New research suggests the influence of industry factors is set to grow, says Joanna Munro at ABN AMRO Asset Management
The industry/country debate has been a hot topic for the past decade, with highly respected academics arriving at quite different conclusions. Beckers et al1 wrote that “global and national influences are of roughly equal importance”, while Heston and Rouwenhorst2 found that “industries explain less than 1% of country index volatility”. With the recent Financial Analysts Journal publication of Rouwenhorst’s latest research3, concluding that “in_Europe country effects continue to dominate industry effects” it is clear that, post-Emu, the industry/country debate is still very much alive.
At ABN AMRO Asset Management we believe that companies increasingly operate in a global marketplace, that a global industries approach is the right way to add value to our equity portfolios and that, in Europe in particular, industry effects are becoming as important as country influences. This belief is supported by the on-the-ground observations of industrial structure from our portfolio managers and industry analysts. They see, for example, that GM and Chrysler face similar challenges to Toyota in the global marketplace for automobiles, and that the profitability of the UK company Glaxo/Wellcome is related to the success of the Canadian company, Biochem Pharma, in one of its strongest franchises, the HIV market. Despite this anecdotal support, the research evidence on the subject has been unclear and contradictory.
What the academics say
One of the simplest ways to see the changing relationship between country and industry factors is to analyse the correlations of country and industry indices over time: however, correlations don’t provide us with hard evidence about the relative importance of industry and country factors. Why is that?
In practice, industries differ in their geographical distribution and countries differ in their industrial make-up. This means that when we observe a change in correlation between two indices we don’t know if it’s the result of the industry element of the index return or the country element. As one example, the Swiss market index has more than 40% exposure to the health industry, which is not represented in the Italian index. Is the relatively low correlation of the Swiss market with Italy (0.28 since 1990) due to its Swiss-ness or its health exposure?
Given the above, the majority of industry/country research in the past decade has focused on isolating ‘pure’ country and industry impacts. In the early 1990s, the Barra group of Beckers, Grinold, Rudd and Stefek published a paper which looked at Europe only and carried out analysis at the stock level4. They found that industry factors alone explained 21% of volatility. However, in making that statement it seems that the group was capturing at least part of a ‘general market’ influence in its industry factors and that the additional explanation provided by industry factors was only around 5%.
A few years later Heston and Rouwenhorst ran a stock level analysis within Europe using monthly data and a broad seven sector industrial classification. They used an approach which has been generally adopted since then, estimating industry and country factors return for each period by giving each stock a 1/0 exposure to the various industry and country factors and estimating ‘best fit’ factor returns. They explicitly included a general market factor in their estimation, so that the returns to industries and country factors can be thought of in excess return terms. In their conclusions they focused on their equal-weighted result that industry factors explained less than 1% of the variation in index returns. However, the value-weighted results found a somewhat higher level of explanation, just over 7%.
In 1996 Beckers, with Connor and Curds, published a further piece that is particularly interesting as it provides a comparison between global and European results and the use of broad sector and finer industry categorisations. They adopted the Heston and Rouwenhorst estimation methodology, using an equally-weighted approach. They found that sectors provided less explanation than the finer industry divisions, that industry effects were relatively more important within Europe than globally and that there was strong evidence of a trend towards integration in Europe. Although they stated in their conclusion that ‘global and national influences are of roughly equal importance’ they found relatively limited industry effects at a global level (an additional 5% explanation from the finer industry classification).
However, their results within Europe do suggest a different pattern from the Heston and Rouwenhorst work, with country effects having only 1.4 times the impact of industry effects, versus Heston and Rouwenhorst’s equally weighted result that country effects had around 2.1 times the impact of industry effects. The differences are likely to be mainly due to the different data period (Beckers included post-Maastricht data in his analysis) and the different industry classification (for this result, Beckers was using 36 industries rather than Heston and Rouwenhorst’s seven sectors). Currency may also be a factor as Beckers has used local currency excess returns while Heston and Rouwenhorst made their estimation in Deutschmark terms.
With the arrival of the single European currency, Rouwenhorst directly addresses the question of increasing European integration post-Maastricht in his recent paper in the Financial Analysts Journal. Using the same approach as Heston and Rouwenhorst (1994) and again using the seven sector classification, he looked at data from 1978 to August 1998, both in total and by five-year period. In his summary he concluded that country effects continue to dominate industry effects in Europe and that, quite surprisingly, there is no evidence of an increase in integration post-Maastricht.
In April 1999, Frank Russell published a research commentary: ‘Risk Contributors: Country versus Sector’5, which concluded that “country effect was more important ... than the sector effect”. They looked at 22 global developed countries from May 1993 to January 1998 using both 12 economic sectors and the 38 MSCI industries. In the same way as Griffin and Karolyi in their paper of 19986, rather than working with stock level information, they used returns to the MSCI country/industry sub-indices to estimate ‘pure’ factor country and sector returns. On a global basis, using the broad sector factors, they found that country factors had 2.3 times the impact of industry factors. However, using the finer industry classing, the ratio fell to 1.6 times.
