White's Method

Harry Dexter White used data on outstanding stock compiled by financial analysts (Neymarck, Leroy-Beaulieu, Raffalovitch, and Théry) at the end of the nineteenth century and published in contemporary financial journals.⁷

From these figures, based on five-year averages, White deducts 1 percent of the outstanding stock each year and calculates a time series of the stock in overseas assets. By assuming a rate of return on this stock, he obtained an estimate of French annual earnings from foreign capital and an estimate of annual capital flows. However, there are some problems with this approach: notably, White assumed higher returns on foreign assets than on French assets (see Table 1).

Furthermore, it is difficult to defend the premise that French international assets and incomes increased regularly without any regard to the changing economic environment: for example, the depression which beset France from 1876 to 1895. Normally, one would expect French capital exports to mirror these cycles, yet White's statistics do not reflect this.

Lévy-Leboyer criticized White for accepting the data provided by these financial analysts who he claimed were not only biased but also guilty of artificially boosting figures in order to retain resident subscribers' confidence.⁸

Financial Analysts' Direct Method

The Financial Analysts' Direct Method is based on data for new foreign issues in Paris that analysts assume were held in France. However, that was not necessarily the case: assets issued in Paris might have been purchased by investors outside of France. The assumption could lead one to over-estimate the movement of capital. Neymarck makes the ad hoc assumption that nonresidents purchased one-tenth of the assets issued in Paris and others have followed his lead when calculating estimates. Ultimately this unsupported assumption leaves the appraisal of France's net external position on shaky ground with little empirical support.

Additionally, it is difficult to ignore the fact that French residents may have subscribed to foreign assets from nondomestic markets. By contrast, we can assume that British residents did purchase foreign assets in their own domestic market in view of the fact that London was a leading financial market and competitive enough to concentrate the vast majority of foreign issues subscribed to by English residents.⁹

Nonetheless, the same argument does not apply in the case of France because of the difference in scale between the two financial markets. Moreover, the French government's well-known interference in domestic capital markets may well have driven French residents to use foreign capital markets for some of their transactions. However, one could consider that a decrease in the French-held stock in foreign assets is very likely to have been linked to a contraction of foreign issues in Paris and consider foreign issues in Paris as an approximate indicator of portfolio movements.

Existing estimates of foreign issues include: 20–22 billion francs in 1898¹⁰

; 27 billion francs in 1906¹¹; 37 billion francs in 1910¹²; 40 billion francs in 1912¹³; and 45 billion francs in 1913.¹⁴

Pierre des Essarts studied a sample group of share deposits in the Banque de France and found that the number of foreign assets held by French households in 1897 was 31 percent of the total portfolio.¹⁵

He then suggested reducing the proportion to 25 percent overall, in view of the fact that French owners of deposit accounts in the Banque de France were not representative of the average French subscriber.

Essentially, the abundance of questionable statistics prompted Lévy-Leboyer to propose an alternative method called the “indirect” method.¹⁶

Lévy-Leboyer's Indirect Method

Lévy-Leboyer reacted to the uncertain estimates of the Direct Method with an indirect estimate based on the balance of payments, which is similar to the method Albert Imlah used for Britain.¹⁷

First, he used the Tableau Général des Douanes to extract material goods, specie, transports, services, and tourism. Second, he estimated financial income and conjectures the capital movement balance (see Table 2). Although Lévy-Leboyer shifted the focus of the problem from asset stock to capital income, he did not resolve the underlying problem.¹⁸

Financial incomes logically depend on the invested stock of capital, the capital return rate, and the reinvested earnings. Lévy-Leboyer focused on the capital return rate and assumed that French foreign assets earned between the domestic asset yield in the Paris market and the rate of return on foreign British investment. He identified two periods: during the first period, from 1850 to 1876, he assumed that the rate of return on French capital exports equaled the domestic rate; during the second period, from 1877 to 1913, he considered the rate of return to be the arithmetical average of the French domestic and the British foreign investment rates of return. As opposed to the British, the French rarely invested in new countries with booming growth, therefore their profit on foreign investments must have been less than that of their British counterparts. After a financial loss correction for several years, Lévy-Leboyer arrived at the financial incomes in column 4, Table 2. By adding up partial balances (goods, services, and specie), we obtain the “net foreign direct investment” balance. The transition from flows to stock is derived from adding up the annual balances in column 6, assuming a net external position equal to zero in 1833.

