BRITISH IMPERIALISM AND FINANCIAL GLOBALIZATION BEFORE 1914

Between 1865 and 1914 more than £4 billion flowed from Britain to the rest of the world, giving the country a historically unprecedented and since unequalled position as a global net creditor—“the world's banker” indeed; or, to be exact, the world's bond market. By 1914 total British assets overseas amounted to somewhere between £3.1 and £4.5 billion, as against British GDP of £2.5 billion.²

This portfolio was authentically global: around 45 percent of British investment went to the United States and the colonies of white settlement, 20 percent to Latin America, 16 percent to Asia and 13 percent to Africa, compared with just 6 percent to the rest of Europe.³ Adding together all British capital raised through public issues of securities, as much went to Africa, Asia, and Latin America between 1865 and 1914 as to the United Kingdom itself.⁴

It has been claimed by Michael Clemens and Jeffrey Williamson that there was something of a “Lucas effect” in the period between 1880 and 1914, in other words that British capital tended to gravitate towards wealthy countries rather than relatively poor countries.⁵

Yet the bias in favor of rich countries was much less pronounced than it has been in more recent times. In 1997 only around 5 percent of the world's stock of international capital was invested in countries with per capita incomes of a fifth or less of U.S. per capita GDP. In 1913, according to Maurice Obstfeld and Alan Taylor, the proportion was 25 percent.⁶ Very nearly half of all international capital stocks in 1914 were invested in countries with per capita incomes a third or less of Britain's, and Britain accounted for nearly two-fifths of the total sum invested in these poor economies.⁷ The contrast between the past and the present is striking. Whereas today's rich economies prefer to “swap” capital with one another, largely bypassing poor countries, a century ago the rich economies had very large, positive net balances with the less well-off countries of the world.⁸

How important was the empire as a destination for British capital? According to the best available estimates, more than two-fifths (42 percent) of the cumulative flows of portfolio investment from Britain to the rest of the world went to British possessions.⁹

An alternative measure—the imperial proportion of stocks of overseas investment on the eve of the First World War—was even higher: 46 percent. And about half of this amount went to relatively poor British colonies, not to the much more prosperous areas of white settlement. An obvious hypothesis might therefore be that investors a century ago were more willing to invest money in relatively poor countries because a high proportion of these countries were not sovereign states but were under the political control of the investors' own country.

Did membership of the British Empire give countries access to the British capital market at lower interest rates than they would have paid as independent states? Contemporaries and an older historical literature had little doubt that it did. Writing in 1924, Keynes noted that “Southern Rhodesia—a place in the middle of Africa with a few thousand white inhabitants and less than a million black ones—can place an unguaranteed loan on terms not very different from our own [British] War Loan.” It seemed equally “strange” to him that “there should be investors who prefer[ed] … Nigeria stock (which has no British Government guarantee) [to] … London and North-Eastern Railway debentures.”¹⁰

More recently, Michael Edelstein has argued “that the British capital market treated empire borrowers differently from foreign borrowers.”¹¹ An obvious explanation for an “imperial discount” on bonds issued by British colonies is that they were in some way guaranteed by the British government and therefore in a legal sense indistinguishable from British bonds in terms of default risk.¹² However, Edelstein rejects this explanation:

Another explanation may lie in the effect of legislation specifically calculated to encourage investors to buy colonial bonds. At the turn of the century, two laws were passed, the Colonial Loans Act (1899) and the Colonial Stock Act (1900), which gave colonial bonds the same “trustee status” as the benchmark British government perpetual bond, the “consol.”¹⁴

At a time when a rising proportion of the national debt was being held by Trustee Savings Banks, this was an important stimulus to the market for colonial securities.¹⁵ However, the importance of this legislation should not be exaggerated. The average difference between noncolonial and colonial yields was above 250 basis points between 1880 and 1898 and about 180 basis points between 1899 and 1913—in other words the premium on noncolonial bonds was actually higher before the Colonial Loans Act and Colonial Stock Act came into force. Prior to the First World War, these acts were the only formal encouragements to investors to favor colonial bonds.¹⁶

There are, however, other, less formal reasons why prewar investors may have incorporated an imperial discount when pricing bonds. The Victorians imposed a distinctive set of institutions on their colonies that very likely enhanced their appeal to investors. These extended beyond the Gladstonian trinity of sound money, balanced budgets, and free trade to include the rule of law (specifically, British style property rights) and relatively noncorrupt administration—among the most important “public goods” of late-nineteenth-century liberal imperialism.¹⁷

Debt contracts with colonial borrowers were more likely to be enforceable than those with independent states. It would be rather puzzling if investors had regarded Australia as no more creditworthy than Argentina, or Canada as no more creditworthy than Chile.

