France: Financial Sector Assessment Program—Technical Notes—Stress Testing Methodology and Results; Integration into Global Financial Markets; and Public Intervention in Financial Markets—Obstacles to Monetary Transmission
Author:
International Monetary Fund
Search for other papers by International Monetary Fund in
Current site
Google Scholar Close
Publication Date:
08 Jun 2005
eISBN:
9781451813630
Language:
English
Keywords:
ISCR; CR; rate; French; bank; yield; Crédit Agricole; bank activity; activity in France; fixed-income market; market integration; Securities markets; Stocks; Foreign banks; Loans; Global
Download PDF (949.5 KB)

These Technical Notes on France explain integration of global financial markets. The stress tests for the France Financial Sector Assessment Program (FSAP) were designed to yield as comprehensive and detailed a picture as possible within the constraints of the approach. Retail activity by foreign banks in France is small, but significant. The financial landscape in France remains characterized by a large number of idiosyncrasies that affect monetary transmission. Macroeconometric models point to a smaller reaction to monetary policy in France than in other large euro-area economies.

Abstract

These Technical Notes on France explain integration of global financial markets. The stress tests for the France Financial Sector Assessment Program (FSAP) were designed to yield as comprehensive and detailed a picture as possible within the constraints of the approach. Retail activity by foreign banks in France is small, but significant. The financial landscape in France remains characterized by a large number of idiosyncrasies that affect monetary transmission. Macroeconometric models point to a smaller reaction to monetary policy in France than in other large euro-area economies.

II. Integration of Global Financial Markets

This note reviews the international linkages of the French banking and insurance sectors, and with the integration of organized capital markets. The overall level of international integration is comparable to that of other major industrial countries. Banks undertake a large volume of business in the Europe-wide interbank market, and lend extensively, but mostly to industrialized countries. Retail banking activity abroad is undertaken mainly by a few major financial groups, and is typically more important for the recipient country than for the metropolitan parent. Retail activity by foreign banks in France is small but significant.

The insurance sector is, in important ways, more integrated internationally, with important cross-border ownership links in both directions. The market for reinsurance is very internationalized.

Equity and bond markets have been growing in importance as vehicles for international capital movements. The correlation of returns among equity markets has displayed a trend increase, as has the co-movement of volatility. The establishment of Euronext in itself does not seem to have had a major impact on correlations among returns. Correlations among bond returns have been more stable in recent years, but display occasional episodes of high correlation.

A. Introduction31

53. French financial institutions compete with other major financial institutions around the globe, and France represents a major international financial market.32 Hence, international factors can influence the stability of the French financial system, both as a source of risk and as a means of diversification. Its structural development will be affected by trade in financial services across borders, foreign entry into the French market and the scope for expansion by French institutions abroad, and the search by investors for the efficient allocation of capital worldwide. Therefore, an examination of international linkages is an important part of the financial sector assessment.

54. Strong interlinkages between French and global financial institutions and markets are inevitable given the linkages that exist in the real sector. The French economy as a whole is very open, and macroeconomic developments are strongly influenced by economic conditions abroad. As a member of the EMU, monetary conditions are determined on a Euro area-wide basis. Economic cycles are broadly synchronized with those in major trading partners (over the past decade, the correlation of quarterly changes in industrial production with those in Germany the U.K., and the U.S. has been around 0.5, and that with Italy has been over 0.65). Exports and imports are both about 30 percent of GDP. French corporations also have large foreign operations and direct investments; in some recent prominent cases, ambitious overseas expansion caused financial distress and over-borrowing by some large conglomerates. Hence, French financial institutions have a strong incentive to follow French firms abroad, and even much of their domestic business will have an external aspect. Likewise, financial market performance will be correlated with that of markets in other countries, even if there were no cross-country portfolio holdings.

55. But cross-country portfolio stocks and flows are very large. Table 9 shows that France’s gross foreign claims and obligations are large in absolute terms and relative to GDP, and they have been growing in importance; the decline in 2002 is due to the appreciation of the euro. Financial obligations are spread across a range of investment vehicles, with no one category dominant. Securities, which consist mainly of equity and government obligations, make up the largest share of France’s external obligations. Among France’s external assets, holdings of securities have been increasing and by 2002 matched the stock of direct investment. The gross positions of monetary financial institutions (largely banks) are also large, and sundry other items are sizeable.

Table 9.

France: Foreign Assets and Liabilities

(All at market values; end of period)

article image
Source: Banque de France and staff estimates.

56. Such connections through investment portfolios are most immediately relevant to the condition of the financial sector. In addition, the financial sector itself conducts business with the rest of the world by selling or buying financial services either directly or through subsidiaries.

57. This note concentrates on such linkages that arise within the financial sector, rather than on international linkages through the other sectors; only a part of the holdings of international obligations and claims, and the direct investment described above is intermediated through the financial sector. Attention focuses on the international operations of French financial institutions—especially banks and insurance companies—through direct exposure and through the activities of their subsidiaries; the activities of foreign institutions in providing financial services in France; and the connections between securities markets in France and elsewhere.

B. Banking Sector33

58. French banks play a major role in channeling cross-border financial flows. By way of illustration, their international lending is on scale comparable to that of other major industrialized countries (Figures 3 and 4). They are particularly active in lending to other industrialized countries, primarily via the EMU-wide interbank market (see below). Their lending to developing countries is less in absolute terms and also smaller relative to that of banks from some other major industrialized countries.34 Nonetheless, the amounts are large, especially relative to the size of the recipient economies.

Figure 3.
Figure 3.

Bank Lending to Developed Countries by Major Countries, end‐2003

(In billions of U.S. dollars)

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Source: BIS, and staff estimates.
Figure 4.
Figure 4.

Bank Lending to Developing Countries by Major Countries, end-2003

(In billions of U.S. dollars)

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Source: BIS, and staff estimates.