This choice, however, is arbitrary and needs to be examined.¹⁹

Because capital flows are deduced from upper balances, an overvalued (or undervalued) rate of return on foreign assets amplifies (or reduces) capital flow estimates. Thus, both the direct and indirect estimates of previous authors are inadequate methods of estimating French balance of payments.

The Assertion of French Foreign Investment Economic Inconsistency

On the basis of his statistics, Lévy-Leboyer developed three major ideas: there were consistent levels of French capital exports; French foreign investment was funded by returns on previous portfolio investments and the geographical distribution of French foreign investments was not driven by economic factors.

First, the hypothesis of consistent French capital exports is no longer tenable. Three different economic periods can be distinguished. First, during the 1860s strong growth of capital exports was associated with global railroad expansion. A period of slowdown followed from 1876 to 1895, corresponding to the great depression affecting France at the end of the nineteenth century.²⁰

Lévy-Leboyer noted that, at the same time, French government bond issues attracted part of domestic savings usually devoted to foreign investments. Finally, after 1895 French capital exports expanded again.

Secondly, Lévy-Leboyer found that French investment abroad was funded by returns on previous portfolio investments. According to his calculations, capital income reached 211 percent of the total capital exports during the 1880s.²¹

And finally, Lévy-Leboyer argued that the geographical distribution of French foreign investments (Table 3) was not driven by economic motives. Comparing the geographical distribution of French foreign assets to that of British foreign assets, there is a clear predominance of French establishments in Eastern and Southern Europe, and a failure to invest in “younger countries” where the British invested 40 percent of their much larger total investment abroad between 1900 and 1913. In this view, a booming world was closed to French capital, and France remained disconnected from the international dynamics of capital flows to the New World.

Table 3 presents the geographical distribution of French foreign investments calculated by Rondo Cameron for the year 1892 and by Herbert Feis for the years 1900 and 1913, excluding colonial investment, which represented 5 percent in 1892 and 8 percent in 1913.²²

In Lévy-Leboyer's data in Table 3, a European preference seems to dominate. However, this tends to decrease by the end of the period in favor of “emerging countries.” According to Lévy-Leboyer, France remained too committed on the old continent, and did not manage to diversify its external investment. He judged this geographical polarization as being inefficient, too risky, and inconsistent with commercial flows. Lévy-Leboyer maintained that France neglected considerations of profitability in setting up business abroad. In order to corroborate this hypothesis, Lévy-Leboyer calculated the correlation coefficient (R²) between the French bond return rate and French foreign capital exports. From 1876 to 1895, when these capital export movements were quite low, French foreign investments seemed to grow as French domestic rates decreased. The correlation coefficient (R²) had the right negative sign. But for the rest of the period, from 1840 to 1875 and from 1896 to 1913, the coefficient had the “wrong” positive sign. French capital exports grew with domestic yields. Based on these statistical yields, Lévy-Leboyer rejected the financial explanation of French capital relocation.

Lévy-Leboyer's use of interest rate levels would seem to be an inappropriate indicator to choose when selecting assets.

