The Legacy of War on Fiscal Capacity∗

Didac Queralt† May 13, 2016

Abstract States make war, and wars make states. The second clause of Tilly’s dictum assumes that the fiscal effort that states exert to wage war persists over time. This paper investigates the effect of premodern warfare on long-term fiscal capacity as a function of two types of war financing instruments: taxes and loans. Tax-financed wars are argued to exert lasting effects on state capacity, as new taxes require enhancements of the state apparatus as well as complementary fiscal innovations. Loan-financed wars may not contribute to long-term state capacity, as countries might default once war is over, thus preempting any persistent fiscal effect. Importantly, the way war is financed might be endogenous. To cope with this possibility, I exploit unanticipated crashes in the nineteenth-century international financial market, which temporarily banned states from borrowing regardless of their (un)observed characteristics. The analysis shows that countries that fought war while international lending stopped have today higher fiscal capacity, measured by various tax ratios as well as the size of the tax administration. Altogether, the paper advances the conditions under which war exerts positive and lasting effects on state capacity.

∗First Draft: June 2015. I am grateful to Ben Ansell, Laia Balcells, Thomas Brambor, Mark Dincecco, Hector Galindo, Francisco Garﬁas, Scott Gates, Margaret Levi, Pilar Nogues-Marco, Shanker Satyanath, Peter Schram, Ken Scheve, David Stasavage, Joachim Voth, Tianyang Xi, and seminar participants at Stanford University, Carlos III University, and CEIFR-Peking University for comments and suggestions. †Institute of Political Economy and Governance, iPEG-Barcelona; [email protected]

1 1 Introduction

War is devastating but it also offers an unmatched opportunity to transform the state. The magnitude of the resources that a country needs to amass in order to wage a military campaign offers rulers the incentives to invest in state-making while reducing domestic resistance to taxation. War clears the path to fiscal centralization (Dincecco, 2011; Tilly, 1990), the professionalization of the tax administration (Ardant, 1975), and the adoption of various forms of revenue generating policy, such as mercantilism (Findlay and O’Rourke, 2007), excises (Brewer, 1988), and income taxes (Scheve and Stasavage, 2010). Fiscal innovations are often accompanied by complementary organizations, including treasuries, national cus- tom services, and central banks (O’Brien, 2001), improved budgeting technologies (Dincecco, 2011), and new mechanisms of conscription (Fischer and Lundgreen, 1975) that contribute to expand the scope of the state. Far from disappearing, the financial innovations that make war possible are expected to exert lasting effects on the extractive capacity of the state (Ar- dant, 1975; Besley and Persson, 2011; Brewer, 1988; Desch, 1996; Dincecco and Prado, 2012; Hoffman and Rosenthal, 2000; Mann, 1980; Rosenthal and Wong, 2011). In other words, states make wars as much as war makes states. The bellicist hypothesis of the creation of the modern state has received a lot of attention. Both qualitative and quantitative evidence suggest that this hypothesis accounts particularly well for the process leading to state-building in the Western World (Bonney, 1999; Dincecco, 2011; Ertman, 1997; Gennaioli and Voth, 2015; Hintze, 1975). By contrast, the evidence is mixed in other regions: some scholars find positive evidence for the bellicist hypothesis in Latin America (Thies, 2005), Asia (Barkey and Parikh, 1991; Stubbs, 1999), and Africa (Thies, 2007), while others find negative results: Centeno(2002), and Kurtz(2013) in Latin America; He(2013) and Taylor and Botea(2008) in Asia; and Herbst(2000) and Dincecco, Fenske and Onorato(2016) in Africa. The lack of traction of the bellicist hypothesis in the periphery is attributed to a combination of moderate incidence of inter-state warfare and strong ethnic divisions, both of which are expected to weaken the incentives to invest in