For a number of reasons, then, it is possible that the imposition of British rule practically amounted to a “no default” guarantee; the only uncertainty investors had to face concerned the expected duration of British rule. Before 1914, despite the growth of nationalist movements in possessions as different as Ireland and India, political independence still seemed a fairly remote prospect for most subject peoples. At this point even the major colonies of white settlement had been granted only a limited political autonomy. Thus, in the words of P. J. Cain and A. G. Hopkins: “One of the key reasons why the colonies could borrow cheaply [was that] they offered almost complete safety.”¹⁸

DETERMINANTS OF BOND SPREADS

The possibility exists, nevertheless, that other considerations mattered more to investors than the extent to which a country's sovereignty had been reduced by imperialism. The recent literature on the determinants of risk premia has centered on these other factors.

An alternative approach focuses on monetary policy rather than colonial status. Michael Bordo and Hugh Rockoff argued that adherence to the gold standard worked as a credible “commitment mechanism,” reassuring investors that governments would not pursue time-inconsistent fiscal and monetary policies.¹⁹

Investors rewarded this binding policy commitment by charging—ceteris paribus—lower risk premia. The gold standard worked in this respect as a “Good Housekeeping seal of approval.” A commitment to gold convertibility, they calculate, reduced the yield on a country's bonds by around 40 basis points.²⁰ Using a somewhat larger sample, Obstfeld and Taylor confirmed that gold standard membership lowered spreads.²¹ In this analysis, therefore, it was membership of the informal and voluntary gold “club” rather than membership of the British Empire that lowered the yields paid by some emerging markets. As Obstfeld and Taylor conclude, “Membership in the British Empire was neither a necessary nor sufficient condition for preferential access to London's capital market before 1914.”²²

As a contingent commitment, however, membership in the gold standard was nothing more than a promise of self-restraint under certain circumstances. Independent countries on gold were not members of some kind of monetary union. They retained the right to suspend convertibility in the event of an emergency such as a war, revolution, or a sudden deterioration in the terms of trade. Such emergencies were in fact quite common before 1914. Argentina, Brazil, and Chile all experienced serious financial and monetary crises between 1880 and 1914. By 1895 the currencies of all three had depreciated by around 60 percent against sterling. This had serious implications for their ability to service their external debt, which was denominated in hard currency (usually sterling) rather than domestic currency.

A second hypothesis is that investors were primarily interested in the fiscal policies of borrowing countries. Marc Flandreau and Frédéric Zumer have recently suggested that the most important risk factors were public debts, the corresponding amount of debt service, and the relation between these burdens and tax revenues.²³

They find that, once differences in indebtedness are taken account of, gold standard adherence was insignificant. In addition, they present evidence that contemporary economic thinking about default risk centered on debt sustainability and the soundness of public finances.²⁴

A third determinant of risk premia may simply have been political events. According to Ferguson, revolutions, governmental crises and wars were regarded by nineteenth-century investors as increasing the likelihood of defaults by the countries affected.²⁵

Finally, Clemens and Williamson have identified demographic characteristics, natural resource endowment, and education as significant determinants of yield spreads.²⁶

To determine whether or not membership in the British Empire genuinely lowered borrowing costs, it is therefore imperative to control for these and other factors. British colonies may simply have been able to borrow at lower rates than other foreign countries because they were on the gold standard, had more sustainable fiscal policies, were less susceptible to political crises, or were simply better situated relative to trade routes and temperate climatic zones.

YIELD DATA AND ECONOMIC CONTROL VARIABLES

We constructed the largest possible sovereign bond database for the period 1880–1913. Price data for government bonds quoted and traded in the London market were copied by hand from the leading financial publication of the time, the Investor's Monthly Manual. Some additional quotations were taken from the London Stock Exchange Weekly Intelligence, the London Stock Exchange's official weekly gazette. The bonds chosen had to pass three strict criteria to qualify as benchmark issues. First, they had to be payable in London in either sterling or gold, enabling us to focus exclusively on country risk and to ignore the currency risk inherent in bonds denominated in other currencies.²⁷

Secondly, the selected bonds had to be issued in large volumes and actively traded. Finally, the bonds needed to be long-term, typically of a maturity of over ten years, and to have quotations for at least three consecutive years.