Political Forces Dominated the Process of French Foreign Investment

The existing literature tends to contrast major creditor countries.²³

Great Britain undertook market-oriented foreign investment directed to North America, Latin America, and Oceania, with these three regions accounting for 70 percent of British investment in 1913. In contrast, investment in Europe and its periphery mainly came from France. This geographical difference was coupled with different priorities when choosing which sector of the economy to invest in. Matthew Simon estimated that from 1865 to 1914, Great Britain allocated 70 percent of its foreign investment to social overhead capital with mining coming in at 12 percent, in a distant second place.²⁴ Railroads alone represented 40 percent of British asset holdings in 1913. In contrast, during the same period, Paris focused on financing government expenditures in Eastern and Central Europe and in the Middle East. In his study focusing on the period between 1870 and 1914, Feis developed the thesis that economic considerations dominated British portfolios while political considerations drove French portfolios:²⁵

The idea that politics and diplomacy were significantly more influential in France than in Great-Britain, a concept first expressed by Feis and later extended by White and Cameron, constitutes the core proposition of the “older literature.” This thesis relies entirely on the geographical orientation of French capital abroad. France neglected booming economies such as the United States, Canada, Australia, and South-Africa while concentrating on less dynamic areas such as Southern and Eastern Europe, Russia, the Ottoman Empire, and Egypt. Lévy-Leboyer turned these choices into the thesis of “economic inconsistency” when they were compared to those made by the British. The geographical location of its financial investments abroad indicates that France wanted to underwrite its hegemony within its sphere of influence with private direct and portfolio investments in Russia or in the Balkans to complement its political and diplomatic alliances with these countries.

This one-dimensional explanation of the differences between British and French capital exports overlooks many considerations. First, there are similarities between France and Great Britain: colonial investment was particularly low for France (see Table 3) as it was for Great Britain, which weakens the political argument in both cases. Secondly, one should be cautious in accepting the stringent distinction the older literature tends to make between flows to European governments (in other words, characterizing French investment abroad), and flows to the private sector in the New World (in other words, mainly representative of British investment abroad). As Ragnar Nurkse noted, British government involvement, notably in foreign railroads, the largest category of private-sector borrowing, could have existed indirectly through grants, subsidies, or loan guarantees.²⁶

The British intervention in Egypt could also be explained by the political goal of securing British influence over the Suez Canal. Michael Clemens and Jeffrey Williamson shed new light on the connection between public and private investment.²⁷ They concluded that in Great Britain's case, flows to private sector investment opportunities were encouraged by previous investments in government-financed projects. Ultimately, these points may bring the French and British cases closer together.

As Albert Fishlow pointed out:²⁸

We intend to follow Fishlow's lead and test whether or not economic arguments can explain why Europe attracted French investments. Ideally, we would like to be able to include political variables in our test, which would make it easier to evaluate the currently accepted view. Unfortunately, the core proposition of the older literature was not actually tested but simply extrapolated from the geographical distribution of French capital exports. In fact, assessing this thesis directly appears to be impossible as its core proposition is not statistically testable because political assumptions cannot be operationally defined. Explanatory variables cannot be constructed in a way that would permit them to be used in a regression involving either political or diplomatic influences. What dummy variable would capture what Neymarck described as “informal contacts” between the French government and the Bourse de Paris?²⁹

Did diplomatic or political influences increase or decrease over this period? The evolution of the geographical distribution of French investment abroad seems to suggest that they decreased, but this does not appear to correspond to the international politics before the First World War. However, it is possible to test whether the French foreign portfolio allocation can be explained solely within an economic model.

Shares and Bonds Index and Rates of Return on Foreign Assets Held by Residents

On a weekly basis, L'Economiste Français indicated the prices of the “leading foreign bonds” (principaux fonds étrangers) for fixed income assets and “main foreign shares issued on the Paris Stock Exchange” (principales actions étrangères côtées à la Bourse de Paris) from 1873 to 1913. We have used these data to construct two separate indices: one for the price of foreign stocks and one for the price of foreign bonds. The foreign bond index is a simple arithmetic annual average of the prices on the last day of the year for 18 foreign bonds issued in Paris at various times between 1873 and 1913 (Spain 4 percent, Italy 5 percent, Austria 4 percent and 5 percent, Hungary 4 percent, Greece 5 percent and 4 percent, Germany 3 percent, Great Britain 3 percent, Norway 3 percent, Sweden 3.5 percent, Russia 4 percent and 5 percent, The Ottoman Empire 4 percent, Egypt 4 percent and 4.5 percent, Argentina 5 percent, and Brazil 4 percent). The foreign stock index is a simple arithmetic mean of the prices of six European railway shares on the last day of each year, including northern and southern Spain, northern and southern Italy, the Austrian Empire, and Portugal.