2 fiscal capacity (Besley and Persson, 2011). In this paper, I advance a new explanation for the mixed evidence of the bellicist hypothesis around the globe. In particular, I argue that war itself is not as important as the way it is financed. That is, I claim that the effect of war on long-term fiscal capacity is conditioned on the ultimate mix of taxes and loans. States that do not only rely on debt to wage war, but implement fiscal reforms to raise revenue through taxation, should benefit the most from war-making, at least in the long-run. Financing war through taxation requires investing in the fiscal capacity of the country, involving the transformations that are associated with the bellicist hypothesis: that is, fiscal centralization, the enhancement of the tax administration, and the adoption of new taxes and complementary fiscal institutions that, ultimately, rubricate the legitimacy of the state to assess private wealth and monitor compliance. These reforms are expected to outlast war. The main reason for that is political: national rulers are reluctant to cede power back over the newly created fiscal institution once war comes to an end, as it serves them to consolidate their power vis-à-vis local elites and other political rivals (Besley and Persson, 2011; Dincecco, Federico and Vindigini, 2011; Gennaioli and Voth, 2015; Tilly, 1990). By contrast, financing war with loans, particularly if these are external —as it is usually the case in the developing world—, does not necessarily lead to any of the above. Some states invest in enhancing the tax capacity to service debt once war is over, while others do not, thus preempting any significant enhancement of fiscal capacity with respect to prewar years. Reputation concerns do not suffice to prevent default (Bulow and Rogoff, 1989; Reinhart and Rogoff, 2009). First, defaults are accepted in reasonable circumstances: e.g. losing a war (Slantchev, 2012; Tomz, 2007). Second, even serial defaulters eventually regain access to international markets, either after substantial debt forgiveness or in exchange for state monopolies and land properties (Jorgensen and Sachs, 1988; Marichal, 1989). One way or the other, war-making does not necessarily translate into an enhanced capacity to tax. Additionally, access to external credit, specially when it is inexpensive, might lead

3 to excess borrowing and destabilize national accounts in ways that prevent post-war fiscal consolidation (Centeno, 2002; Tin-bor Hui, 2004). To sum up, I argue that while financing war with taxes (even if only partially) makes a clear contribution to state making, the effect of financing war with loans, mainly, is uncertain. To test this hypothesis, this paper assesses the effect of types of war financing on long- term fiscal capacity, while controlling for alleged causes of the weak traction of the bellicist hypothesis outside Europe: i.e. low warfare and high ethnic divisions. In particular, I investigate the effect of participating in pre-modern inter-state war between 1816 and 1913 on current fiscal capacity, measured by the percentage of personal income tax to GDP, as well as the size of the tax administration, value-added taxes, and modern census technologies. To avoid sample selection issues that characterize previous work, the analysis includes over a hundred countries from all around the globe. Rulers cannot impose new taxes without negotiating (Levi, 1988). The adoption of new taxes comes with payoffs, chief among them, political compensations in the form of representation (Bates and Lien, 1985). To minimize political payoffs, rulers are expected to borrow from abroad when they have the opportunity, that is, when they have access to the international capital markets (Slantchev, 2012; Shea, 2013). Based on this theoretical corpus, the empirical analysis assumes that, whenever rulers have access to external credit, they will borrow to finance the means of war.1 Building on this assumption, I first compare the effect of fighting war when a given country has access to the international capital market vis-à-vis times in which the country has no access because it is in default. In the latter case, I expect the ruler’s incentives to invest in domestic taxation to be strongest, thus contributing to expand the stock of fiscal capacity permanently. Results confirm the expectation. More importantly, they are robust to alternative sources of revenue, these being domestic borrowing, money printing, and financial repression.

1The only exception is France under Napoleon.

4 A second battery of empirical analyses addresses the endogeneity of having access to external loans in the nineteenth century. Specifically, I exploit crashes in the international capital market to identify periods in which states cannot borrow from abroad irrespective of their (un)observed characteristics. This quasi-natural experiment, allows me to study what is the effect of fighting war in periods in which international capital dries relative to fighting wars in periods in which international capital flows. Results show that fighting war when the international lending market is down is associated with higher fiscal capacity today. To the contrary, making war while international credit flows is at best inconsequential. Results are robust to sample changes, and various war and statehood definitions, as well as to initial and contemporaneous factors that condition the effect of war on long-term fiscal capacity: population density (Karaman and Pamuk, 2013), ethnic fractionalization (Besley and Persson, 2011; Besley and Reynal-Querol, 2014; Kurtz, 2013), war intensity measured by the number of casualties (Dincecco and Prado, 2012; Rasler and Thompson, 1985), geographic controls, and World War I participation (Scheve and Stasavage, 2010). Once access to international lending is exogenized, I address potential selection issues into war. That is, countries that go to war despite the unavailability of external loans might be different in ways that affect fiscal capacity today. This form of endogeneity is addresses threefold: first, omitted variable bias is minimized by controlling for initial state capacity, measured by a state antiquity index (Putterman, 2007) and the capacity to conduct a modern census by 1820. Second, anticipation issues are addressed by focusing on wars initiated before the sudden-stop of external credit, thus disconnecting the decision to go to war from availability of external credit. Third, I investigate the effect of war on fiscal capacity by focusing on countries that do not choose to go to war, but are dragged into it. Additionally, the Appendix includes a reduced-form model in which war by country i is instrumented by war participation by its adjacent neighbors. These analyses confirm that war make states if it is substantially financed with taxes, not loans only. The last empirical section addresses the short-term effects of war-making as a function of