French banks’ foreign exposures

59. Statistics from the Bank for International Settlements (BIS) provide information on the magnitude of French banks’ consolidated claims on the rest of the world, and their geographical distribution (Table 10). The total, currently over €1,000 billion, is equivalent to about one quarter of the banking system’s total assets. However, claims on other industrialized countries in Europe constitute one half of the total, and claims on the U.S. amount to more than 20 percent. The next largest exposure is to offshore financial centers, part of which may represent claims on subsidiaries of financial institutions from other industrialized countries. Claims on other countries constitute less than a tenth of the total, a share that has been declining. The absolute amounts have been roughly stable over the past five years, except for a decline in claims on Latin America and the Caribbean from 2002, which is presumably attributable to the financial crises suffered by some countries it the region.

Table 10.

France: French Banks’ Consolidated Claims on the Rest of the World

(End of period)

article image
Source: Bank for International Settlements, and staff estimates.

60. More detail is provided by available information on (resident, nonconsolidated) French banks’ claims on other countries (Table 11 and Figures 5-7). Total claims are less than 20 percent of total assets, and are dominated by lending. This lending goes mostly to Europe (notably the U.K.), but the U.S. and East Asia (mainly Japan, Korea and China) are also important borrowers. Other relatively large exposures are to (near) investment-grade emerging markets such as the Czech Republic, Morocco, and Poland; exposures to noninvestment grade emerging markets are relatively small and widely diversified. Holdings of foreign bonds have been growing in importance. Again, European issues predominate; in contrast to the pattern for lending, issuers from countries such as Italy and the Netherlands are relatively important. Equity holdings by banks are more modest and have been roughly stable in absolute value since end-1999. The geographical distribution is wide, with all the main markets represented. However, some of these claims represent banks’ loans to or equity stakes in their own subsidiaries and joint ventures.

Table 11.

France: French Banks’ Claims on Nonresidents

(End of period)

article image
Source: French authorities, and staff estimates.
Figure 5.
Figure 5.

France: French Banks’ Loans Abroad

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 6.
Figure 6.

France: French Banks’ Holdings of Foreign Bonds

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 7.
Figure 7.

France: French Banks’ Holdings of Foreign Shares

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

61. A further breakdown of French banks’ claims on nonresidents shows that most are against other financial institutions (Table 12). Loans to nonresident financial institutions amount to three quarters of financial institutions’ total loans to nonresidents, and about 30 percent of total loans to all financial institutions. Claims on nonfinancial nonresidents are much smaller, and represent only a small fraction of total lending. These claims only slightly exceed deposit liabilities toward nonfinancial nonresidents. This table also show that 80 to 90 percent of foreign assets and liabilities are concentrated in the commercial banks; the mutuals and other credit-granting institutions have traditionally concentrated on the domestic retail market, and so far are rather less involved in the European money markets (Box 2).

Table 12.

France: Domestic Credit Institutions’ Claims on and Liabilities Toward Nonresidents 1/

article image
Source: French authorities, and staff estimates.

Includes foreign-owned banks in France.

Foreign Operations of French Banking Institutions 1

Major French banking institutions differ greatly in the extent and type of foreign operations that they undertake. All the large groups are engaged in the European-wide interbank market, and wholesale investment banking is an internationally integrated business. All the major banks also have strategies for developing their international business. However, only a few banks already have extensive involvement in overseas markets. In particular, the two private bank groups, and to some extent subsidiaries of the Credit Agricole, are the main internationally active French banks. Despite strong ties to some regions, notably francophone Africa, French banks are less internationally active than some of their competitors from other European countries, and were relatively late in entering growing markets in Eastern Europe and Asia.

BNP Paribas has built up a retail presence in several overseas markets: more than half its revenue and profits from retail operations are generated abroad, and only 40 percent of its lending goes to France, while a quarter goes to North America and a sixth to Western Europe. It has acquired two retail branch networks in the United States, and also has a presence in the WAMU region, the Maghreb and the Mashrak. It has a significant presence in the consumer lending markets in the U.K., Spain, Italy, and Hungary. Its operations in Asia are comparatively small and are undertaken mainly through joint ventures. About a third of its leasing activities are abroad, and it is involved in property management across Europe.

Société Générale has concentrated more on acquiring retail networks in transition countries (notably the Czech Republic and Slovenia) and some developing countries (such as Ghana and Tunisia). These have recently achieved a ROE in excess of 30 percent. It also engages in specialist financing and financial services (car and consumer loans, fleet management) in Europe, investment banking in such countries as Spain, Germany and Italy, and private banking worldwide. International operations in 2003 contributed 11 percent of group revenue and 13 percent of profits.

Crédit Agricole was traditionally a domestically-oriented bank, but acquired first Indosuez, a medium-sized, internationally active investment and private bank, and then Crédit Lyonnais, which brought with it more foreign operations. Crédit Lyonnais scaled back its European network during the 1990s, but still earns about a quarter of its revenues abroad (mainly in the Americas and Western Europe). Crédit Agricole itself has minority shareholdings in several banks abroad, most importantly in Greece, Italy, Poland, and Portugal. It wrote off its investments in Argentina and handed over its operations there to a local bank following the crisis in 2002.

1The information contained in this box comes mainly from material released in the Banque de France Bulletin, published reports by consultants, and banks’ annual reports. The consolidated accounts do not in all regards distinguish between domestic and foreign business, especially with regards to wholesale activities.

62. Much of French banks’ foreign operations are undertaken by subsidiaries and therefore are not booked to the metropolitan head office. The number of subsidiaries abroad has increased as they have developed networks in foreign markets (Table 13). Direct employment abroad has tended to decrease, but year to year fluctuations are affected by individual large deals. This measure also illustrates that foreign operations are mainly undertaken by the commercial banks, and in particular the largest ones. In certain developing countries (mainly francophone ones), subsidiaries of French banks play a dominant role, but the financial magnitudes involved are small relative to business in industrialized and emerging countries.

Table 13.

France: Credit Institutions’ Subsidiaries, Branches and Employment Abroad 1/

(End of period)

article image
Source: French authorities, and staff estimates.

Includes foreign-owned banks in France.