The resulting dataset includes securities from 57 independent countries, colonies, and self-governing parts of the British Empire: in other words, almost the entire universe of foreign borrowing in the London market, reaching not only “from the Cape to Cairo” but also from Boston to Buenos Aires and from Budapest to Beijing.²⁸

The rationale for constructing such a broad sample was to avoid the regional biases that characterized previous studies. Bordo and Rockoff used observations for just ten countries, all either European or American.²⁹ The two most recent investigations of pre-1914 bond yields by Obstfeld and Taylor and by Flandreau and Zumer were based on samples of around 20 countries. The samples in both cases were predominantly European and American. Quite clearly it is difficult to form robust conclusions about the significance of colonial status without including data for at least some Asian and African countries.

Table 1 shows the summary statistics for our current yield series.³⁰

In total, we count about 1,450 observations, roughly 900 for independent countries from Europe, America, Asia, and Africa and about 550 for issuers from the British Empire, drawn from these four continents as well as Australasia. Immediately obvious from the yield data is the significantly lower average yield of Empire borrowers (3.89 percent) compared with the yields of independent countries (6.30 percent).

Older research on financial investment in the age of high imperialism looked only at raw yield data, thus leaving open the possibility that lower colonial spreads were a function of better economic “fundamentals” rather than the explicit or implicit guarantees to investors stemming from empire membership.³¹

The only way to say for sure that there was an empire effect is therefore to regress yield spreads against an appropriate range of additional control variables. The obvious question is which variables to include. In our view, there are powerful methodological objections to the inclusion of anachronistic indicators such as debt to GDP ratios.³² Self-evidently, people usually do not base their actions upon concepts that have not yet been invented or upon figures nobody yet calculates.³³ Rather, if we want to determine how nineteenth-century investors made their decisions, we need to model their behavior deductively on the basis of the data that were available to them at that time.³⁴

The economic data were collected from primary and secondary sources.³⁵

As anyone familiar with the financial press of the period knows, there was a plethora of publications available to investors. Standard reference publications such as Fenn's Compendium, the Investor's Monthly Manual (henceforth IMM), the Stock Exchange Weekly Intelligence and the Corporation of Foreign Bondholders Annual Reports collected and analyzed statistical data on government borrowers in a manner not unlike that of the handbooks on equity investments pioneered by Moody's in the United States. In addition to this dedicated financial press, there was a rapidly growing number of more general statistical publications.³⁶

The subtitle of the 1898 edition of Fenn's Compendium, the self-proclaimed “doyen of all financial books of reference,” neatly summarizes what economic indicators the City of London had access to: it was “a handbook of public debts containing details and histories of debts, budgets and foreign trade of all nations, together with statistics elucidating the financial and economic progress and position of various countries.”³⁷

In many respects, the main problem for contemporaries was not so much the raw data in the numerator—whether public debts, debt service charges, or exports—but the denominator. In the absence of a direct measure of a nation's output such as gross national product, a concept then its infancy, it was far from easy to compare the fundamental resources of different countries. Population was generally acknowledged to be an unreliable choice, though it had the advantage of being readily available, thanks to fairly regular and accurate censuses, and was often used to denominate export capacity. However, in more sophisticated analyses of fiscal sustainability, the debt burden tended to be related to public revenues or to export earnings.³⁸ The same was true of budget and trade balances.

Drawing on the records of the Service d'Études Financières of the Crédit Lyonnais, Flandreau and Zumer have suggested that debt service to revenue was the contemporary indicator that best measured the creditworthiness of borrowers.³⁹

However, for a number of reasons we chose to stick to the more traditional debt to revenue ratio. First, in contemporary statistical publications, the overall debt burden was far more frequently given, and was also, it seems, less frequently subject to revisions. Secondly, as the debt service itself is determined by the interest rate, it is questionable whether it should be used as an independent variable to estimate the interest rate. Nevertheless, we can also work with debt service data for a far larger number of countries than previous studies and will show that our key findings do not depend on the choice of a particular fiscal measure.