5 types of war-financing instruments. If the effect of war on fiscal capacity is cumulative, we should expect some evidence of war-making on fiscal capacity by the end of the period under consideration. In the absence of reliable, crossnational tax data, two alternative proxies to state capacity are used: having conducted a modern census by 1913, and the length of rail lines by the end of that year. Results suggest that countries that fought wars lacking (having) access to external loans between 1816 and 1913 present a higher (lower) probably of having conducted a modern census at the eve of World War I, as well as a more (less) dense railroad network. Overall, results imply that the rulers’ incentives to fight back domestic opposition in order to adopt more extractive taxes and intrusive fiscal administration are strongest when rulers lack cheaper financial alternatives. Financing wars with domestic taxes seems to be a last, politically costly resort, that nevertheless turns beneficial in the long-run. This paper contributes to a growing industry that revisits Tilly’s dictum by conditioning the effect of war on initial conditions. For instance, Karaman and Pamuk(2013) argue that the effect of warfare on state capacity is conditional on the complementarities between urbanization and regime type. Similarly, Kurtz(2013) claims that Latin America does not capitalize the war effort because it lacks enough political cohesiveness to invest in public goods such as a fiscal capacity.2 Fascinating as they stand, these accounts do not address the radically different international context in which countries in the periphery are created as compared to European nations: one in which external credit, even for new countries in the periphery, is unprecedentedly cheap as a result of excess savings generated by the industrial revolution in Western Europe (Taylor, 2006). In this paper, I vindicate Charles Tilly’s theory and conceptual toolkit while adapting it to the distinct international context of the long-nineteenth century. A close reading of his work suggests that state building is as a function of both war-making and domestic capital access. European powers (e.g. Britain, the Netherlands, France) capitalized the

2Refer to Besley and Persson(2011) for a formalization of the argument.

6 fiscal effort of war because they disproportionately borrowed domestically. In the absence of an efficient international lending market that supplied inexpensive capital, as soon as the sixteenth century, European rulers turned to domestic merchants to raise the means of war, either by taxing or borrowing from them (alternatively, capital-intensive cities were coercively annexed for the same purposes). The inward-turn of war financing led to political reform that eventually ended in early forms of limited government (arguably, some states went further than others). Countries in the periphery did not face the same capital constraints as their European counterparts had when they were involved in state-building. From their very inception, states in the periphery had access to unprecedentedly inexpensive external loans despite their low institutionalization and lack of international reputation (Tomz, 2007). Access to easy money, I argue, weakens the incentives to develop domestic credit institutions, expand taxation, and undertake political reform conducive to responsible fiscal policy. In other words, cheap external credit facilitates the means of war while preempting fiscal (and political) reform associated with war-making. Others have pointed out the perverse effects of cheap credit on the incentives to tax domestically: Centeno(2002, p.130-3) suggests this kind of artificial wealth distorted rulers’ incentives to finance war with taxes in post-independence Latin American. Similarly, Thies (2007, p.728) speculates with this possibility to explain the mild effect of warfare on state capacity in post-colonial African. Shea(2013) emphasizes the changing incentives to tax when rulers have access to external credit at affordable prices. In this paper, I advance the political economy underlying the choice of loans vs. taxes, articulate the implications for long-term fiscal capacity, and test them against historical series while addressing endogeneity of credit-access and war-making. Ultimately, the paper offers a better understanding of the conditions under which war exerts positive and lasting effects on state building. Interestingly too, the perverse effects of cheap external credit advanced in this paper speak to the schol- arship analyzing the effect of different forms of non-tax revenue on political accountability:

7 natural resources (Morrison, 2009; Ross, 2001), foreign aid (de Mesquita and Smith, 2013; Moss, Pettersson Gelander and van de Walle, 2006), and ores from the colonies (Drelichman and Voth, 2011). What remains of the paper is organized as follows: First, I articulate the political economy of war financing. Second, I model current fiscal capacity as a function of war participation and credit access in the long-nineteenth century. These models address stepwise endogeneity of credit access and war participation. Third, I investigate the incremental nature of fiscal capacity building by showing the effect of war-making and credit acces at the eve of WWI. Lastly, I revisit the bellicist hypothesis in light of the empirical evidence.

