6.042 1.714 Japan 4.541 11.749 2.799 The Netherlands 2.627 1.788 2.710 New Zealand 2.848 9.476 2.820 Norway 2.214 9.156 2.401 Sweden 2.255 3.247 1.810 UK 2.518 5.280 2.313 USA 2.246 5.005 2.138
Notes: Each series has been detrended using the HP filter, with the smoothing parameter set to 400. Column one reports the mean correlation of per capita GNP across the other 16 OECD countries. Column two reports the mean correlation of per capita consumption across the other 16 OECD countries. Column three reports the volatility of per capita GNP of the country. Column four reports the volatility of terms of trade. Column five reports the volatility of per capita consumption. Volatility is measured by the time-series standard deviation.
The variables we consider are two measures of output (GNP and GDP), two measures of technological shocks (Solow residuals corresponding to the two measures of output), two measures of consumption (total consumption expenditure which includes government spending, and private consumption expenditure), the GDP price deflator, and the terms of trade. Since a good measure of the capital stock for various countries is not available and the capital stock contributes little to the fluctuations of the economy, we use only labour employment data to calculate the Solow residual.(4) Following BKK (1995) we set the capital share equal to 0.36. To consider the data's cyclical patterns, we have detrended the data both by using the Hodrick-Prescott (HP) filter (1997) and by analysing the data in growth rates. Since we find that these two approaches provide quite similar results, we report results based only on the HP-filtered data.
Table 1 is separated into the first column which reports evidence of the international comovement of variables and the last three columns which report within-country evidence of the business cycle. The first column reports the mean of the cross-correlations of the variables between two countries -- the average of the 17 x 16 pairs in the sample without double counting. Column two presents the mean standard deviation of the variables across time within one country. The former is, therefore, a measure of the average contemporaneous relationship of the variable across countries, whereas the latter is an average measure of the intertemporal volatility of the series for all countries. The third column presents the mean correlation of the country's series with its own output as a measure of the series' cyclicality. Finally, column four presents the data's persistence, which is calculated as the first order autocorrelation.
Table 2 reproduces some important statistics in Table 1 for each individual country. Columns one and two present the average comovement of output (GNP) and consumption (total consumption) for a particular country with that for other countries. For example, we calculate the cross-correlations of consumption for Australia with that of the other 16 OECD countries and report the average of them for Australia. Columns three, four, and five report the volatilities of output, terms of trade, and consumption, respectively, for each country.
While closed-economy business cycle features for aggregate data are summarized well elsewhere, we point out two stylized facts that are relevant for our purpose.(5) First, when we compare the volatility of output to total consumption, consistent with the permanent-income hypothesis, there is evidence that consumption (2.298) is smoother than either measure of output -- GNP (2.727) or GDP (2.609). The permanent-income hypothesis predicts that individuals choose consumption based on the income which they expect to persist into the future (i.e. their permanent income), and use savings and borrowing to smooth consumption in response to transitory variations in income. While private consumption does not show the evidence of smoothness as much as total consumption, its volatility is still lower than that of GNP. We believe that the reason for the rather weak evidence of the lower volatility of consumption is the fact that the consumption measure includes durable goods in addition to nondurable goods. BKK (1995) report that for the countries where data are available, the ratio of the volatility of consumption to that of output decreases when durables are excluded: from 0.85 to 0.59 for Canada, from 0.99 to 0.77 for France, from 0.78 to 0.61 for Italy, from 1.15 to 0.96 for the UK, and from 0.75 to 0.52 for the USA.(6) Second, the price level is countercyclical; its correlation coefficient with output (GNP) is -0.323. This evidence of a countercyclical price level has been emphasized by Cooley and Ohanian (1991) for the USA and Chadha and Prasad (1994) for the G7 countries.
In BKK's extension of the RBC model to an international environment, they construct a two-country model where each country is subject to exogenous technological shocks.(7) If technological shocks (hence output) are not perfectly correlated across countries, then, since an individual in each country can share risk internationally (for example by holding a stock portfolio that is internationally diversified), the cross-country correlation of consumption should be much higher than the cross-country correlation of output or technological shocks. In the data, however, they find that just the opposite is true. Similarly, Obstfeld (1994) analyses the cross-correlations for consumption and output growth rates between an individual country and the rest of the world, as well as between OECD countries. He reports that, although the cross-correlations of consumption growth rates have risen since 1973, they are always lower than those for output. Since this difference between theory and data is extremely robust to changes in the parameter values of the model, BKK (1995) term this a quantity anomaly.