Another indicator watched by contemporaries was the budget deficit to revenue ratio. As Cain and Hopkins have argued, the principles of “Gladstonian finance”—which aimed at budget surpluses during peacetime in order to repay existing public debt—were all but sacrosanct in the eyes of the “gentlemanly capitalists” of the City of London.⁴⁰

In addition, we collected information on those countries that breached the “London consensus” on good housekeeping by defaulting on their obligations; the Annual Reports of the Corporation of Foreign Bondholders contain detailed information on defaulters. Because default damages reputation, we constructed a control variable for default within the preceding ten years.⁴¹

Apart from public debt data, the second class of economic statistics readily available to late-nineteenth-century investors related to foreign trade. That there was a link between trade and creditworthiness was obvious to contemporaries because countries needed to earn foreign exchange in order to service their external debts. Export capacity was also seen as a proxy for wealth and the state of economic development. Because we wanted to capture the risks stemming from both large external deficits and low levels of international trade, we collected data for both the trade deficit and the sterling value of exports per capita. Modern studies of country risk tend to use GDP per capita as a proxy for risk-reducing factors such as more stable politics or better institutions. The City of London had to settle for something less than that before the First World War, but it was looking for analogous information.

Given the importance attributed by some scholars to gold standard adherence, we also wished to control for the positive effects of being on gold. The question of whether or not a country's currency was—de facto or de jure—convertible into gold is in itself a difficult issue; indeed, it is far from clear-cut even for well-researched economies such as Austria and Italy, both of which “shadowed” the gold standard without officially having fully convertible currencies.⁴²

Nonetheless, because considerable attention has been paid to the role of gold adherence in reducing country risk, our estimations include two dummies for gold standard adherence. Following Christopher Meissner as well as Obstfeld and Taylor, we use the “strict” gold coding.⁴³ We also take account of Obstfeld and Taylor's point that “the market's view of gold standard adherence [ought] to depend on whether a country [was] in full compliance with its debt contracts.”⁴⁴ Finally, we took the idea seriously that internal or external political conflicts may have been important determinants of yield fluctuations.⁴⁵

Table 2 summarizes the core economic control variables used in the statistical analyses. It will be seen that they are comparable, though not identical, to the variables used by Flandreau and Zumer.⁴⁶

By applying them to a much larger sample of countries, however, we are able to pose a question they did not consider: How far yield spreads reflected the fundamental differences in political status that distinguished independent borrowers from those that were members of the British Empire. The important point to be borne in mind is that our approach may tend to underestimate the empire effect by assuming that it is possible to separate colonial status cleanly from “fundamentals” such as fiscal, monetary, and trade policy, or indeed political stability, all of which were almost by definition affected by the imposition of British rule. As Table 2 shows, empire borrowers were slightly less indebted than independent countries. They were more likely to be on the gold standard than independent states, though we still have enough cases of British possessions off gold to distinguish empire membership from gold standard membership. Exports per capita were markedly higher inside the empire than outside it (the dominions and colonies exported about four times more per capita than independent countries), which tends to confirm conventional wisdom about the relative openness of the imperial trade regime.⁴⁷

ESTIMATING THE EMPIRE EFFECT

In order to gauge the size of the empire effect on country risk premia, we first investigated the relationship between the spread over consols, i.e., the difference between the yield on a country's bonds and the yield on consols, and the economic control variables discussed previously. We look to the coefficient of the empire dummy (coded 1 if a borrower was a British possession) for an estimate of the empire effect.

The estimation of panel or time-series cross-section data has become a standard method of exploring large datasets in economic history. Pooling enables us to increase the amount of informative data, through combining variation across countries with variation over time. It also makes it possible to control for exogenous events affecting all units at a point in time, thus to control for time effects—a crucial advantage here because we need to take account of global interest rate shocks affecting all countries in a specific year.⁴⁸

We borrow an estimation method that has become the standard for datasets like ours in quantitative research in comparative political economy: OLS with panel corrected standard errors (PCSE).⁴⁹ This method allows for the inclusion of a unit-specific AR1 term to correct for serial correlation, while retaining the unbiased OLS coefficient estimates and calculating reliable “panel-corrected standard errors.”⁵⁰