2 The Political Economy of War Financing

The two most common instruments of war financing are taxes and loans (Slantchev, 2012; Sprague, 1917).3 Resorting to one or the other is a matter of possibility —has the state enough capacity to tax its citizens and/or access to lending markets?—as much as of political opportunity —who gains and who loses upon borrowing and taxing (Flores-Macias and Kreps, 2013; Stasavage, 2011). Taxation is politically delicate, as it involves some form of extraction from elites, the masses, or both. In taxing the wealthy, rulers can rarely impose new taxes without their consent. Often times, taxes have to be negotiated and consulted with economic elites (Levi, 1988; Tilly, 1990). In pre-modern Europe, rulers offered elites political payoffs that, ultimately, established modern political rights (e.g. the Magna Carta), and Parliamentary representation, later on (Bates and Lien, 1985).4 Taxing the masses was not easier, especially when it was accompanied by the military draft: in such circumstances, political concessions were required to prevent tax revolts from below (Ardant, 1975; Hintze, 1975;

3Printing money, financial repression, selling offices, and confiscation are addressed below. 4Elites’ resistance was also smoothed out with trade privileges (Ekelund and Tollison, 1981; Queralt, 2015), and the sale of offices (Hoffman, 1994). Either way, financing wars through taxation impinges a political cost on the ruler, as it weakens his advantage position vis-à-vis the economic elites.

8 Tilly, 1993). Ultimately, power-sharing institutions were the “price and outcome” of bar- gaining with different members of subject population in overcoming resistance to financing with taxation the means of war (Tilly, 1990, p.64). Financing war by borrowing from domestic sources might come with similar political costs (North and Weingast, 1989). However, domestic borrowing requires levels of capital accumulation that cannot be taken for granted, let alone in the new world.5 When domestic credit markets are small, rulers may borrow from abroad, a practice that, despite being common in pre-industrial Europe (Stasavage, 2011), accelerated after the Napoleonic Wars in parallel to the globalization of financial markets (Reinhart and Rogoff, 2009; Lindert and Morton, 1989; Suter, 1992). Crucially, borrowing from abroad does not suffer from the same political costs and administrative challenges attached to taxation. That is, rulers do not have to concede political rights or representation to international lenders; a good margin suffices. External borrowing does not come with the uncertainties of tax yields either, thus facilitating the planification of military campaigns (Slantchev, 2012). Lastly, external loans prevent sudden tax hikes that might disrupt household allocation decisions, while passing the tax burden to next gen- erations and minimizing political opposition to war (Barro, 1979). Given the short-term advantages of financing wars with external loans, it is hardly surprising that nineteenth- century warfare in the periphery (Asia, Eurasia, Latin- and North America, and Eastern and Southern Europe) was heavily financed with external debt (Centeno, 2002; Flandreau and Flores, 2012; Marichal, 1989). Having access to external credit is consequential to understanding the conditions under which wars make states precisely because taxes and loans may not exert the same lasting effect on fiscal capacity. The bellicist hypothesis implicitly assumes that states service debt following military conflict: that is, rulers exert a fiscal effort (e.g. enhance tax collection) to honor their debt once war is over. However, debt service is uncertain. It depends on,

5For the challenges to create domestic ﬁnancial markets in the periphery, see Calomiris and Haber(2014), della Paolera and Taylor(2013), and Taylor and Williamson(1994).

9 first, the financial capability of the state —e.g. war losers are less capable to meet fiscal obligations (Centeno, 2002; Gennaioli and Voth, 2015; Slantchev, 2012)—, and second and most importantly, the ruler’s willingness to repay (Reinhart and Rogoff, 2009; Tomz, 2007). Some honor their debt in full and on time, others do not. Certainly, few countries outrightly repudiate their debt (e.g. Turkey and Mexico in the second half of the nineteenth century, or Russia in the early twentieth century); most rene- gotiate it (Lindert and Morton, 1989). However, renegotiating debt weakens the incentives to invest in fiscal capacity. First, settlements might not involve a transfer of money. Instead of raising taxes to repay, rulers may exchange public properties (including Crown monopolies and revenues, mining, or lands) for old bonds. We can find examples of this in the Latin American world throughout the nineteenth century (Marichal, 1989; Vizcarra, 2009). Second, default might come with substantial debt forgiveness, a common practice in the period considered in the empirical analysis (Lindert and Morton, 1989). For instance, the 1870’s default settlements in Latin American represented effective debt relief of almost 50% for Bolivia, Chile, Peru, and Paraguay, and 40% in Costa Rica.6 Third, even when debt is not forgiven, renegotiations usually involve reductions in interest and extensions of maturi- ties that, in turn, may relax the incentives to enhance the extractive capacity of the state (Marichal, 1989). All in all, financing war with loans does not necessarily translate into an enhanced fiscal capacity. By contrast, the more war is financed with taxes, the stronger fiscal capacity should be after military conflict. Financing war with taxes implies financial innovations that transform the physiology of the state (Ardant, 1975): namely, new and professional administrations, central banks, fiscal unifications or advances in indirect and direct taxation outlast war, thus permanently enhancing the fiscal capacity of the state with respect to prewar levels. Key to explain the legacy of war on state capacity, this collection of “self- strengthening reforms” (Tin-bor Hui, 2004) tend to outlast war. Once in place, it is in the