Foreign banks’ operations in France

63. French banks not only compete in markets in other countries, but also face competition from foreign banks in France. The operation of the latter also links conditions in France to the performance of those institutions in other countries. An overview of this exposure is provided by BIS data on non-French banks’ claims on France (Table 14). The total rose by a third in the early years of the EMU, but has since leveled off at around €600 billion. Interbank lending within the Eurozone makes up much of the total and accounts for most of the increase. Claims on the public sector (presumably mostly in the form of government debt obligations) has shown a trend increase and now constitutes about one eighth of the total. Claims on the nonbank private sector are about double that, but has varied little since 2000, when a major British bank acquired a mid-sized French bank.

Table 14.

France: Consolidated Borrowing from Foreign Banks

(In billion euros; end of period)

article image
Source: Bank of International Settlements, and staff estimates.

64. Foreign banks have about a ten percent market share; this level of foreign penetration is comparable to that seen in other major European countries (Table 15). Foreign banks have a higher share of lending than they do of deposit taking. The number of foreign banks active in France has been roughly constant, but the volume of their activity jumped in 2000 due to the aforementioned take-over. There has also been exit: attempts by foreign banks to enter the French market for housing loans and internet banking were not successful, and one foreign-based internet bank recently withdrew from the market.

Table 15.

France: Foreign Bank Activity in France

article image
Source: French authorities, and staff estimates.

Includes foreign-owned banks in France.

65. While the interbank and investment banking markets are well integrated across Europe, cross-border retail business remains limited due to differences in language, customs and legal systems. Furthermore, insofar as a bank’s close connections with its clients is an intrinsic aspect of its competitive advantages (and indeed a main reason why intermediaries exist), maintaining a branch network is a precondition for reaching a significant market share. Since building a new branch network involves large, up-front fixed costs, acquisition of an existing network is likely to remain the most attractive means of entry. Yet, most of the branch networks in France are under mutualist ownership, which precludes a take-over. Rather, French banks may continue to seek to acquire branch networks in other countries, or they may enter into cooperation agreements to develop joint products or exploit economies of scale in such areas as information processing.35 Improved profitability and capitalization may provide French banks with the means to take over banks elsewhere in Europe as momentum for integration builds.

C. Insurance36

66. The French insurance sector is among the six largest in the world. It has extensive linkages with the insurance sectors in other countries, and with global financial markets. These linkages take the form mostly of ownership connections, reinsurance, and financial investments, rather than the writing of cross-border policies.

67. Foreign insurance companies own subsidiaries in France with a combined market share of about 20 percent in nonlife business, and rather less than that in the life sector. These companies are for the most part subsidiaries of large insurance groups from Germany, Italy and the United Kingdom.

68. French insurers have important subsidiaries in other countries. In particular Axa, the largest, has extensive insurance operations in the U.S., Europe (mainly the U.K., Germany, and Belgium), Japan, Hong Kong, Singapore, and Australia, conducted through subsidiaries.37 It also has smaller operations elsewhere, for example in Morocco, and operates a medium-sized bank in Belgium, but in 2003 it pulled out of insurance business in Argentina and Brazil. Less than a fifth of Axa’s group revenue is generated in France; North America, or Germany, and the U.K. combined are equally important. In recent years, Axa has earned more net income on life business in the U.S. than in France. More generally, the diversification of its operations across countries and lines of business tends to stabilize group profits. Any one member company could presumably suffer large losses, but at least in principle Axa group’s liability is limited.

69. For domestically-incorporated companies, premium revenue from abroad is relatively small. Premium income from outside France amounts to less than two percent of total premiums, and that comes overwhelmingly from within the EU. Non-French business is almost all in the nonlife sector. These statistics imply that French companies have little direct exposure to insured risks outside France.38 Companies report that writing of policies in foreign jurisdictions is complicated by differences in legislation and legal systems—even within the EU—and therefore they prefer to work through locally-incorporated subsidiaries.

70. Data are not available on the allocation of assets by currency or country. However, prudential regulations and the companies’ incentives ensure that the vast bulk of assets take the form of claims on industrialized countries, and in particular euro-denominated government bonds from OECD countries. Hence, the vulnerability of the insurance sector to country or exchange rate risk is small.39

71. French insurers make extensive use of foreign reinsurers, while French commercial reinsurers are comparatively small and concentrate on business in France. Thus, in 2002, total reinsurance premiums amounted to €19.3 billion (just under 15 percent of premium income), but less than half that amount went to French reinsurers. Foreign earnings of reinsurers in France amounted in that year to just over €1 billion, two-thirds of which was earned by French companies. The implication is that France is transferring abroad a significant portion of insured risk, while not taking on the concentration of risk that is characteristic of reinsurance. Therefore, the French financial system as a whole has little exposure to vulnerabilities of the reinsurance sector. Progressively more severe difficulties at foreign reinsurers would, however, be reflected first in higher reinsurance costs, then in the nonavailability of some types of reinsurance, and, ultimately, in the nonreimbursement of reinsured claims.

D. Equity and Bond Market Interdependence 40

Introduction

72. The advent of new technology and the elimination of capital controls has helped France’s capital markets to become closely integrated into the global marketplace over the past several decades. One indicator of the degree of integration is the role of French government bonds as a benchmark for short-to medium-term Euro-denominated interest rates, which would imply that European wide news and events would have an influence on these rates.

73. Market commentators have noted a more recent trend increase in correlations across major equity markets, including the French market.41 They have attributed this to several factors, including the globalization of financial markets as well as the greater integration of growth cycles across major industrialized economies.42 Another factor, which has garnered rather less attention, is the link between increase correlation and the rise in capital market volatility since the late 1990s. In the case of the French equity markets, an additional factor is the establishment of Euronext and its introduction of a common trading platform for trading French, Dutch, and Belgian stocks may have also increased equity market linkages across these jurisdictions, as investors could view them as constituting one equity market. In what follows we document the degree to which French equity and bond markets are in fact correlated to other major equity and bond markets, and assess its dependence on the above noted factors.

Historical correlations: some stylized facts

Cross-country securities holdings and trading

74. The proportion of French securities held and traded by foreign investors is relatively high. In France, roughly 40 percent of the trading activity in government bonds is accounted for by foreign investors. This is a similar in scale to most large mature economies, as their bonds tend to be used as benchmarks and/or in core (risk-free) portfolio allocations for global fixed-income investors.