Even though we cover more countries and a longer sample period using a different data set, we obtain the same evidence. The international comovement of variables are summarized in the first column of Table 1. The correlation of consumption across countries (0.277 or 0.283) is on average lower than that of output (0.311 or 0.288). The first two columns of Table 2 show the average cross-correlation of output (GNP) and consumption (total consumption) for each country with the other countries in the sample. We find that in six out of 17 countries, the cross-correlation of consumption is higher than that of output, but the difference is marginal. In the other countries, we have strong evidence that the cross-correlation of output is higher than that for consumption.
The evidence that the cross-correlation of consumption is lower than that of output is contrasted with the finding that the volatility of consumption is lower than that of output discussed earlier. As emphasized by Hess and Shin (1997), to understand the contemporaneous and intertemporal relationships between consumption and output, it is helpful to distinguish between intertemporal and intratemporal (contemporaneous) risk sharing. Altug and Miller (1990) and Cochrane (1991) point out that the permanent-income hypothesis (intertemporal risk sharing) and consumption insurance (intratemporal risk sharing) are distinct propositions. In the context of our study, intratemporal risk sharing analyses the representative household's ability in each country to insure against idiosyncratic shocks to their national output -- this increases the contemporaneous cross-correlation of consumption above that for output. Alternatively, intertemporal risk sharing studies a representative household's ability to smooth consumption over time -- this reduces the intertemporal volatility of consumption relative to that for output.
Our basic argument is as follows: since consumption is less volatile over time than output, then there is evidence of intertemporal risk sharing (i.e. the permanent-income hypothesis). However, since the cross-correlation of consumption is quite low, there is much less evidence of intratemporal risk sharing (consumption insurance). Since a household can smooth consumption overtime without owning stock in foreign firms, evidence of consumption being too uncorrelated across countries is evidence that households' portfolios are insufficiently diversified internationally.(8) This `home bias problem' has been discussed widely with respect to the lack of international diversification in households' portfolios (e.g. Baxter and Jermann, 1997).
There hive been various attempts theoretically to introduce other factors ignored in BKK's basic model in order to generate a lower cross-correlation of consumption than that of output, although to date none has been very successful. The first approach points out that the standard RBC model assumes technological shocks solely responsible for business fluctuations. In this case, if we assume a homogeneous individual in both countries, then it would be optimal for individuals to consume equally across countries by averaging their outputs. Hence, there will be perfect correlation of consumption across countries. However, if an economy is faced with other shocks, such as taste shocks which imply different preferences for consumption in different countries, then the cross-correlation of consumption will be lowered even within an equilibrium model. Incorporating taste shocks formally into the basic model, Stockman and Tesar (1995) find that the cross-correlation of consumption can be lower than that of output. However, this raises two important issues: how can economists justify taste shocks as another source of fluctuations as they are unobservable. If we cannot quantify taste shocks from the data, it is hard to be convinced that the observed lower cross-correlation of consumption is in fact due to taste shocks. Second, if random taste shocks play a large role in driving important relationships in an otherwise intertemporal, optimizing model, this directly calls into question the basis for using utility-maximizing models in the first place.
The second approach emphasizes the existence of non-tradable goods. Since risk sharing across countries cannot be made through non-tradable consumption goods, the cross-correlation of consumption of non-tradable goods across countries will be lowered. If consumption of non-tradable goods affects consumption of tradable goods, differences in consumption of non-tradable goods will lead to differences in consumption of tradable goods and hence lower cross-correlation of tradable consumption goods. Stockman and Tesar (1995) adopt this approach, but find that introducing non-tradable goods alone does not lower the cross-correlation of consumption enough, so it is still higher than that of output in the model.
Finally, incomplete international capital markets can lower the cross-correlation of consumption because the ability to share risk across countries will be lowered. Incomplete international capital markets imply that there is some limitation on the individual's ability to share risk internationally. But the incompleteness, if it is not entirely incomplete, also implies that there is some partial ability for individuals to share risk internationally. If the individual has some ability to share risk internationally, then the cross-correlation of consumption, if any, will be higher than that of output. Conze et al. (1993) and Kollmann (1996) show that, with incomplete markets, the cross-correlation of consumption decreases but is still predicted to be higher than that for output.
(iii) International Relative Prices
When we analyse the international business cycle there appears one more price variable to consider -- the terms of trade, defined as the ratio of the implicit price deflators for imports and exports. According to this definition; it measures the relative price of imported goods, and hence it is the inverse of the conventional definition. However, since this definition is more consistent with the concept of the real exchange rate used in the study of international macroeconomics, we remain with this definition. Salient business cycle features of the terms of trade for developed countries are summarized in Mendoza (1995). These features include the relationship between the terms of trade and net exports. Three relevant business cycle features of the terms of trade are reported in Table 1. First, it is weakly countercyclical; the average correlation with output is-0.222 for 17 OECD countries. Second, it is quite persistent; the average first-order autocorrelation is about 0.6 on quarterly data. Third, it is highly volatile; on average its volatility is 2.2 times higher than that of output.