6Data from Marichal(1989, Table 4) and Jorgensen and Sachs(1988).

10 best interest of the national ruler to keep them around, as they serve her to consolidate power vis-à-vis local elites and political opponents (Besley and Persson, 2011; Dincecco, Federico and Vindigini, 2011; Gennaioli and Voth, 2015; Tilly, 1990). Ultimately, the more wars are financed with taxes relative to other means, the more likely is to achieve high fiscal capacity in the long-run. To illustrate the basic intuition behind this logic, I briefly present some evidence for Chile, one of the countries with higher state capacity in Latin American today.7 Chile participated in three wars in the nineteenth century: the Confederation War, 1836-1839, against Peru and Bolivia; the Chincha Islands War, 1865-1867, against Spain; and the Pacific War, 1879-1883, against Peru and Bolivia again. The first war was a smaller one, but the latter two required a vast mobilization of resources at a national scale.8 Importantly, these wars were fought in different financial contexts: the Confederation War and the Pacific War (first and last) were fought while Chile was in default. By contrast, the Chilean-Spanish War was fought while the country had access to the international financial market. In light of the political payoffs of taxation, rulers are generally inclined to finance war with loans rather than taxes. More specifically, I expect rulers to resort to taxation only when they are pushed by circumstances: that is, when they are banned from more politically neutral options such as external borrowing. The way Chile financed war in the nineteenth century is consistent with this logic. Figure1 plots the share of tax revenue and public foreign debt as percentage of GDP from 1833 to 1913. The years in which Chile was at war are shaded. However, I differentiate wars fought while Chile was in default (light gray) —thus excluded from the international markets—, from wars fought while Chile had access to the international credit market (dark gray).

Figure1 here

The ﬁrst lesson drawn from Figure1 is that wars are ﬁnanced with both debt and taxes.

7And one of the few for which historical series of ﬁscal outcomes are available. 8The Confederation War did not achieve the 1,000 battle deaths, a convention to be included in standard war databases.

11 However, consistent with the argument advanced in this paper, the debt/tax mix is less favorable to taxes when rulers have access to the international credit market. Take the two larger wars, the Chilean-Spanish War, 1865-1867, and the Pacific War, 1879-1883: In 1865, Chile was allowed to borrow from international lenders, and it did. Chile financed war against Spain with external loans, which rose over 350% with respect to prewar years. In stark contrast, tax revenue remained virtually unchanged during this period. Things were different in 1879: Chile was again at war, but this time the country was in default, thus excluded from the international credit market. Because war costs were pressing, Chile had to finance the Pacific war out of its own pocket. Among other fiscal reforms, “[i]n May of 1879, in desperation, Congress passed the mobiliaria, the income tax it had rejected the previous year” (Collier and Sater, 2004, p.147). Following this reform, tax revenue in Chile reached unprecedented levels, and, importantly, yields never went back to prewar levels. That is, the fiscal effort exerted to wage war outlasted conflict precisely because it involved investment in fiscal capacity. Importantly, significant tax reforms were only adopted when rulers were forced to by circumstances. The Chilean example suggests that the effect of war on fiscal capacity hinges on the financial instrument used to wage war. The rest of this paper investigates whether this logic generalizes around the globe, taking into account potential endogeneity issues in access to external credit and the decision to go to war.