75. Equity holdings by non-French residents is roughly 35 percent. This seems high in comparison to the U.S., where 9 percent of domestic equity is held by foreigners. However for the more open, non-European G-7 economies, the comparison with France is less stark as 22 and 15 percent of domestic Japanese and Canadian (respectively) equity is held by foreigners. More importantly, for European economies, France’s 35 percent figure is not at all uncommon, with the range for most EU countries between 25 and 38 percent. This is largely due to the large cross-European securities holdings arising out of European integration.

76. These data provide a rough indication of the potential financial linkages that exist via the strength in cross-country portfolio flows. Although these flows tend to gage the linkages that might arise due to the integration of financial systems, prices across securities markets in different countries are nonetheless also driven by news and information on (real-economy) fundamentals, which are, as mentioned in the introduction of this note, strongly linked across mature economies.

Equity and bond price dynamics: Volatilities and correlations

77. The evolution of major equity indices are plotted in Figure 6.43 The figure highlights that, as in many industrialized counties, the recent movements in French share valuations has been associated with the rise and fall of the technology equity bubble, by the long global expansion that ended in 2000, and by a series of crises and events that periodically buffeted these markets (e.g., LTCM crisis, September 11, 2001).

78. Market returns are clearly correlated. As depicted in Figures 7 and 8, correlations between national equity markets have been rising over our sample. These figures present rolling window estimates of the bilateral correlation coefficients between French and other major equity markets, including the average of all bilateral correlations for our set of markets.44 Also, the estimated correlation coefficients between the U.S. and a sample of these equity markets are presented in Figure 9. With the correlation coefficient estimates rarely declining below the 0.50 mark, French stocks appear highly integrated with other major equity market centers, essentially those in Europe. In particular, since the early 2000s, their price movements have become even more correlated with other major equity markets. A rise in the average equity market correlation is noticeable starting roughly in the early 1990s. The increase in correlations is sharper from late 1999, and reaches new highs in 2002, plateauing thereafter at around 0.70.

Figure 8.
Figure 8.

Equity Market Indices

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 9.
Figure 9.

French Equity Correlations

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

79. The correlation between the French equity market and the other Euronext markets in the Netherlands and Belgium increased markedly in late 2001 (Figure 8). This rise in equity price co-movements within this set of three markets roughly coincides with the implementation of the common trading platform in October of 2001. The correlation coefficient estimates before and after the move to the common trading platform in October 2001 presented in Table 16 provide supporting evidence of the upward trend.

Table 16.

France: Equity Market Correlations Pre and Post Move to Common Euronext Trading Platform

article image
Bottom part of matrix reports the estimated correlation coefficients after the move to the common trading platform, while the top reports these estimates for the period that preceded the move.

80. The French bond market also displays a high degree of integration with other industrialized bond markets. Bond index movements are presented in Figure 10, with French and U.S. bond market correlations presented in Figures 11 and 12,.45 Correlations across bond markets are generally positive, ranging on average between 0.2 and 0.8. These estimates are lower on average than those observed in equity markets. Cross-country bond return correlations between France, and Germany, the U.K. and, to a lesser extent, the U.S., have increased moderately over time, likely reflecting in part increasingly integrated fixed-income markets. However, overall, bond market correlations seem more stable than those observed across equity markets.

Figure 10.
Figure 10.

French Equity Correlations

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 11.
Figure 11.

US Equity Correlations

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 12.
Figure 12.

Bond Market Indices

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 13.
Figure 13.

French Bond Correlations

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 14.
Figure 14.

US Bond Correlations

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

81. Although economic globalization and greater integration of capital markets have been put forward as driving force behind the rise in asset return correlations, the coincident rise in volatility leads to the possibility that traditional correlation measures provide a misleading indication of this increase in global market integration. Over the last few years, academic work has shown that traditional (unconditional) correlation measures may be biased and dependent on the level of volatility observed.46

82. Volatility has in fact risen across most asset markets over the second half of our sample period. Moreover, French equity and bond returns have exhibited higher volatility than those in the U.S. (Figures 15 and 16, but they are broadly in line with that observed elsewhere in Europe; the difference may largely be a reflection of the relatively heavy weight in the CAC40 of financial and technology firms. When comparing the changes in correlation, particularly in equity markets, with the periods of heightened volatility displayed in Figure 15, one can observe that correlations tended to be higher during periods of heighten volatility. This may help explain the results presented in Table 16, as the move to the common trading platform and other measures to integrate markets also coincides with an extended period of frequent and sharp volatility increases. After conditioning for changes in volatility, it is possible that the majority of the increase in correlation observed in all markets could be attributed to a rise in volatility, rather than other macroeconomic or institutional factors.

Figure 15.
Figure 15.

Equity Market Volatility

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 16.
Figure 16.

Bond Market Volatility

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Market volatility and equity and bond return correlations

83. In order to get a more precise picture of the changes that have occurred in the level of interdependence between France’s capital markets and markets in other major industrialized economies, we use a procedure that estimates return correlations, conditional on the level of market volatility (i.e., correcting the correlation measures for volatility changes). Specifically, we estimate a model in which capital market prices are driven by both a common factor and idiosyncratic shocks, where the variance of disturbances can vary over time. This framework also allows one to test whether or not there has been any change the degree of “structural” interlinkages between these markets. The same approach is also used to test whether there was a regime change in the correlation between the French, Dutch, and Belgian equity markets following the move to a common Euronext trading platform on October 29, 2001. Details of the model, procedures and results are contained in Appendix I.

84. Our findings indicate that indeed these markets have become more integrated (while controlling for changes in volatility), thus supporting the idea that globalization, broadly defined, has influenced correlations across these markets.