BKK (1995), in a two-country model, find that if fluctuations are driven solely by technological shocks, the model mimics most properties of the terms of trade except that its volatility is very low as compared with actual data; the volatility of the terms of trade for the theoretical model is about one-seventh of that observed in US data, and, as a bench-mark, about one-third of volatility of output generated in the theoretical model. Since the model's prediction that the volatility of terms of trade is much lower than that of output is quite robust to parameter changes of the model, they term this inconsistency of theory and data the terms of trade, or price variability, anomaly. Another `price' inconsistency between theory and data, although not emphasized by the authors, is that the terms of trade generated from the model is quite procyclical, while the actual data show that the terms of trade is weakly countercyclical (-0.222).
In contrast to the model's theoretical predictions, we report in Table 2 that the volatility of output (column three) is lower than that of the terms of trade (column four) except for The Netherlands. The ratio of the volatility of the terms of trade and that of output is at maximum of over four for Norway. For both the UK and the USA, it is higher than two.
There have been fewer attempts made to explain the price variability anomaly. However, a natural extension might be introducing money (more realistically, multiple currencies) into the model. Potentially, money plays two roles to contribute to high volatility of the terms of trade. First, since asset markets and goods markets are separated, additional shocks to asset markets change currency prices and hence lead to more variable terms of trade. Second, since nominal prices and real prices are separated, sticky nominal prices in one country contribute to less synchronized international prices and hence a more volatile terms of trade. Of course, an early objective of RBC models was to mimic the data without any currencies, and therefore an adaptation of the RBC model to satisfy the price anomaly could be thought to be against the spirit of the approach.
III. INTRANATIONAL BUSINESS CYCLES
While the study of international business cycles has provided a number of discrepancies between theory and data, one can argue that the models are unrealistic as they ignore the frictions inherent in international economics. Hence, business cycle models that do not account for these frictions will always be incapable of replicating the data. To overcome this limitation, researchers have begun to analyse business cycles within a country as a way of understanding business cycles across countries.
There are two main reasons to believe that analysing intranational business cycles is a natural experiment for understanding international business cycles. First, regions within a country have no tariffs or quotas between them to impede trade and there are no capital market restrictions to reduce cross-regional capital flows. As such, there are far fewer trade frictions for goods or capital within countries than across them. Second, regions within a country share a common currency. Therefore, to the extent that multiple currencies create non-fundamental volatility in exchange rates which distorts cross-country economic activity, analysing business cycles within a country circumvents these additional frictions which RBC models do not incorporate. Together, these points suggest that intranational data should provide a more sympathetic environment for understanding business cycles (especially real ones) across geographic regions.
(i) Sources of Intranational Business Cycles
As emphasized above, it is convenient to summarize business cycle features by the comovement and volatilities of the relevant series. With respect to measuring business cycles within countries, we maintain the RBC perspective that productivity measures are paramount when understanding output fluctuations. From the international literature, recall Costello's (1993) important finding: productivity growth is more correlated across industries within one country than across countries for a particular industry. This suggests that industry productivity growth is characterized more by nation-specific factors than worldwide industry-specific factors. Based on this, a number of researchers have attempted to see what role geography and industries play in intranational productivity. Kollmann (1995) examines intranational productivity for the USA, using data provided by Hulten and Schwab (1984) as well as more recent data, and obtains the opposite stylized fact. He finds that cross-region correlations of productivity growth within a given industry are typically larger than cross-industry correlations of productivity growth within the same region.(9) He conjectures that, because of the close integration of US regions, industry-specific technological innovations spread more rapidly across regions than across independent countries.(10)
Independently, both Hess and Shin (1995) and Ghosh and Wolf (1996) examine intranational productivity growth within the USA using disaggregate state-level annual data from 1963 to 1991 for a number of industries. The following results are from Hess and Shin, and are representative of the literature. The seven industries examined are: mining, construction, manufacturing, transportation, trade, services, as well as the finance, insurance, and real estate industry. The disaggregate productivity growth measures for industry i, in state j at time period t are: output growth ([e.sup.G.sub.ijt]), the Solow residual using just the labour input ([e.sup.L.sub.ijt]),the Solow residual using the labour input and the aggregate growth in capital for that industry ([e.sup.KL.sub.ijt]), and output growth per worker ([e.sup.LP.sub.ijt]).(11)
Each disaggregate productivity measure ([e.sub.ijt]) can be decomposed into an aggregate, industry-specific, state-specific, and idiosyncratic component as follows:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