These two indices contain the most representative shares and bonds published by L'Economiste Français, but they do present certain limitations. First; a very small number of assets represents each country. Additionally, a single asset is often assumed to represent one country. Finally, many countries are not represented in either the stock or bond index. Unfortunately, no other source regularly published the foreign assets issued in Paris from the end of the nineteenth century to the first World War; so we are forced to use the available economic information found in L'économiste français, and to consider this as leading assets.³⁰

See Figure 1.

The observed yield for each foreign bond has been calculated as the ratio of coupon to price and the rate of return index as the simple arithmetic mean. We chose this simplification for three reasons. First, the amount of each subscription is unknown. Second, although calculating the return at maturity on each foreign bond would have been more rigorous, it is not possible as the data necessary for such a calculation are incomplete. In order to do this we would need to know the amount of the coupon, the dates of distribution of this coupon which are unknown for several bonds in this group, the redemption price of the bond if it was redeemable (some were), the price at the end of the year, and the advance redemption possibility, which must not be undervalued because it may significantly increase the return rate. Rather than limit our sample to those bonds on which we can calculate actuarial returns we have decided to consider a larger number of assets because portfolio diversification played a key role. Finally, because we are looking at a period of just over 20 years, from 1891 to 1913, we have decided to consider them as perpetual rents. In this case, the actuarial rate of return can be approximated by the coupon to price ratio. Thus, in this data-constrained simplification, calculating the coupon to price ratio is technically justified.

For stocks, we calculated a similar ratio.³¹

But it must be noted that important breaks in dividend series occurred. In the case of missing values, we have used a previous dividend. Over the period, railway dividends tended to be quite steady. Nevertheless, the calculated rate of return on foreign shares could not be considered very strong. See Figures 2 and 3.

Stock in French Foreign Assets

Our estimates of French foreign stocks are based on data on new issues: a rehabilitation of the direct method of estimation. Françoise Marnata's studies provided us with residents' foreign shares and bond subscriptions from 1892 to 1913.³²

These data come from the Crédit Lyonnais archives and provide detailed information on residents' purchases. For the prewar and the interwar period the “Financial Studies Office” of the Crédit Lyonnais registered each and every purchase of French and foreign financial assets made by French residents. It constitutes a complete and accurate historical source to estimate the foreign financial stock held by residents. Referring to the methodological appendix developed by this office, these statistics were designed to accurately calculate the precise amount French residents paid for each foreign or domestic issue in Paris; the second purpose was to assess the distribution and the evolution of investors' preferences between fixed and variable income assets, and between guaranteed and nonguaranteed loans.

The construction of estimates for the stock in overseas assets, however, does raise some statistical problems. First, we need an estimate of the initial stock. Lévy-Leboyer supposed a stock equal to zero in 1833.

³³

This introduces a bias at the beginning of the series as capital flows occurred before that date. However, overtime the bias decreases because the major capital outflows occurred after this date. A more important problem arises because the value of the outstanding assets changes continuously. To account for this, one needs to make the distinction between shares and bonds, as they follow different cycles. Lévy-Leboyer's stock simply cumulated the sums of annual flows, as follows:

where: S_t is the stock of portfolio outflows at date t, and F_t is the corresponding flow.

In order to obtain a stock at market value, which is appropriate for portfolio analysis, we have: set the initial stock in 1891 according to financial analysts=estimates; used the annual flows calculated by Marnata after this date; adjusted the market value of the stock using share and bond prices published by L'Economiste Français weighted by the geographical distribution of French foreign assets (from Table 3).

Our procedure can be summarized as follows:

where: SFB_t is the “Stock of Foreign Bonds”; FFB_t is the associated flow calculated by Marnata, with figures expressed in French francs; SFS_t is the “Stock of Foreign Shares”; and FFS_t is the associated flow calculated by Marnata, with figures expressed in French francs.