85. We also examine if the Euronext driven integration of the French, Dutch and Belgian equity markets has had any direct impact on the degree of equity price co-movement across these markets. We find that for the French and the Dutch markets, there is little empirical support for the idea that a move to a unified trading platform resulted in greater cross-market correlation in equity returns. However, the Belgium stock market has become somewhat less correlated with the French and Dutch markets as a result since Euronext started, but not to an economically significant degree. This last finding may be attributable to differences in the composition of those markets; the Belgian exchange contains fewer multinational companies, and a number of major Belgian companies have been acquired by multinationals from other countries

APPENDIX I: A Multivariate Regime-switching Model for Equity and Bond Returns

Data

86. In order to document the extent of the linkages across French equity and bond markets with other major markets, we use weekly data for the period from August 5, 1987 to August 4, 2004.47 For equities, our analysis is based on the CAC40, DAX, S&P 500, and FTSE indices. We also include in part of our analysis the Dutch AEX and Belgian BEL20 indices. (Note, the BEL20 data spans January 1, 1991 till August 4, 2004.). For bonds, the data chosen are 7- to 10-year indices for France, Germany, the U.S. and U.K. These bond indices allow our analysis to be less affected by the durational differences that arise across sovereign benchmark bonds. Durational differences typically need to be taken into account when examining changes in the level of interests rates combined with changes in their volatilities.

Model specification

87. We postulate a multivariate model for returns that are driven by idiosyncratic and common shocks. These returns are correlated due to a common unobserved factor (i.e., the common shocks).48 However, the common shock is drawn from two separate distributions, with different variances, which represent two different regimes or states of the world. The main advantages of implementing this multivariate regime-switching model is that it captures changes in co-movements across asset returns, and avoids the many exogenous ad hoc assumptions about the timing of these changes typically required in standard linear econometric models.

88. Let the variables r1t, r2t,… and rnt denote the returns for n assets.49 With the assumption that each variable follows a distributed lag process, the model can be formalized as:

r it = α i + Φ i ( l ) r it + u it i = 1 , 2 , 3… n ( 1 )

where αi is the intercept term, Φi(l) is the polynomial specific to variable i, and uit is the disturbance term. The latter is assumed to be correlated: E(uitujt) ≠ 0 for i ≠ j. Moreover, as we elaborate below, the variance of the disturbance term is allowed to vary over time, taking on a high or low value in different periods. In order to ease exposition, we restrict our discussion of the model to the case of four asset returns.50

89. The assumption that these variables are correlated is suggested by the analysis done in the previous section, which shows that on average returns are positively correlated for both bonds and equities. This positive correlation further suggests the existence of a common, unobserved factor for the shocks uit, so that we can decompose the disturbance terms into common and idiosyncratic structural shocks:

u 1 t = g c + δ 1 z ct + σ 1 z 1 t u 2 t = g c + δ 2 z ct + σ 2 z 2 t ( 2 ) u 3 t = g c + δ 3 z ct + σ 3 z 3 t u 4 t = g c + δ 4 z ct + σ 4 z 4 t

where zct is the common shock, gc the mean of the common shock, and zit are the idiosyncratic shocks for variables i = 1,2,..,4. The idiosyncratic shocks are assumed to be uncorrelated with each other and with the common shock, and to have zero mean. In addition, their variances are normalized to one, giving their loading coefficients (δi’s for the common shock and σi’s for the idiosyncratic shocks) the interpretation of standard deviations.

90. We first discuss how we can examine the relationship between correlation and increases in volatility. Under the assumption that the size of equity (bonds) returns are smaller (larger) during periods of turbulent markets—corresponding to relatively low investor risk appetite—one would assume that the mean of the common shock would be smaller (larger) than the mean of the shock during more tranquil times, or indeed negative (positive).51 Moreover, the level of correlation between the returns may also be quite different during turbulent periods, when investors show relatively little risk appetite. In order to account for this, we allow the common shocks to be drawn from two distinct regimes. Let us denote these as regime 0 and regime 1, where regime one is assumed to have a smaller mean for the shocks and, in principle, a larger variance (and higher covariance) in the case of equity returns.

91. Let Sct be a state variable subject to two regimes and associated with the common mean, gc and shock zct. = {0, 1}. This allows us to rewrite the model in state dependent form:

u 1 t = g c ( S ct ) + δ 1 ( S ct ) z ct + σ 1 z 1 t u 2 t = g c ( S ct ) + δ 2 ( S ct ) z ct + σ 2 z 2 t ( 3 ) u 3 t = g c ( S ct ) + δ 3 ( S ct ) z ct + σ 3 z 3 t u 4 t = g c ( S ct ) + δ 4 ( S ct ) z ct + σ 4 z 4 t

92. In this notation, the different δs, as well as the common mean (gc), are a function of the state variable, Sct. The variance-covariance matrix for this system of residuals is thus given by:

Σ 0 = [ σ 1 2 + δ 1 , 0 2 δ 1 , 0 δ 2 , 0 δ 1 , 0 δ 3 , 0 δ 1 , 0 δ 4 , 0 δ 2 , 0 δ 1 , 0 σ 2 2 + δ 2.0 2 δ 2 , 0 δ 3 , 0 δ 2 , 0 δ 4 , 0 δ 3 , 0 δ 1 , 0 δ 3 , 0 δ 2 , 0 σ 3 2 + δ 3 , 0 2 δ 3 , 0 δ 4 , 0 δ 4 , 0 δ 1 , 0 δ 4 , 0 δ 2 , 0 δ 4 , 0 δ 3 , 0 σ 4 2 + δ 4 , 0 2 ] ( 4 )

where δi,0 = δi(Sct = 0) for i=1,2,3 and 4 Σ0 = Σ(Sct = 0). This implies that the variance of each variable, uit, will be the summation of the idiosyncratic variance σi2 and the square of the common shock loading factor δi,0. The covariance (and correlation) of the residuals are defined by the cross-products of the loading factors δi,0. For example the variance of u1, is σ12+δ1,02 and covariance of u1 with u2 is δ1,0 δ2,0 in regime 0. A symmetric version of the variance-covariance matrix is given for Sct = 1, where δi,1 replaces δi,0 for all i. Note that not only will the covariance of each series vary with the regime, but so will the variance itself since the common shock nonetheless has a direct additive impact of the total shock uit in each regime. For example, given that the mean of the common shock for regime one is assumed to be smaller than that of regime zero: State 0 is the high mean regime, while state 1 is identified as the low mean. However, it should be made clear that we allow the common shock loadings for each individual series, i, to differ across regimes. This implies that we allow for the possibility that variance of the returns may in fact be greater during either state zero or one in any jurisdiction (i). For example, the correlation between the CAC 40 and S&P returns in the high volatility regime could in principle be positive while the correlation between the CAC40 and DAX in this same regime is negative.