The first term ([e.sub.at]) represents the aggregate shock, which is the weighted average of the productivity shocks across the whole economy. The second term, the industry productivity measure ([e.sub.it]), is the weighted average of the productivity growth in that industry across all states minus the aggregate shock. The third term, the state productivity measure ([e.sub.jt]) is the weighted average productivity a state receives less the average shock which would occur for the state if each industry of the state received the average industry shock. Our adjustment eliminates any effects to the state-specific shock which are due to the different composition of industries among states. The final term, [u.sub.ijt], is the residual term which reflects idiosyncratic shocks.(12)
To answer the question of whether industries or regions are more correlated within a country, Table 3 presents the average cross-correlations for productivity across states (top panel) and industries (middle panel). Although we use a longer sample period, a finer geographical classification (states instead of regions), and a wider set of productivity growth measures, Kollmann's result holds -- industries are more correlated than states. This is demonstrated by the fact that the average cross-correlation of productivity growth across states within a given industry is greater than the average cross-correlation of productivity growth across industries within the state, namely corr ([e.sub.ij], [e.sub.ij],\i) [is greater than] corr ([e.sub.i,j], [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. Therefore, industries comove owing to national factors rather than to state factors. To-understand the source of this comovement across states, it is worth noting that the state-specific shocks, which have been corrected for the state's industrial structure, are uncorrelated across states (i.e. corr ([e.sub.j], [e.sub.j]') is essentially zero). Together with the finding of a high correlation of economic activity at the state level as presented by corr ([e.sub.ij], [e.sub.ij],\i), this indicates that the comovement is driven by industrial composition.
[TABULAR DATA NOT REPRODUCIBLE IN ASCII]
Does geography not matter for understanding intranational business cycles? If there are geographic contemporaneous externalities from either the supply or demand side, then one would expect that economic activity in bordering (neighbouring) states would be more highly correlated than non-neighbouring ones. The bottom panel of Table 3 presents the average correlations for bordering states (B) and non-bordering states (NB). These findings demonstrate that state level activity is more correlated for neighbours as opposed to non-neighbours. The most striking results are for the correlation of the state shocks in the bottom two rows of the panel. The finding that state shocks are more correlated across neighbours is important since these shocks control for the possibility that they maybe correlated simply because of their similar industrial bases.
As emphasized above, to understand business cycle fluctuations a researcher needs to understand volatility as well as comovement. To quantify what drives the volatility in productivity, we decompose the variance of the disaggregate productivity into the variance of its components and covariance terms. The first column of Table 4 presents the variances from the decomposition of the four productivity measures (the covariance terms are negligible and are not reported). Note that only for this table is volatility measured by its variance. The reason for this is that, neglecting covariance terms which are almost zero, the variance of the sum equals the sum of the variances -- a property that standard deviations do not share. The volatility decomposition reveals two important stylized facts. First, the variance of idiosyncratic shocks, ([u.sub.ijt]), dominates for all productivity measures, accounting for about three-quarters of the variation in disaggregate productivity. Second, the variation at the industry level is much greater than at the state level.
[TABULAR DATA NOT REPRODUCIBLE IN ASCII]
To provide a more complete description of the intranational business cycle, columns two and three of Table 4 present the cyclicality and persistence of productivity just as we did for international data in Table 1. For cyclicality, the disaggregate productivity measures ([e.sub.ijt]) are found to be pro-cyclical (approximately 0.4), while their industry, state, and idiosyncratic components are acyclical. This suggests that cyclical behaviour can be well approximated by national cycles. The results for persistence indicate that economic activity becomes more persistent at higher levels of aggregation: the aggregate shock is most persistent, the industry and state shocks are less persistent, and the idiosyncratic shock displays no persistence.
Consistent with these findings, a number of researchers (e.g. Norrbin and Schlagenhauf, 1988; Clark, 1997) have used richer econometric techniques to separate out the effects of nation and region-specific and industry-specific factors in employment growth for the USA.(13) Using a dynamic indicator model, Norrbin and Schlagenhauf report steady-state estimates of the relevant contributions of each factor: 58 per cent of employment variation is explained by national factors, 34 per cent by industry factors, less than 1 per cent for international factors, and only 4 per cent each for regional and idiosyncratic factors. Clark uses aggregate data for regions and industries within the USA and finds that regions have a more important role to play in explaining US employment fluctuations than is found in disaggregate data. He finds that 40 per cent of the variance of the business cycle is region-specific, even after controlling for industry mix, and that regional shocks are propagated across regions. This is consistent with the results presented in the bottom panel of Table 3 and the extensive research on intranational productivity by Ghosh and Wolf (1996).