For the year beginning 1891, we consider a stock equal to the simple average of all estimates published for this year by financial analysts, following Marnata's assumed portfolio of three-fourths bonds and one-fourth shares in 1891.

FBPI_t is the “Foreign Bond Price Index” calculated from L'Economiste français weighted by the geographical distribution of French foreign assets (from Table 3).

FSPI_t is the “Foreign Share Price Index” calculated from L'Economiste français weighted by the geographical distribution of French foreign assets (from Table 3). See Table 4 and Figure 4.

Estimating the value of foreign financial assets is a prerequisite to making an appropriate portfolio analysis. According to the portfolio theoretical analysis, arbitrage concerns the portion of the global portfolio held in foreign assets.³⁴

In the following section, we test whether a portfolio model can account for French foreign financial asset relocation from 1892 until 1913, a period which provides us with reliable data from primary sources.

Previous Use of this Methodological Framework in Historical Analysis

Edelstein developed a portfolio model evaluating the microeconomic foundations of English foreign investments during the period from 1870 to 1913. In his model, Edelstein focused on English demand for stock in American railways which was Great Britain's main investment abroad during this period.³⁶

He found that there was a positive correlation between British holdings of American railways and the railway asset return rate, the British domestic investment return rate as well as with British wealth. We cannot focus our analysis on individual financial assets because we do not have the data to replicate this procedure. Instead we compare the overall returns of our aggregate foreign stock and bond indices to overall returns of French domestic stocks and bonds. Because French foreign investments were targeted to “Europe and its periphery” (see Table 3) we have chosen variables relevant to that region in our model.

We would like to follow Edelstein's method of investigating the driving forces behind English capital exports and his formulation of the debate in terms of “push or pull factors.”³⁷

Did English foreign financial asset demand correspond to a wealth effect in the country that exported capital, or was it pulled by the financial needs of host countries (for example, the public debts of issuing countries)? We want to investigate whether French financial investment in Europe and its periphery was pulled by the industrial and financial needs of the host countries that offered investment opportunities in public funds or in share participation with sustained industrial growth perspectives? Were foreign public debt issues a conclusive argument for French expansion abroad? Such questions are a source of debate and controversy. Cameron and Lévy-Leboyer attempted to point out a sort of crowding-out effect of domestic investment by foreign investment.³⁸ We do not share this view: arguing that foreign investments are a pure loss for domestic growth seems to be an extreme point of view. Instead, we consider foreign investment as part of a portfolio optimizing strategy. International savings transfers are logically motivated by expected yield return. Moreover, the return on investment ultimately increases domestic wealth. In addition, it cannot be assumed that foreign investment reduces domestic investment by the same amount. If domestic returns are not profitable, domestic investment may not occur. We also consider a potential “push effect” by assuming that the behavior of international investors depends on their desired wealth level. If wealth held in foreign financial assets decreases, because of a drop of foreign stock market prices, the investor is motivated to save more in order to restore his desired wealth, which implies an increase of the domestic rate of savings. On the other hand, an increase of stock market prices, which raises the investor's wealth, may discourage domestic savings.

Issues in Applying Economic Modeling to Historical Data

Basic international portfolio models (CAPM) give the appropriate framework to answer our questions. The microeconomic basis of the portfolio approach establishes that asset demand functions in foreign currencies depend on speculative and hedging investment behavior. The optimal portfolio is divided into two parts: one insensitive to yields (usually qualified by minimum variance) reflecting risk aversion; the other, called speculative portfolio, sensitive to yield expectancies.³⁹

In the case of an international portfolio, an exchange rate risk must be introduced, which motivates investors to hedge against it.⁴⁰ The demand function of foreign assets is composed of three arguments, which reflect portfolio choice: a constant hedging against inflation; an exchange rate level to hedge against exchange rate risk; and a yield difference for speculation.

Implementing this theoretical framework requires the following: The shares of foreign funds and equities in the global portfolio, the exchange rate of the Franc against a panel of currencies, and domestic and foreign observed yields.