3 Data

Cross-national conflict-specific data regarding how war is financed is not available in any systematic way. In order to test how types of war financing affect fiscal capacity in the long-run, I propose the following strategy: comparing the relative impact of war in periods in which countries have access to external capital to periods in which they do not. The logic of this test is based on the political economy of war financing: When external funding is not

12 an option, I expect the incentives to resort to domestic taxation to be strongest. That is, it is only when they are pushed by circumstances, that rulers assume the political costs of raising taxes to finance war. In order to identify periods in which countries do not have access to external loans, I follow two strategies: first, I focus on episodes of default (or endogenous access to credit), and investigate how war fought while being in default (thus, forcing rulers to invest in domestic taxation) affect long-term fiscal capacity as compared to war fought while having access to international lending. Second, I exogenize external credit access by exploiting crashes in the international financial market, which temporarily bans access to external loans irrespective of (un)observed characteristics of the country. In order to estimate the effect of warfare on fiscal capacity, I follow Dincecco and Prado’s (2012) strategy. These scholars use nineteenth-century measures of warfare and war casualties as instruments of current fiscal system. Dincecco and Prado(2012) find that countries that fought more wars and suffered the largest number of casualties during the 1816-1913 period have higher ratios of direct taxes to GDP by the year 2000. The lower cut-off, 1816, is deliberately picked to maximize the number of cases in the sample. Most countries in the periphery are created only in the nineteenth century. The upper cut-off, 1913, serves three purposes: First, by focusing on pre-modern wars, one guarantees that the fiscal effort that a country exerts is driven by military need. The boom in welfare spending following World War I (Lindert, 2004) makes it harder to identify the effect of war on fiscal capacity, as re- cipients of social programs create their own demand for higher taxation. In other words, by dropping post-1913 wars, we make sure that the fiscal contract is mainly military. Second, the two World Wars are unprecedented in lethalness, resource mobilization, but also in the rules of social fairness. Scheve and Stasavage(2010) show that even elite’s resistance to high taxation disappeared in light of the unparalleled human costs, thus rendering the Great War hardly comparable to conventional warfare. Third, and related, given the huge financial

13 costs of the Great War,9 most participants were countries with high fiscal capacity to begin with. Including total wars in the analysis would only exacerbate problems of selection. Even though the design follows Dincecco and Prado(2012), I model the impact of pre- modern warfare on current fiscal capacity differently. Specifically, I condition the lasting effect of war on types of financial instruments, not just casualties, while using an finer proxy for current fiscal capacity: namely, the ratio of personal income taxes to GDP.10 This tax is agreed to represent the end point of fiscal capacity building (Mares and Queralt, 2015; Tilly, 1990; Webber and Wildavsky, 1986). It is highly extractive, pushes the incidence of taxation onto the higher income groups, and requires a sophisticated bureaucratic apparatus capable of assessing private sources of income and monitor compliance of an atomized tax base. In light of its implementation challenges, the income tax results from purposeful investment in the tax administration, thus, it makes states. The income tax is not the only important tax in high fiscal capacity states (e.g. Sweden relies heavily on indirect taxation). However, no state can be classified as having high fiscal capacity without effective income taxation. For this reason, this tax sets a clear benchmark from which we can establish how far each country has gone in building tax capacity.11 Among the various income taxes, I work with the Personal Income Tax (PIT), which requires the capacity to enforce income-tax withholding, and a sophisticated organization to assess and monitor compliance of a highly-atomized tax base. PIT data (normalized to GDP) is drawn from various sources. Chief among them is the IMF Global Financial Statistics (GFS).12 Consistent with the theoretical claims, I work with PIT raised by the central government, as war is expected to makes states by centralizing fiscal powers. To minimize influence of

9Refer to Centeno(2002, p.21) and Ardant(1975, p.224) for a detailed account of the unprecedented costs of WWI. 10Models will control for war casualties, nevertheless. 11Alternative proxies of fiscal capacity, such as total revenue to GDP (Besley and Persson, 2011; Hendrix, 2010), or direct tax to GDP (Dincecco and Prado, 2012) confound high fiscal capacity with tax-handles (Musgrave, 1969), such as the land, poll, business, trade and excise tax, all of which are usual in the developing world, where the income tax is only modestly effective. 12IMF data represents almost 80% of the data. Appendix Table A-6 shows that data augmentation with alternative sources does not bias the estimates of interest. Further data details can be found in Appendix Section A.