(ii) The Intranational Comovement of Output and Consumption
The quantity anomaly pointed to by BKK (1995) found that, for international data, consumption growth rates were less correlated across countries than output. This suggests that households are insufficiently diversifying the country-specific risk that they face. A possible explanation for this is that households strongly prefer to hold assets from their home country -- the so-called `home bias' problem -- which discourages them from completely diversifying their risk.(14) of Course, one can also argue that financial markets are not complete enough to allow households to diversify their country-specific risk.
In order better to understand consumption risk sharing, a number of researchers have recently examined consumption across regions within a country. These studies have explored the cross-correlations of consumption and output across geographic regions for Canada, Japan, and the USA. We present their basic findings in Table 5. Three main conclusions can be drawn from the intranational data on the relationship between consumption and output. First, there is still strong evidence that the quantity anomaly remains. While theory predicts that the cross-correlation for consumption should vastly exceed that for output, only for Canadian provinces in the Crucini (1995) study is this the case. Evidence for states within the USA provided by Crucini (1995) and Hess and Shin (1997), as well as for prefectures within Japan provided by van Wincoop (1995), demonstrates that the consumption correlation lies either below or roughly equal to that for output.(15) Together, this suggests that risk is not being shared completely across geographic regions of Japan and the USA.(16)
In Table 6 we present the cross-correlations for non-durable consumption, output, and Solow residuals for the 19 states of the USA where data is available -- see Hess and Shin (1997). For all states except Louisiana and Texas , the consumption correlation is smaller than that for output or Solow residuals. Interestingly, these two states are the oil-producing states in our sample. Not surprisingly, while the other states have highly correlated outputs, the outputs for Louisiana and Texas are negatively correlated with those of other states.
The second main conclusion that can be drawn from intranational evidence is that, as compared to the results in Tables 1 and 2 for international data, the consumption and output correlations in Tables 5 and 6 are both higher within countries than across them. This provides strong evidence for the conclusion that regions are more integrated within a country than countries are integrated with one another. Finally, Table 4 also presents evidence on the average volatilities of consumption and output for Japan and the USA. Consistent with intertemporal risk sharing/consumption smoothing, the volatility of consumption lies below that of output. This is similar to the evidence found for international data: there is evidence of risk sharing across time (i.e. the permanent-income hypothesis or intertemporal risk sharing), but not intratemporal risk sharing across regions or countries.
[TABULAR DATA NOT REPRODUCIBLE IN ASCII]
These findings are consistent with research which attempts to establish whether regional risk is adequately diversified, and has quantified the channels of this risk sharing. In an interesting new path of research, Asdrubali et al. (1996) build from earlier workby Sala-i-Martin and Sachs (1992) and Atkeson and Bayoumi (1993) in distinguishing the roles that non-labour earnings (capital markets), taxes and transfers (the federal government), and credit markets play in making consumption more correlated across US states and less volatile across time -- in their terms, the smoothness of consumption. They find that 62 per cent is smoothed by markets transactions, 13 per cent by the federal government, and 25 per cent is left unsmoothed. These findings are consistent with those in the literature which find an imperfect role of risk sharing, both across states and across time in intranational data.
(iii) Intranational Evidence on the Price Anomaly
The price anomaly is the finding that the volatility of the terms of trade between countries far exceeds what theory predicts. As a bench-mark, BKK compare the terms-of-trade volatility to that for output, and report that while theory predicts that the latter should exceed the former, the data reveal that the opposite is true. While RBC theorists who neglect the role of money in their models (let alone multiple currencies!) find this result to be anomalous, households and firms know the dramatic impact that exchange-rate movements have on their travel, asset, and investment decisions. In fact, the existence of non-fundamental volatility in exchange rates is one of the primary economic factors in the discussion of a common currency for Europe. One wonders, therefore, how measures of the terms of trade across states within a country perform. If the terms of trade are still too volatile within a country (with a common currency such as the USA) and the price anomaly remains, then the benefits to a common currency would certainly be reduced.