Collecting appropriate data on these variables presents some practical problems.

One might conclude that exchange rate hedging is a simple problem because this period is situated during the gold exchange standard, an era characterized by fixed exchange rates. However, Marc Flandreau and Jacques Le Cacheux showed that in Europe, during the gold standard and despite gold convertibility, exchange rate risk existed.⁴¹

There were concentric circles of stability: The franc belonged to the first group of steady currencies, the lira and the peseta to a second group, characterized by a plus-or-minus 10-percent fluctuation amplitude against the franc; the ruble to a third group, characterized by extreme volatility. One can easily suppose that the more volatile the currency, the more its variances influence portfolio choices. Therefore, according to theoretical predictions and observed exchange rate fluctuations, in our regressions, we use an average exchange rate of the franc against a panel of main currencies reflecting the global geographical structure of French capital export (as illustrated in Table 3).⁴²

A second problem concerns the relationship between observed return rates (calculated as the ratio of coupon to price) and expected return rates appropriate for investment decisions. Here, we are restricted by data and forced to make a difficult choice. We assume, as others have, that economic agents have rational and true expectations (in other words, what they plan actually occurs). Therefore, in our econometric test, this enables us to take the calculated rate of return of the t+1 year as an indicator of the expected rate of return in t.

Last but not least, objections can be made concerning the relevance of a portfolio approach applied to an annual series of capital stock and yields. Arbitrage and portfolio choices are usually supposed to concern a short-term horizon (a monthly or even daily adjustment). Nevertheless, as legitimate as this objection may be, we feel that the periodicity of historical data should not prevent one from learning about the past by using economic modeling.

Empirical Investigation

Given the models presented, two tests have to be carried out: one concerning foreign fixed income financial assets and the other concerning foreign variable income financial assets held by French residents. In the former case, the endogenous variable is the “ratio of foreign bonds to total French and foreign assets held by French residents.” In the latter, it is the “ratio of foreign equities to total French and foreign assets held by French residents.” In both cases, two kinds of variables will be tested as explanatory factors of the variations in these shares:

Financial variables: the expected rate of return on domestic and foreign assets. According to international CAPM models, these two rates of return emphasize speculative behavior. Very often, empirical portfolio models wrongly take only short-term yield differences into account; Here, only long-term yields are used, which is economically justified as specie, with short-term capital flows, are excluded from the calculated stock; and

Nominal variables: a weighted exchange rate and variables for “pull factors” (Public Debt Index (PDI) in “Europe and its periphery area” in the case of foreign fixed income assets, and Industrial Production Index (IPI) of the same area for foreign shares).⁴³

The Public Debt Index of “Europe and its periphery” is an indicator of “pull factors” that can play a role in foreign fixed income financial assets allocation. The PDI is built as follows: public debt in local currency of Great Britain, Denmark, Norway, Sweden, Russia, Austria, Hungary, Switzerland, Germany, Netherlands, Belgium, Spain, Italy, Bulgaria, Serbia, Romania, the Ottoman Empire, and Egypt are taken from International Historical Statistics.⁴⁴ They are converted into francs and presented as an index, basis 100 in 1913.

The data set is composed of variables that are all annual series covering the 1890 to 1913 period (see Table 5).

Time Series Properties of Variables

The analysis of the stationarity of the series is an essential step to determine the correct econometric specification to implement in our study. A stationary series is characterized by a time-invariant mean and a time-invariant variance. We have implemented three unit root tests.⁴⁵

The results of these tests are not reported here because of limited space, but they can easily be summarized as follows because clear patterns emerge from them: the unit-root null hypothesis cannot be rejected at the 5-percent level of significance for most of the nine variables under consideration.⁴⁶ The only exception is for r, where the KPSS tests indicate that this variable is stationary around a linear trend. However, the Schmidt-Phillips and Elliott tests do not reject the existence of a unit-root in this variable. Tests have been applied to the variables taken in first differences and reject the nonstationary hypothesis for all series. This leads us to conclude that the series in this study are well characterized as an I(1) Process (integrated of order one), some with nonzero drift.⁴⁷