14 abnormal values, I work with the average PIT value as a percentage of GDP from 1995 to 2005.13 Since PIT might capture both capacity and willingness to tax, a second outcome variable is considered, one that emphasizes the infrastructural component of fiscal capacity: the Size of the Tax Administration circa 2005, measured as the number of staff employed by the tax administration per thousand capita. Additionally, for robustness purposes, Appendix Table A-12 report models of Value-Added Tax (VAT) as a percentage of GDP, even though this form of indirect taxation is less costly to implement than the income tax (Bird and Gendron, 2007). The nineteenth century is popularly known for being a peaceful time. However, the hundred-year peace is only a phenomenon of the developed world (Flandreau and Flores, 2012). Wars did take place in this period, but mostly outside Western Europe. Appendix Figure A-2 plots the location of these wars based on current state borders. That Figure confirms that, consistent with the hundred-year peace, there is little military conflict in the European core. However, countries in other regions, most prominently Asia and Latin America, were at war for almost half of the period (also with European powers). Most wars in this period were inter-state, involving European powers but also non- internationally recognized states (Butcher and Griffiths, 2015). Wars were fought against colonial powers and also between neighboring countries, specially in Africa, Latin America and Southeast Asia. In an effort to move beyond the experience of the developed world with war-making, I work with Wimmer and Min’s (2009) war database, which includes all military disputes by internationally and non-internationally recognized states around the world since 1800.14 The use of non-internationally recognized states in the analysis assumes that theses political entities exert a fiscal effort in financing war comparable to recognized states. This is the case of, for instance, the wars of independence in Latin America (Centeno, 2002; Marichal,

13For robustness, Appendix Table A-6 reports outcomes for slightly diﬀerent time periods. 14To make it into this dataset, military conﬂicts must experience more than 1,000 casualties.

15 1989), the African wars before and after the arrival of the Europeans (see Reid(2012) and Gardner(2012); Frankema(2011), respectively), or the inter-state wars over succession disputes in Southeast Asia (Butcher and Griffiths, 2015). Importantly, results do not hinge on this assumption, as shown in both Table6, in which only states recognized by the international system by the time they go to war are considered, and Table8, in which Wimmer and Min(2009) data is replaced by the more conservative war list in Sarkees and Wayman’s Correlates of War.15 Wimmer and Min’s data stand out in three additional ways: first, wars are mapped onto current state boundaries, making it possible to track which states should inherit the legacy of war making, as well as investigate the effect of fighting war at home and abroad.16 Second, Wimmer and Min(2009) distinguish civil from secessionist war. As part of the robustness tests, these type of wars (defined as fights against the political center with the aim to establish an independent state) are considered. After all, they may contribute to revenue maximization in a similar fashion than inter-state wars.17 Third, in Wimmer and Min(2009) non-proxy wars fought by colonial subjects against third territories are attributed to the colonial subject, and not to the metropolis, thus maximizing the match between war makers and fiscal outcomes: e.g. wars fought by Egypt under Ottoman rule are not attributed to Turkey but Egypt, which led the military campaign. Altogether, I consider 147 armed conflicts between 1816 and 1913: 114 of them are inter- state wars, and 33 are secessionist wars. For each of these wars, I establish whether access to external credit is possible, either by identifying periods of default or exploiting crashes in the international lending market. Specifically, I count the number of years at war in which countries had access to the international capital market during this period, and the years in which they did not. To account for characteristics of war other than duration, I control for

15Refer to Appendix A for further details. 16Most wars can be easily matched to current states. A minority cannot: these are extinct political entities the territory of which overlap with more than one modern state. Refer to Appendix A for further details about matching old to new political units. 17Civil wars are excluded from the analysis because their contribution to state building is yet to be established. Refer to Appendix Section A for further details.

16 the number of casualties and location. In order to establish whether a country actually had access to the international financial markets, the first set of empirical analyses focuses on periods in which states were in default, as listed in Reinhart and Rogoff(2009). These authors define sovereign default as the failure of a government to meet a principal or interest payment on the due date (or within the specified grace period). Among the main causes of default, there is war, which reinforces the main insight of the theoretical discussion: financing war with loans does not guarantee an improvement in the fiscal capacity of the state with respect to prewar levels. Reinhart and Rogoff code periods of external default since 1800 for 68 countries, as defined by their current territory. I work with 63 out the 68 countries in their sample, all for which full data is available. The sample includes countries of the five continents and accounts for approximately 90% of world GDP by 1913. The median duration of default episodes in the period under consideration is six years (Reinhart and Rogoff, 2009, p.81) Critically, while in default, countries are excluded from the international lending market (Tomz, 2007), which I expect to strengthen the ruler’s incentives to invest in the tax capacity of the state. Given the levels of income tax today, and wars and default episodes in the long nineteenth- century, I exploit cross-sectional variation in order to estimate:18

PITi,1995−2005 = α + β1(#years at war between 1816-1913 | no access to external loans)

+β2(#years at war between 1816-1913 | access to external loans)

+Xiδ + γ + ρ + i (1) where X denotes a vector of controls, γ and ρ are batteries of colonial origins and region fixed effects, and is the error term. First, in the absence of external loans, I expect war-making to strengthen the ruler’s incentives to invest in fiscal capacity, contributing to long-term fiscal capacity, β1 > 0. Second, in light of the commitment problems of war debt repayment, I expect a null (if not negative effect) of war-making when countries wage war

18The analysis is cross-sectional. The data structure is not time-series-cross-sectional, as there is no time-varying tax data for the nineteenth century for most of the countries in the sample.