Our research agenda is complicated by the fact that we are interested in coefficient estimates for a largely time-invariant variable. There are only three borrowers in our sample which became (de facto or de jure) colonies within the period: Egypt in 1882, and the Transvaal and the Orange Free State after the Boer War in 1900. As case studies their experiences are instructive. Spreads on Egyptian bonds were as high as 500 basis points in 1880. After the imposition of British rule and the restructuring of public finances and public debts, they fell to 270 basis points in 1882 and declined further to about 130 by the end of the 1880s.⁵¹

A similar story can be told for the southern tip of Africa, where the two Boer republics of Transvaal and Orange both had bonds quoted in London in the 1890s with yields of about 200 to 300 basis points above the British benchmark. After the war in 1900, the new colonies contracted large loans in London (with the blessing of Westminster), increasing their debt-to-revenue ratios from the low levels of the 1890s (about 100 percent of revenues) to more than 500 percent. At the same time, the yield spread fell to around 20 basis points over consols.

The main implication of this limited time-variation of empire membership is that there are two ways to get a reliable estimate of the financing advantage of colonial borrowers. In a standard fixed-effects model the empire effect would appear in the country fixed-effects. The estimated unit effects of the model would show whether or not empire issuers had on average lower overall spread levels than independent borrowers.⁵²

The drawback is that all time-invariant differences are included in the fixed effects. The alternative is to drop the fixed-effects and to run a pooled OLS regression. Yet this approach could suffer from omitted variable bias if cross-sectional heterogeneity were no longer captured by different intercepts.⁵³ However, if the unit effects are spanned (or accounted for) by a linear combination of the time-invariant regressors, then pooled OLS would still be the estimator of choice.⁵⁴ As will be seen, the pooled and the fixed-effects models yield very similar results.

THE EVIDENCE

The results of our benchmark regression, 1, lend strong support to the idea of an empire effect. All other things being equal, the yield on a bond would be about 100 basis points lower if the issuer came from the British Empire.⁵⁵

The finding is backed by the number of observations (1,294), which is more than double the number in previous investigations with a comparable number of controls.⁵⁶ In regression 2 we test whether or not it makes a difference to include country fixed-effects. We drop the empire dummy and include individual country dummies (but keep the year-dummies from our benchmark specification). For a summary comparison, we can now look at the mean fixed effect of the unit effects of the empire group and the mean of independent countries. The result is reassuring: a statistically significant (the null here is a mean of zero) group effect appears, and the difference between the empire and independent countries is both significant and large at more than 150 basis points. The other coefficients match closely those of our benchmark regression, 1. In regression 3 we limit our sample to “developing countries,” in other words capital-poor countries.⁵⁷ Here, the empire effect reaches more than 180 basis points, suggesting that being part of the British Empire was particularly important for the borrowing of less-developed African and Asian colonies.

But can we be sure this is truly an empire effect? Could there be a third factor (correlated with, but independent of, colonial status) that increased market confidence? Obviously, it was not the geographic position, measured by distance from the core or climatic conditions. British colonies were spread over all continents and climate zones. Nor can empire be considered a proxy for the impact of European settlement or the introduction of liberal parliamentary institutions, because the extent of settlement and political representation varied greatly across the empire, yet the exceptionally low country-effects apply equally to the dependent empire and the more autonomous dominions.⁵⁸

What about the macroeconomic control variables? Our results support Flandreau and Zumer's emphasis on public finance as a determinant of pre-1914 bond spreads: the debt to revenue ratio is correctly signed and significant in all regressions. As noted, Flandreau and Zumer have argued that contemporary investors paid more heed to the ratio of debt service to public revenues. It is obvious that, by virtue of their lower spreads, British colonies had to pay less interest on their debt than independent countries. We would therefore expect the empire effect to get weaker if one relies exclusively on the debt service ratio—but not to disappear. Regression 4 provides the corresponding empirical test. It demonstrates that the empire effect does not depend on the choice of the debt indicator. The empire effect remains highly significant. Only its size is, unsurprisingly, somewhat smaller if the debt service serves as the only debt control—about 82 basis points in this specification.⁵⁹

Clearly, empire mattered beyond the differences in debt burdens between borrowers, however scaled.⁶⁰

As for gold standard adherence, our results provide mixed evidence and point to the need for further analysis. Although the gold standard variable (conditional on no default) is correctly signed and has the expected effect of a 15 to 40 basis point reduction in spreads, it passes the significance test only in some regressions. Moreover, we found that, unlike the debt to revenue ratio, the gold standard dummy is rather sensitive to changes in the estimation specification, to influential observations, and to differences in the coding criteria.⁶¹