Parsimonious Dynamic Models and Economic Interpretation

Having demonstrated the presence of a unit-root in all series of our data set, we now test whether these variables are cointegrated. We present the results of the two error correction models appropriate for the analysis of cointegrated times-series. The former concerns the share of foreign bonds, and the latter the share of foreign stocks held by French residents. More precisely, the objective is to study the interdependence between these variables taken in levels, without making any a priori hypothesis on the value of the coefficients linking them, and testing the existence of long-term relationships.

For each hypothesis to be tested, the database is respectively composed of KOB, r′, r, PDI, e, and KAC, rA′, rA, IPI, e.⁴⁸

The econometric methodology adopted here (see Appendix 1 for further details) yields our two final models (estimated with maximum likelihood) with parsimonious dynamics,

⁴⁹

and fully identified long-term relationships, which conform to our theoretical priors.⁵⁰

These models are both consistent with the theory and congruent with the data.⁵¹

The first concerning the share of foreign fixed income financial assets in the global portfolio is given by:⁵²

This model enables us to reconstruct the economic motives of French financial property abroad:

In the long term, French assets invested abroad follow the level of public debt in “Europe and its periphery,” indicating a “pull effect.” One may think that the importance of international public debt issued by these nearby countries as well as its broad diffusion in Paris guaranteed strong liquidity for these assets, and constitute a factor of safety for French investors. In addition to this economic factor, one also finds financial components explaining French foreign asset behavior. The share of foreign fixed-income financial assets in the global portfolio is positively correlated to the expected return on foreign assets and negatively correlated to expected returns on domestic investment, which is consistent with the portfolio theory.⁵³

It reveals an underlying logic based on the search for profit: French investors neglected French investments as the expected rate of return decreased, and turned to foreign assets when the anticipated rates of return on these investments increased. They restructured their portfolio in response to this yield difference.

In the short term, we notice the influence of public debt variation, which means that the French demand for this kind of asset grows with European public supply. The exchange rate emerges as a short-term explanatory variable but it is insignificant in the long term. In dynamics, the portfolio of foreign fixed-income financial assets positively reacts to the variation of the French exchange rate: an increase in the value of the French franc against that of other currencies in the panel facilitates foreign asset purchase by residents. Finally, one finds in dynamics the positive impact of the variation of the expected rate of return on foreign assets: The more this expected rate of return increases, the more likely it is that residents will hold foreign assets, which conforms to theoretical predictions.

Our results are consistent with the following mechanism: The long-term objective of the French investor (here, the share of foreign bonds in the overall portfolio) is negatively correlated with the level of domestic expected rates of return; it is however; positively correlated to the anticipated level of foreign rates of return, and also positively correlated to the public sector borrowing requirements of “Europe and its periphery.” In the short term, the investor corrects the distance to this long-term target according to the variation of the domestic rate of exchange, the variation of the public-sector borrowing requirements of the area, and the variation of foreign assets' expected rate of return (all signs correctly oriented). Thus, it has been possible to highlight the economic and financial arguments underlying the diversification of French capital abroad from the turn of the nineteenth century until 1914.

For the second model concerning the share of foreign stocks held by French residents in the overall portfolio, we obtain the following parsimonious dynamic specification:

⁵⁴

Here again, the estimated model conforms to theoretical predictions. In the long-term relationship, it appears that the share of foreign stocks held by French residents in the overall portfolio responds positively (responds negatively) to the level of the foreign (French) expected rate of return. As previously seen, beside financial components of the French asset demand, one finds a “pull effect,” which is linked in this case to the level of the industrial production in the area. This result seems very logical as growth perspectives in the area sustain the residents' demand for foreign shares. In the short term, it must be noted that the lagged dependent variable is the only one to intervene in the relationship. Investors adapt their behavior according to the past variation of the portfolio held in foreign shares. Moreover, one can note that the adjustment to the target is relatively short (one period).