Clearly the various researchers have produced some conflicting results and the differences generally arise from three main factors:
q geographical scope of study (Europe vs global);
q industry classification (broad sectors vs finer industries), and
q data period (within Europe, pre- and post- Maastricht).
Logically, we would expect a broad sector approach to have difficulty identifying the true ‘industry’ effects as, for example, publishing stocks will be grouped in the same class as shipping companies.
For our purposes, to allow proper comparison across papers and methodologies, we chose to calculate a relatively crude ‘common’ statistic from three key papers. The statistic we calculated was the ratio of the standard deviation of country effects to industry effects, where the standard deviation has been simply averaged and the data frequency is unimportant, given the ratio.
Why should a high standard deviation indicate greater importance for one class of factors? If a factor is going to explain variability in returns it has to have some variability itself. Other things being equal, the more variation a factor has, the more variation in returns it can explain. In this case, a ratio greater than one indicates that countries are more important than industries, a ratio of one would indicate equivalence and a ratio less than one would indicate a greater role for industry factors than country factors.
The obvious points to make are:
q improving the industry classification from ‘broad’ to ‘narrow’ appears to increase the relative importance of industries;
q industries appear to play a more significant role within Europe than they do worldwide;
q within Europe, the relative importance of industries appears to be increasing recently.
All these ratios are greater than one, suggesting that industry-specific factors are still less important than country specific factors. However, there is one recent result that is missing in the table: the most recent period for Europe using the ‘narrow’ industry classification. Given the increased importance of industries when analysed using that finer classification, we felt it likely that such an analysis would show that industries are now of equivalent importance within Europe. That is the subject of the second part of this article.
ABN AMRO Asset Management research
We used the same approach as Griffin and Karolyi in their 1998 paper and calculated ‘pure’ country and industry returns using the monthly returns to MSCI country/industry indices for 15 developed European countries and 307 MSCI industries from July 1993 to February 1999. Given the time series of ‘pure’ country and industry returns we were able to calculate standard deviations for each factor return over time.
Although we did carry out analysis using the ‘sector’ classification and found results similar to Rouwenhorst (1999), our real interest was in the results using the ‘industry’ classification. List 1 shows the results based on the fine industry (30) and country (15) classification.
The common ratio for these results (ratio of the standard deviation of country effects to the standard deviation of industry effects) was calculated as 1.0 times using the simple average, 1.0 times using the median and 0.9 times using the weighted average. Figure 1, ranking standard deviations, and showing country factors (red) and industry factors (blue) now shows a real interleaving of industry and country factors – with no clear pattern of dominance for either countries or industries. This indicates that, post-Maastricht, industry effects are now at least as important as country effects within Europe.
Summary and conclusion
The relative importance of industry and country influences on stock returns has long been a subject of debate, despite consistent anecdotal evidence in support of increasing global integration. Summarising the academic research we found that:
q industries were relatively more important in Europe than globally;
q using a finer industry classification shows more impact for industrial influences than the broad sector classes, and
q industry influences were increasing in Europe post-Maastricht.
In all the pure factor research we analysed, the results showed that country influences remained dominant over industry effects. However, we noticed an absence of ‘pure’ factor analysis for Europe post-Maastricht, using the fine industry rather than broad sector classification. This encouraged us to undertake our own analysis using the ‘pure’ factor approach for Europe since 1992. We used both sector and industry classifications and found that our sector results were consistent with the literature. Moving to the more natural industry classification (where publishing, shipping and tourism are no longer treated as subject to the same influences) we found that, as anecdotal evidence suggested, in post-Maastricht Europe, industrial effects are now as important as national influences. The overall research is summarised in Table 1.
What of the future? We believe that the relationship between industry and country factors will continue to evolve. In particular, as barriers to cross-border movements of people and goods are removed in Europe, we expect to see increasing geographical specialisation, in line with the US experience where technology is concentrated in California and finance on the east coast. We believe the implications for asset managers are clear: a pure country-oriented approach to stock investing will miss its mark and the benefits of a cross-country industry approach will only increase over time.
Joanna Munro is head of global consulting at ABN AMRO Asset Management in London
6 Griffin and Karolyi, “Another look at the role of the industrial structure of markets for international diversification strategies” Journal of Financial Economics Vol 50 (1998), pp 351–73.
7 The MSCI Europe universe consists of 37 industries. In order to achieve a robust overall result, we excluded from the analysis MSCI industries with very low weights (ie an industry is excluded if a market weight (Feb 99) is lower than 0.2%). Excluded industries were Misc. Materials (26), Data Processing (33), Energy Equipment (36), Appliances (41), Textiles (47), Transportation – Roads & Rail (57) and Trade (59).