93. In analyzing conditional correlations, periods of high volatility are typically designated via some ex post rule (see Favero and Giavazzi (2000)). However, in our model changes in volatility are endogenously designated within the estimation process. We need only specify how (not when) the (state) variables evolve. As such we implement a two-state Markov-switching regime, where we assume that the two state variables change according to the transition probabilities given by:

Pr [ S t = 0 | S t − 1 = 0 ] = q and Pr [ S t = 1 | S t − 1 = 1 ] = p ( 5 )

The Markov switching probabilities are conditional on the previous state, and are able to capture the idea that high-variance states of the world are relatively persistent (at the weekly frequency).52

94. In order to test if the correlations among Euronext exchanges changed after the move to a common trading platform, we adjust the δs parameters, in a three equation framework, to incorporate a dummy variable that takes on the value of 0, before Euronext (October 29, 2001) and 1 thereafter. The δs take on the following configuration:

δ * it = δ i ( S t ) + β i D t ( 6 )

where we assume that the impact from the dummy variable is constant across states of the world. The variance-covariance matrix that is assumed under this condition is:

Σ 0 = [ σ 1 2 + δ 1 , 0 2 δ 1 , 0 * δ 2 , 0 * δ 1 , 0 * δ 3 , 0 * δ 2 , 0 * δ 1 , 0 * σ 2 2 + δ 2 , 0 2 δ 2 , 0 * δ 3 , 0 * δ 3 , 0 * δ 1 , 0 * δ 3 , 0 * δ 2 , 0 * σ 3 2 + δ 3 , 0 2 ] ( 7 )

where the diagonal elements are not affected by the creation of a common Euronext trading platform for the three jurisdictions. This assumes that changes in the variances are not driven by the advent of Euronext, while correlations across French, Dutch and Belgian equity markets, if estimates of βi are significant, are influenced by it. Again, a symmetric version of this variance-covariance matrix is estimated for regime one.

Estimation Results

Equity markets

95. Having explained the econometric methodology, we now describe its application. We first examine equity returns. The returns are placed into two separate groups: 1) CAC40, FTSE, DAX and S&P 500 index returns and 2) CAC40, AEX and BEL20 index returns. As a first step we must estimate the residuals, uit, which are obtained by regressing the return series as in equation (1).53 Estimation of the model described by equation (3) to (5) is carried out by maximization of the log likelihood. The results are presented in Table 17 for the first grouping. Estimation of the model requires 16 coefficients to be estimated.

Table 17.

France: Equity Market Estimation Results

article image
* implies estimate is significance at 5% levels. Variables i = 1, 2, 3, 4 correspond to the returns calculated from CAC40, S&P 500, FTSE, and the DAX indices respectively.

96. For the sample period August 5 1987 to August 4 2004, we find all coefficients except gc(1) to be significant at the 5 percent level. Given that the factor loadings in regime, δi(1) are larger than those in regime zero, δi(0), regime one volatility is greater than that of regime zero.54, 55 Interestingly, the common shock mean in state one, gc(1), is below that of the mean from state zero, implying below-average returns when the common shocks are drawn from this regime. In other words, the high-volatility regime is characterized by small or negative movements in stock prices, which is consistent with the notion that periods of market turbulence are those in which equity values fall sharply.

97. The correlation across equity markets is also higher in regime one. That is, the estimated covariance terms in the variance-covariance matrix or regime one, Σ1, are larger than those estimated for Σ0. For example, the covariance estimate between the CAC 40 and the S&P 500 in regime one, δ1(1)δ2(1) = 14.52, is greater than the covariance in regime zero, δ1(0)δ2(0) = 2.18. (The corresponding correlation coefficients are 0.85 for regime one and 0.42 for regime zero.) The regime dependent covariance estimates for the other equity return pairings display the same characteristics.

98. The estimated probabilities of being in regime one—the high volatility regime (or ‘low’ return time periods)—are presented in Figure 17. These probabilities depict the timing of shifts in (and out) of each regime.56 We can see from the figure that the occurrence of this regime is more frequent (and persistent) in the latter part of the sample, which is also a period where equity markets were depicted as more volatile using the traditional annualized standard deviation measure (presented in Figure 15).

Figure 17.
Figure 17.

Estimated Regime 1 Probability

(CAC40, S&P 500, FTSE, DAX)

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

99. More importantly, these results indicate that equity returns, including those for France, tend to co-move more strongly during periods of high volatility. That is, the results indicate that much of the increase in unconditional correlation measures between French and other major equity markets observed since the mid-1990s is attributable to the higher level of volatility observed during this same period, and only a part if attributable to changes in the fundamental linkages across these markets.57

100. To what extent is the increase in correlation in the later half of the 1990s and early 21’ st century driven by the greater preponderance of volatile periods rather the effects of globalization per se? As shown by Gravelle et al (2003), a test that controls for changes in volatility, in which the null is that there is no change in degree of interlinkages between markets, can be formulated from our regime switching model. The Likelihood-Ratio test statistics is asymptotically distributed as a chi-squared with three degrees of freedom and is calculated to be 39. This exceeds by a wide margin chi-squared critical values for standard confidence levels. The results suggest that the degree of interdependence between the four equity markets has in fact increased over time, even after accounting for the affects of changes in volatility observed across these markets, which have increased unconditional correlations.

101. In Table 18, the estimation of the model which incorporates a dummy variable representing the inception of trading on Euronext on October 29, 2001, are presented. The results reveal that Belgian equity markets became moderately less correlated with the Dutch and the French markets since shares in these countries began trading on the Euronext platform. This is contrary to the notion that trading of French, Dutch and Belgian shares on a common platform would make it easier for investors to view these markets as constituting one market, and thus potentially increasing their interdependence (and correlation) over time.