Unfortunately, owing to data limitations, we cannot directly measure the terms of trade between states in the USA. However, we can obtain a measure of the production-based real exchange rate between states as the ratio of the state's gross,state product deflator (a GDP-type measure) to the national gross domestic product deflator. This measures the relative price of the state's representative basket of output as compared to the representative basket of other states. For intranational data, therefore, we use the production-based real exchange rate as a proxy for the terms of trade.(17) Recall that the ratio of these volatilities was approximately 2 in international data (see Tables 1 and 2), yet theory predicted a ratio of one-third. For intranational data, Hess and Shin (1997) report that the average volatility of the production-based real exchange rate for 19 states is 1.25 per cent per year as compared to 3.35 per cent for GDP. We can see that the ratio is more in line with the theory, and hence the price anomaly is not an issue for intranational data. Table 6 reports the volatilities for the terms of output, the Solow residual, and the production-based real exchange rate separately for the 19 states considered in Hess and Shin (1997). From the table, it is clear that the terms-of-trade variability between states is quite small, with the exception of Texas and Louisiana, which are the two oil-producing states in the sample.
[TABULAR DATA NOT REPRODUCIBLE IN ASCII]
IV. CONCLUSIONS AND POLICY PERSPECTIVE
While juxtaposing international business cycles with intranational business cycles has aided us in understanding the benefits and limitations of RBC models, it can potentially do more. From a policy perspective, intranational business cycles can also help to teach us how the world may look as countries join together to form common markets and perhaps issue a common currency. The three areas we have examined in this paper are business cycle fluctuations, risk sharing, and terms-of-trade volatility, and we discuss each in turn.
First, the results for international productivity fluctuation indicated that geography mattered: industries within individual countries comoved more than individual industries did across countries. In direct contrast, geography mattered much less within a country. As this suggests that business cycles are currently asymmetric across countries (see de Grauwe and Vanhaverbeke (1993) for evidence on European regions), the intranational findings tend to indicate that cycles may become more symmetric as countries reduce trade frictions for goods and capital by reducing tariffs and adopting a common currency.
Second, a comparison of the international and intranational evidence on risk sharing indicates that the integration of capital and output markets is unlikely immediately to enhance welfare from a portfolio diversification standpoint. This conclusion is based on findings that there is too little risk sharing between countries, as well as between regions within the same country. The general problem of too little risk sharing is, more likely, that too many households are currently excluded from participating in any financial markets, let alone international ones. Therefore, governments seeking to improve welfare along this dimension should aid in creating financial institutions that are more inclusive.
Finally, international evidence supports the view that the volatility, of the terms of trade between countries is far in excess of what theory predicts. For intranational data, however, the volatility of a terms-of-trade-type measure is much lower. Based on these two findings, the likely culprit driving the excessive volatility of the international terms of trade is the multiple currencies involved. It is likely, therefore, that the adoption of a common currency will lead to a reduction in the non-fundamental relative price volatility between countries. As relative price changes (whether between goods or countries) should reflect changes in fundamental supply and demand conditions, the adoption of a common currency would be likely to lead to higher welfare if it reduces the non-fundamental volatility of the terms of trade. Of course, other economic and political factors are relevant to the adoption of a common currency.
Altonji, J. G., and Ham, J. C. (1990), `Variations in Employment Growth in Canada: The Role of External, National, Regional and Industrial Factors', Journal of Labor Economics, 8(1), 198-236.
Altug, S., and Miller, R. A. (1990), `Household Choices in Equilibrium', Econometrica, 58(3), 543-70.
Asdrubali, P., Sorensen, B. E., and Yosha, O. (1996), `Channels of Interstate Risksharing: US 1963-1990', Quarterly Journal of Economics, 111(4), 955-1006.
Atkeson, A., and Bayoumi, T. (1993), `Do Private Capital Markets Insure Regional Risk? Evidence from the United States and Europe', Open Economies Review, 4(3), 303-24.
Backus, D. K., Kehoe, P. J., and Kydland, F. E. (1992), `International Real Business Cycles', Journal of Political Economy, 101), 745-75.
_____ _____ _____ (1993), `International Business Cycles: Theory vs. Evidence', Federal Reserve Bank of Minneapolis Quarterly Review, Fall, 17, 14-29.
_____ _____ _____ (1994), `Dynamics of the Trade Balance and the Terms of Trade: the J-curve?', American economic Review, 84(1), 84-103.
______ _____ _____ (1995), `International Real Business Cycles: Theory and Evidence', in T.F. Cooley (ed.), Frontiers of Business Cycle Research, Princeton, NJ, Princeton University Press.
Baxter, M., and Crucini, M. J. (1993), `Explaining Saving/Investment Correlations', American Economic Review, 83(3), 416-36.
_____ Jermann, U.J. (1997), `The International Diversification Puzzle is Worse than you Think', American Economic Review, 87(1), 170-80.
Burns, A. F., and Mitchell, W. C. (1946), `Measuring Business Cycles', NBER Studies in Business Cycles No. 2, New York.