As expected, both defaulters and previous defaulters were heavily penalized by the City, but the budget deficit seems to have had no significant effect on spreads. (One possible explanation is that investors did not regularly follow the budget balances of various countries, but concentrated on debt indicators instead as an excess of expenditure over revenues would show up in the debt figures.) The picture is different for the external trade indicators. Richer countries (measured by exports per capita) paid less interest and, other things being equal, a country that ran an export surplus would have lower borrowing costs. Our estimations also lend some support to the argument that current political factors were important spread determinants: any incidence of internal political conflict raised spreads by as much as 70 basis points (see Table 3).

Why do we find strong evidence for an empire effect of about 100 basis points, where Obstfeld and Taylor concluded it did not exist? Apart from the bigger sample, different controls for time-specific asset market shifts could drive the result. Such shifts can be controlled for by using time-dummies (or any other market-wide measure) that affect all borrowers in a given year, which is what we opted for in the benchmark regression. The time dummies from our regression show a clear downward trend over the period, briefly interrupted in the crisis years of the early 1890s (Figure 1). The picture mirrors the general trend towards spread convergence discussed by Obstfeld and Taylor as well as Flandreau and Zumer.

However, Bordo and Rockoff as well as Obstfeld and Taylor took a different track. They included a measure of systematic risk—a weighted “world spread” over consols in every year with country-specific slopes or “betas”—following the capital asset pricing model (CAPM) and its predictions about pricing of assets according to systematic risk.⁶²

The coefficient on this variable indicates how closely the spread of a country conformed to the variation in the “average” risk of foreign bonds as perceived by British investors. A coefficient greater than one would signal that a bond of a given country was more strongly affected by an increase in market risk than the average borrower. This could have an impact on our estimation of the empire effect, if colonies as a group experienced much less correlation with the market-wide risk than independent countries.

To test this, we constructed a debt-weighted world spread for any given year.⁶³

In estimation 5 we regress the spread on the usual controls plus the average world spread in a fixed-effects framework, and look again at the group effects. The difference between colonial borrowers and independent countries does indeed fall dramatically, and more importantly, it turns statistically insignificant at conventional thresholds.

Why does the empire effect fade once one switches to CAPM-style controls? Our large sample enables us to identify the underlying causes with great certainty: the key driver for the different findings is the much lower covariance of colonial interest rates with the average risk of foreign bonds (or significantly lower individual betas in CAPM-language).⁶⁴

The mean correlation of empire borrowers (0.24) with the market-wide average risk premium on foreign borrowing was close to zero (the risk-free rate) and far below the coeefficient of independent countries (1.34). The difference is of considerable magnitude (1.10) and highly significant. Following the logic of the CAPM, colonial bonds were safe assets, whereas bonds of independent countries carried a high systematic risk. In other words, the country risk of colonies was much less sensitive to changes in the perception of the average riskiness of foreign investment. Even in times of crisis (during the Baring crisis) colonial risk premia remained low. The empire effect was therefore strongest during crisis periods; it was lower when the market sentiment towards foreign investment was more positive.

The lower sensitivity of colonial assets to market risk actually confirms the empire effect hypothesis. Investors treated colonial bonds differently, as reflected in the exceptionally low betas.⁶⁵

They were, in effect, slightly higher-yielding substitutes for risk-free British consols. In a specification with CAPM-style controls, however, the fundamentally different risk characteristics of colonial borrowers and independent economies are effectively swept away by the country-specific coefficients. It is not surprising, then, that the country dummies do not show a large empire effect anymore. This would seem to explain why previous studies considered the idea of an empire effect to be an optical illusion of contemporaries.

BOND SPREADS WITHIN THE BRITISH EMPIRE

The last part of our analysis of spread determinants is devoted to an equally old question in the study of the British Empire: Who profited most from preferential access to the London capital market—the dependent empire, the Dominions, or India? Ceteris paribus, which did investors see as the safest place to put their money? Looking in detail at loan issues in the period under investigation, Lance Davis and Robert Huttenback concluded that “within the British Empire, India consistently paid less for capital than either the dependent colonies or those with responsible government.”⁶⁶

Does this finding—based on groupings of yield data without further controls—stand up to the inclusion of economic controls for the level of debt, the external position, and the state of development?