Table 18.

France: Euronext Test Estimation Results

article image
* shows significance at 5% levels. Variables i = 1, 2, 3 correspond to the returns calculated from the CAC40, AEX, and BEL20 indices respectively.

102. Table 18 presents the results of the estimation of a three variable system that comprises the returns from French, Dutch and Belgian equity market indices as well as the common factor loadings that include a dummy variable (equation (6) and (7)) that takes on the value of 1 after October 29, 2001.

103. It is only the Belgian dummy parameter estimate, β3, that is found to be statistically significant. The dummy parameter is negative, implying that since the inception of Euronext trading, the Belgian equity market is in fact somewhat less correlated with French and Dutch markets than it would otherwise would be. However, the size of the effect measured by the dummy variable is not economically important and could reflect other unobserved factors that roughly coincide with the timing of the consolidation of trading on Euronext. In particular one explanation is that the Belgian equity market is very different in nature to the French and Dutch stock markets. These differences have increased, more as a result of changes brought by EMU than because of the creation of the Euronext platform. The Dutch and French markets are both very internationalized, and home to many large multinational firms. By contrast, the number of companies listed on the Belgian exchange, and its liquidity, started to decline soon after the launch of the euro. Indeed, as a result of merger and acquisition activity, a lot of initially Belgian companies are now part of larger European groups listed in Paris or Amsterdam. Hence, the Belgian market has become more local than Paris and Amsterdam, and this may account for the identified decline in correlation.

104. Estimation results indicate that, in periods of either high or low volatility, the French equity market is positively correlated (on average) with its Dutch and Belgian counterparts, indicating a persistent degree of interdependence. In addition, Figure 18 shows that the timing of high volatility (and high correlation) periods generally overlaps with those estimated for the four equity markets displayed in Figure 17. The test for no change in the degree of interdependence is strongly rejected, indicating that, after controlling for changes in volatility, the French equity market’s linkages to the Dutch and Belgian markets has grown overtime.

Figure 18.
Figure 18.

Estimated Regime 1 Probabilities

(CAC 40, AEX, BEL20)

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Bond markets

105. Estimation of the multivariate Markov-switching model was also undertaken for bond returns across France, Germany, the U.S. and U.K.

106. In estimating the independent-switching model, we followed the same steps as that for equity returns, except that the specification for the transition probabilities is the following:58

P 0 = Pr [ S t = 0 ] and P 1 = Pr [ S t = 1 ] = 1 − Pc . ( 8 )

These probabilities are unconditional and capture the idea that the high-variance shocks are unpredictable (and not very persistent in this case). The estimation results are presented in Table 19.

Table 19.

France: Bond Market Estimation Results

article image
* shows significance at 5 percent levels. Variables i = 1, 2, 3, 4 correspond to the returns from French, U.S., U.K., and German bond indices respectively.

107. As was the case for equity markets, regime one is, except for Germany, the high-variance regime. For German bond returns, the loading factor for the common shock is found to be 1.58 in regime zero, which is greater than the regime one loading factor. This means that for German bond returns, the more volatile in regime is regime zero rather than one. Moreover, German bond returns are in general more correlated to other bond returns in regime zero than in one. For example, the German bond market covariance with the French market in regime zero is estimated to be 0.60, while it is 0.37 in regime one. However, the pair-wise correlation estimates across the U.S. French and UK bond markets are higher in regime one.

108. The results reveal a lack of regime dependency over the sample period. That is, as indicated in Figure 19, the estimated probabilities of being in regime zero (the low volatility regime) at time t never climbs above 0.5, which implies that the volatility and correlation estimates do not jump from one level to another (or are not drawn from two statistical distribution. However, the estimation of an independent-regime switching specification of the model shows that there are brief, transitory periods in which bond returns are characterized as being in a low volatility and correspondingly low correlation regime (see Figure 20).59

Figure 19.
Figure 19.

Estimated Markov-Switching Regime 0 Probabilities

(Bonds)

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

Figure 20.
Figure 20.

Estimated Independent-Switching Regime 0 Probabilities

(Bonds)

Citation: IMF Staff Country Reports 2005, 185; 10.5089/9781451813630.002.A002

109. The test in which the null is no change in “economic” interdependence across bond returns yields a test statistic of 144.2 indicating the null can be rejected. As was the case for equity markets, this suggests that the degree of interdependence between the four bond markets has in fact changed over time, even after accounting for the affects of changes in volatility. This supports the hypothesis that bond markets, including the French bond market, have become more highly integrated as globalization effects have taken hold in the late part of the 1990s.

References

  • Avouyi-Dovi, S. and D. Neto (2004) “Equity Market Interdependence: The relationship between European and US stock markets” Banque de France, Financial Stability Review, No. 4, June 2004.

    • Search Google Scholar
    • Export Citation

  • Bordo, M., B. Eichengreen, and D. Irwin (1999) “Is Globalization Today Really Different from Globalization One Hundred Years Ago?,” NBER Working Paper No. 7195.

    • Search Google Scholar
    • Export Citation

  • Favero C. and F. Giavazzi (2000) “Looking for Contagion: Evidence from the ERM,” NBER Working Paper No. 7797.

  • Forbes, K. and R. Rigobon (2002) “No Contagion, Only Interdependence: Measuring Stock Market Movements” Journal of Finance, Vol. 57(5), pp. 2223–61.

    • Search Google Scholar
    • Export Citation

  • Gravelle T., M. Kichian and J. Morley (2003) “Shift Contagion in Asset Markets,” Bank of Canada Working Paper 2003-5, February 2003.

    • Search Google Scholar
    • Export Citation

  • Hamilton, J.D., (1996), Time Series Analysis, Princeton University Press, Princeton.

  • International Monetary Fund (2003), Global Financial Stability Report, World Economic and Financial Surveys (Washington, September).

  • Kim, C.-J., J. Morley, and C. Nelson (2004) “Is There A Positive Relationship between Stock Market Volatility and the Equity Premium?,” Journal of Money, Credit, and Banking, Vol. 36(3), pp. 339–360.