Cardia, E. (1991), `The Dynamics of a Small Open Economy in Response to Monetary, Fiscal, and Productivity Shocks', Journal of Monetary Economics, 28(3), 411-34.
Chadha, B., and Prasad, E. (1994), `Are Prices Countercyclical? Evidence from the G-7', Journal of Monetary Economics, 34(2), 239-57.
Clark, T. (1997), `Employment Fluctuations in US Regions and Industries: The Roles of National, Region-Specific, and Industry Specific Shocks', Journal of Labor Economics, forthcoming.
Cochrane, J. (1991), A Simple Test of Full Consumption Insurance', Journal of Political Economy, 99(5), 957-76.
Conze, A., Lasry, J. M., and Scheinkman, J. (1993), `Borrowing Constrains and International Comovements', in R. Becker et al. (eds), General Equilibrium, Growth and Trade, Volume 2, The Legacy of Lionel McKenzie, Academic Press, 460-89.
Cooley, T. F., and Ohanian, L. E. (1991), `The Cyclical Behavior of Prices', Journal of Monetary Economics, 28(1), 25-60.
Costello, D. (1993), `A Cross-Country, Cross-Industry Comparison of Productivity Growth', Journal of Political Economy, 101(2), 207-22.
Crucini, M. J. (1995), `On International and National Dimensions of Risk Sharing', mimeo, Ohio State University. de Grauwe, P., and Vanhaverbeke, W. (1993), `Is Europe an Optimal Currency Area: Evidence from Regional Data', in P. Masson and M. Taylor (eds), Policy Issues in the Operation of Currency Unions, Cambridge, Cambridge University Press, 111-29.
Finn, M. G. (1990), `On Savings and Investment Dynamics in a Small Open Economy', Journal of International Economics, 29(1-2), 1-21.
Ghosh, A., and Wolf, H. C. (1996), `Texas or Textiles? Geographic Versus Sectoral Shocks in the US Business Cycle', mimeo, New York University Business School.
Gregory, A. W., Head, A. C., and Raynauld, J. (1995), `Measuring World Business Cycles', mimeo, Queen's University. Hamilton, J. D. (1989), `A New Approach to the Economic Analysis of Non-Stationary Time Series and the Business Cycle', Econometrica, 57(2), 357-84.
Hess, G. D., and Iwata, S. (1997), `Measuring and Comparing Business Cycle Features', Journal of Business and Economic Statistics, forthcoming.
_____ Shin, K. (1995), `Intranational Productivity Growth in the US', mimeo, University of Kansas.
_____ _____ (1997), `Intranational Business Cycles in the United States', Journal of International Economics, forthcoming.
Hodrick, R.J., and Prescott, E.C. (1997), `Postwar US Business Cycles: An Empirical Investigation', Journal of Money, Credit and Banking, 29(1), 1-16.
Hulten, C.R., and Schwab, R.M (1984), `Regional Productivity Growth in US Manufacturing: 1951-78', American Economic Review, 74(1), 152-62.
Kollmann, R. (1995), `The Correlation of Productivity Growth across Regions and Industries in the United States', Economics Letters, 47(3-4), 437-43.
_____ (1996), `Incomplete Asset Markets and the Cross-Country Consumption Correlation Puzzle', Journal of Economic Dynamics and Control, 20(5), 945-61.
Kydland, F., and Prescott, E. C. (1982), `Time to Build and Aggregate Fluctuations', Econometrica, 50(6), 1345-70.
Mendoza, E. G. (1991), `Real Business Cycles in a Small Open Economy', American Economic Review, 81(4), 797-818.
_____ (1995), `The Terms of Trade, the Real Exchange Rate, and Economic Fluctuations', International Economic Review, 36(1), 101-37.
Norrbin, S. C., and Schlagenhauf D.E. (1988), `An Inquiry into the Sources of Macroeconomic Fluctuations', Journal of Monetary Economics, 22(1), 43-70.
Obstfeld, M. (1994), `Are Industrial-Country Consumption Risks Globally Diversified?', in L. Leiderman and A. Razin (eds), Capital Mobility: the Impact on Consumption, Investment and Growth, Cambridge, Cambridge University Press.
Sala-i-Martin, X., and Sachs, J. (1992), `Fiscal Federalism and Optimum Currency Areas: Evidence for Europe from the United States', in M. Canzoneri, P. Masson, and V. Grilli (eds), Establishing a Central Bank: Issues in Europe and Lessons from the US, London, Cambridge University Press.
Stockman, A. C. (1988), `Sectoral and National Aggregate Disturbances to Industrial Output in Seven European Countries', Journal of Monetary Economics, 21(2-3), 387-409.