Regressions 6 and 7 exploit our dataset to give a more comprehensive answer (see Table 4). They essentially confirm the conclusions of Davis and Huttenback; Indian bonds had a distinctly lower risk premium than either dependent or self-governing borrowers within the empire. In both estimations, we found India's financing advantage to have been worth about 30 basis points. This result is not surprising because, unlike some other colonial bonds, “Indian government bonds carried the backing of the British government and were listed in the official rosters of the London stock exchange with ‘British funds’.”⁶⁷

COUNTRY RISK AND CAPITAL FLOWS

The City of London viewed British possessions as safe places to invest. As a result, distant colonies gained access to the London capital market at cheaper rates than comparable sovereign states. But what implications did this have for the amounts of capital that flowed from Britain to her empire? In other words, did the empire effect mean more capital as well as lower interest rates? Any answer to this question depends both on counterfactual argumentation and ceteris paribus assumptions and must therefore remain highly speculative. Nevertheless, such questions have been raised before and played an important role in the debate on the costs and benefits of British imperialism, and we cannot therefore ignore them.⁶⁸

Edelstein, for example, estimated that had they been independent, colonies would have paid twice the actual interest rate. As a consequence, they would have received only as much capital per capita as other comparably developed independent countries. He concluded on that basis “that the non-white-settler colonies would have had British investments one fifth their actual £140 and £480 million levels in 1870 and 1913.”⁶⁹ By the same token, the self-governing parts of the empire would have received about 30 percent less capital.

Figure 2 allows a first visual impression of the patterns of international borrowing in the London capital market, and underlines the risk aversion of British financial investors. More than 60 percent of aggregate public borrowing in the boom years between 1900 and 1913 was concentrated in the low-risk segment of the market (spreads of less than 100 basis points), while another 30 percent went to public borrowers whose spreads were less than 200 basis points above the British consol. This tendency looks even more pronounced if borrowing is denominated by population. It therefore seems reasonable to assume that, if the colonies suddenly had gained their independence, capital flows would have fallen rather substantially. Higher default risks would have depressed the expected return for investors, rendering foreign investment less attractive and drying up the supply of capital.

To illustrate these considerations in an empirical framework, we can try to estimate the relationship between the amount of foreign borrowing in London and risk premia. We would expect to find that capital flows decreased with higher country risks, because investors tend to limit their exposure to high-risk assets.⁷⁰

However, whereas risk premia are certainly an important determinant of capital flows, comparably important roles are played by investment opportunities, the institutional environment and many other country-specific fundamentals. Despite these complexities we try to derive some illustrative insights from estimating the determinants of capital flows in a simple cross-sectional model accounting for country-specific “pull” factors and global “push” factors.⁷¹

First, we use five-year averages to level out cyclical effects, reduce the impact of outliers, and get a reliable picture of the underlying factors. Second, we control for a number of other plausible determinants, regressing the yearly average capital inflows per capita not only on spreads, but also on population (to control for country size), the period average of population and export growth (for demographic and growth trends), and on the ratio of rail miles in operation to country size (as a proxy for opportunities for “reproductive investment”). Finally, we include U.K. interest rates as a “push” factor.

Regression 8 includes the full sample of British investment, while regression 9 limits the sample to independent countries and regression 10 to the less developed independent countries outside Western Europe and North America (see Table 5). The main finding that runs through all regressions is that higher risk premia were indeed associated with lower flows. The elasticity of flows to independent countries in relation to the risk premium, about 0.3 at average regressor values, was also substantial. For the less developed countries in our sample, the impact was even greater. Other things being equal, a country received more than twice as much capital per capita if its risk premium was only half of the developing country average. And cutting risk premia by half is what empire membership probably implied for the poor African and Asian colonies. In view of the highly exploratory character of the estimation, we are inclined not to read too much into these results. However, the significance of market size and the positive sign on population growth are in line with previous studies.⁷²

At the very least, it seems legitimate to conclude that the higher country risks that would have been consequent on an “Edwardian decolonization” would, in turn, have reduced capital flows to Britain's former possessions. For British investors did not place voluminous bets on risky governments; they extended relatively more credit to the low-risk segment of the market. Given this preference, the appeal of investing in the empire is obvious.