    • Search Google Scholar
    • Export Citation
31

Prepared by Daniel Hardy.

32

The international linkages are of long standing. The precursors of many of today’s major French financial institutions were important actors in the globalized economy of the nineteenth century.

33

Prepared by Daniel Hardy.

34

Banks from countries such as Japan, Spain, and the UK have traditionally had strong presences in certain geographical areas such as East Asia and Latin America that have fewer connections to France.

35

One mutualist bank already has limited links with mutual banks in other countries, notably Germany, including ownership links through their apex organizations.

36

Prepared by Daniel Hardy.

37

This information is taken from Axa’s published reports and presentations.

38

They may, however, have more significant exposure to risks to French companies’ assets abroad, for example, in shipping and aviation.

39

The June 2004 issue of the Banque de France’s Financial Stability Review contains a survey of the market for credit derivatives and similar instruments. The survey is summarized in the FSSA. The evidence presented there suggests that French insurance companies’ involvement in this market, in which the counterparts are largely American financial institutions, is not yet large relative to their balance sheet size.

40

Prepared by Toni Gravelle.

41

Recent work at the Bank de France looks into similar issues examined in this study (see Avouyi-Dovi and Neto (2004))

42

See Bordo, Eichengreen, and Irwin (1999), who show that since the mid-1970s, globalization has led economies and financial markets to be more integrated. See also Chapter 3 of the September 2003 issue of the Global Financial Stability Report (IMF 2003) for more on the dependence of the financial market volatility and recessions.

43

The data are described in Appendix I..

44

Both correlation coefficient and volatility measures in this study are based on an exponentially weighted moving average of past returns, where weights decay by a factor of 0.90. These exponentially weighted measures put greater weight on more recent observations rather than, as is the case with the traditional measure, an equal weighting across all observations in the sample.

45

Equity and bond correlation coefficient estimates for other country pairs are available upon request.

46

See for example Forbes and Rigobon (2002) for a discussion how volatility may bias unconditional correlation measures.

47

The data sources are Data Stream and Bloomberg.

48

Using a similar methodology, other studies find significant shifts in the degree of interdependence, after controlling for volatility, in emerging market bonds and advanced countries’ exchange markets (Gravelle et al. 2003).

49

Returns are measured as the percentage log difference in price.

50

Another reason behind this limitation is due to computational constraints. The optimization routine for calculating the parameter estimates becomes increasingly time consuming and unstable, as the number of equations and in turn parameters in the common factor system (equation (2) or (3)) increases.

51

That is, investor who become more risk averse or more uncertain about future asset fundamentals, would seek higher expected returns. This would translate into contemporaneous declines in current asset prices (i.e., negative contemporaneous returns). See Kim, Morley and Nelson (2004) for more details on this.

52

Hamilton’s (1996) textbook offers a detailed exposition of the Markov-switching econometric approach.

53

Lagged parameter coefficient estimates from equation (1) were all insignificant. As a result, actual equity returns were used in equations (3) and (4) in the place of any residuals derived from equation (1). Idiosyncratic means were also added to equations (3) to allow for the individual equity returns to have different average means rather than a common, regime dependent mean. However these were found to be insignificant and the original specification (equation (3)) is used in the analysis.

54

Note this need not be the case. The estimation procedure allows for the possibility that some loading factors to be smaller in state one δi(1)<δi(0)) while for others it is the reverse.

55

More specifically, regime one French equity volatility is, σ12+δ1,12 = 20.25 is greater than in regime zero σ12+δ1,02 = 6.44. The equity market return volatilities display a similar pattern.

56

Researchers typically define a regime switch as having occurred if the probability of being in the new state is greater than 0.5.

57

A further question, which goes beyond the scope of this note, is what causes periods of heightened volatility, and whether such episodes are themselves linked to globalization.

58

Again, lagged parameter coefficient estimates estimated from equation (1) were all insignificant. As was the case for equity returns, actual bond returns were used in equations (3) and (4) in the place of the residuals derived from equation (1). Idiosyncratic means were also added to equations (3) to allow for the individual returns to have different average returns. But these were found to be insignificant and the original specification is used in the analysis.

59

Technically, hypothesis testing should be carried out to ascertain the correct number of regimes (one versus two) and/or the correct type of switching process (Markov-switching versus independent-switching). However, this would require the calculation of bootstrapping or Monte Carlo methods and lies outside of the scope of this study.

Same Series

Other IMF Content

Other Publishers

  • Collapse
  • Expand
Cover IMF Staff Country Reports
France: Financial Sector Assessment Program—Technical Notes—Stress Testing Methodology and Results; Integration into Global Financial Markets; and Public Intervention in Financial Markets—Obstacles to Monetary Transmission
Author:
International Monetary Fund
Volume/Issue:
Volume 2005: Issue 185
Publisher:
International Monetary Fund
ISBN:
9781451813630
ISSN:
1934-7685
Pages:
121
DOI:
https://doi.org/10.5089/9781451813630.002
Figure 3.

Bank Lending to Developed Countries by Major Countries, end‐2003

(In billions of U.S. dollars)
Figure 4.

Bank Lending to Developing Countries by Major Countries, end-2003

(In billions of U.S. dollars)
Figure 5.

France: French Banks’ Loans Abroad
Figure 6.

France: French Banks’ Holdings of Foreign Bonds
Figure 7.

France: French Banks’ Holdings of Foreign Shares
Figure 8.

Equity Market Indices
Figure 9.

French Equity Correlations
Figure 10.

French Equity Correlations
Figure 11.

US Equity Correlations
Figure 12.

Bond Market Indices
Figure 13.

French Bond Correlations
Figure 14.

US Bond Correlations
Figure 15.

Equity Market Volatility
Figure 16.

Bond Market Volatility
Figure 17.

Estimated Regime 1 Probability

(CAC40, S&P 500, FTSE, DAX)
Figure 18.

Estimated Regime 1 Probabilities

(CAC 40, AEX, BEL20)
Figure 19.

Estimated Markov-Switching Regime 0 Probabilities

(Bonds)
Figure 20.

Estimated Independent-Switching Regime 0 Probabilities

(Bonds)