_____ Tesar, L. L. (1995), `Tastes and Technology in a Two-Country Model of the Business Cycle: Explaining International Comovements', American Economic Review, 85(1), 168-85.
van Wincoop, E. (1995), `Regional Risksharing', European Economic Review,
(1) This paper has benefited from helpful comments by Stephen Nickell. Part of this paper was written while Gregory D. Hess was a Visiting Scholar at the Federal Reserve Bank of Kansas City. The opinions expressed are those of the authors and do not necessarily reflect the views of the Federal Reserve Bank of Kansas City or the Federal Reserve System.
(2) A problem with using simple Solow residuals to measure total factor productivity is that they are sensitive to the omission in the production function of cyclical utilization factors for both capital and labour -- the latter perhaps due to labour hoarding. This shortcoming applies to the literature cited throughout our paper.
(3) The countries included are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, The Netherlands, New Zealand, Norway, Sweden, the UK, and the USA. All data are obtained from World Data CD-ROM 1995, provided by the World Bank, except employment data for Germany, Italy, and the UK, which are not available from this data set. To add the employment data for these three countries, we supplement our data set with the OECD Main Economic Indicators. We exclude Spain, Iceland, and Switzerland, because employment data are not available in both data sets. Luxembourg is also excluded because it does not have terms-of-trade data. To sum, four OECD countries are excluded from the sample.
(4) While investment expenditures certainly contribute to the shape of the business cycle, since the stock of capital is large relative to the flow of investment, we can make this simplification. This has also been done in work by BKK. Costello (1993) uses a proxy for capital, energy consumed, in her Solow residual measure.
(5) For example BKK (1995) point out that, despite some heterogeneity in international business cycle experience across the major industrialized countries over the last 20 years, most regularities emphasized in Kydland and Prescott's (1982) closed-economy study stand up.
(6) Since their sample period analysed is shorter (1970Q1-1990Q2) and they use quarterly data, their findings are not directly comparable to Table 2.
(7) For other extensions, to name a few, see Finn (1990), Cardia (1991), Baxter and Crucini (1993), Mendoza (1991), and BKK (1995).
(8) Altug and Miller (1990) are quite clear on the relationship between risk sharing and consumption smoothing. The former implies the latter, but the latter does not imply the former.
(9) There are nine geographic regions determined by the US Census Bureau.
(10) As a caveat, however, since growth may be in part due to the `catch-up effect' -- poorer countries grow faster than richer ones as predicted by the Solow model -- a country that is `catching up' is likely to do so in all industries, so that the correlation of activity by industries within a country may be biased upwards. This is unlikely to influence the within-country results as the initial levels of output across regions in a state are more similar than those across countries.
(11) More specifically, let [y.sub.ijt] be real output of i industry in state j at time [Alpha.sub.i] is the time-series average labour income share of total output in industry i, assumed constant across all regions. Then [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII], and [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where In is the natural logarithm, [Delta] is the time-difference operator, and [k.sub.it] is real capital of i industry in aggregate at time t.
(12) More formally, [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] is the weight of each shock. [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] where [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].
(13) See Altonji and Ham (1990) for an analysis of Canadian employment data.
(14) For a recent treatment of the home-bias problem, see Baxter and Jermann (1997).
(15) The differences in data used by Crucini (1995) and Hess and Shin (1997) are as follows. The data used by Hess and Shin are gross state product (GSP) data which are available in constant dollars only since 1977. Also, official retail sales data from the Bureau of Economic Analysis are only available for 19 states in the USA from 1978. On the other hand, Crucini uses retail sales data for a longer time period and for all 50 states plus the District of Columbia, although from an unofficial source (published in The Survey of buying Power Data Service). The data also contain expenditures on durable goods, which are not compatible with theory. Additionally, nominal personal income for each state is converted to real data by a common aggregate deflator.
(16) Also, since households hold the majority of their wealth in human capital, which is `quasi-fixed', primarily in local economic activity, due to
(17) In Hess and Shin (1997) we demonstrate that a price anomaly exists between theory and international data for the production-based real exchange rate that is almost identical to that for the terms of trade. In international data we must convert relative prices into a real exchange rate
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||business cycles affecting regions within a country and affecting countries|
|Author:||Hess, Gregory D.; Shin, Kwanho|
|Publication:||Oxford Review of Economic Policy|
|Date:||Sep 22, 1997|
|Previous Article:||The labour market over the business cycle: can theory fit the facts?|
|Next Article:||Asset market hangovers and economic growth: the OECD during 1984-1993.|