17 while having access to external credit, β2 6 0. A negative sign for β2 would suggest that the ﬁscal disequilibrium associated with excess borrowing combined with the exchange of state monopolies for default settlements can fully reverse the eﬀect of war on state-making.

Also importantly, the expectation β2 6 0 works against the Ricardian Equivalence, which implicitly assumes no commitment problem of debt repayment. If borrowing and taxes are equivalent in the long-run, we should expect β1 ≈ β2 > 0, everything else constant. As part of Expression1, all models below include: first, a battery of Region fixed effects that account for continent-specific characteristics in the frequency of war, access to credit, and statehood timing. Second, a battery of Colonial Origins indicators, as I expect access to external credit of colonies, their opportunities to go to war, as well as the tax structure that they build up to be influenced by the metropolis.19 For reference, Appendix Table A-13 investigates whether British colonies (and military allies) are comparable to the remaining countries, given their unequal relationship with the financial capital of the world. All models include a vector of potential confounders affecting the level of PIT today as well as war participation, credit access, or both, back in the nineteenth century: First, a measure of initial wealth, as wealthier countries are more likely to go to war and have stronger fiscal capacity in the first place (Tilly, 1990; Gennaioli and Voth, 2015). In the absence of systematic GDP data for the early nineteenth century, I follow Acemoglu, Johnson and Robinson(2005) and Dincecco and Prado(2012) and include a measure of Population Density as of 1820, which is argued to be the best proxy of a country’s wealth in the early industrial revolution (Tilly, 1990, p.17). Second, I also include two geographic characteristics that could affect both sides of the equation. The first one, Sea Access, is the percentage of the land surface area of each country that is within 100km of the nearest ice-free coast (Nunn and Puga, 2012). If expect sea access to correlate with trade activity (thus access to international lending) and monetization, a precondition for income taxation (Tilly, 1990). By the same

19Accominotti, Flandreau and Rezzik(2011) show that UK colonies accessed the international lending markets in the same terms as the metropolis, as it was perceived that London would assume the default risk. See Obstfeld and Taylor(2003) for an opposite view. Following Persson and Tabellini(2003) and Dincecco and Prado(2012), I include three colonial origins dummies: British, Iberian, and all other colonies.

18 token, I expect territories with sea access to be military valuable, thus increasing their likelihood of experiencing war. The second geographic control is the percentage of territory that is Desert (Nunn and Puga, 2012). I expect deserts to inhibit industrial growth, and preempt monetization. But desert territory might also work as a natural barrier to foreign invasion, thus reducing the frequency of war.20 Lastly, I control for a close substitute to tax revenue that could also shape the incentives to go to war (or being attacked): being an Oil Producer (Wimmer and Min, 2009). Arguably, this variable gains relevance for the later years of the period under consideration.

4 Endogenous Access to External Credit

To establish a benchmark, column 1 in Table1 tests for the unconditional version of the bellicist hypothesis: that is, does long-term fiscal capacity increase in the number of years at war in the long-nineteenth century, holding everything else constant? Or more generally, does war make states? Results are mixed (consistent with what others have found): the coefficient for # of Years at War between 1816-1913 is positive but not significant.

Table1

Column 1 should be compared to column 2 and remaining specifications, in which I distinguish the effect of war fought while in default, β1, from war fought while having access to international credit markets, β2. Both point estimates are positive, but, consistent with the political economy of war financing, only the former is significantly different from zero. A one-standard deviation increase in the number of years at war while in default increases income tax to GDP in 0.41 points. This is a 15% increase with respect to the PIT’s sample mean.

20Other geographic controls are endogenous and should not be included in the analysis: e.g. size. Others correlate with military capacity but not with the dependent variable: e.g. terrain ruggedness.

19 On the contrary, column 2 suggests that wars that are fought when countries have access to international markets do not exert any persistent effect on fiscal capacity. This is consistent with the commitment problem above indicated. Nothing guarantees that once war is over, countries service debt within the pre-established timeframe and conditions. Some countries honor their debt (by enhancing its fiscal capacity as to amass the required funds), others do not. Column 3 controls for the baseline propensity to default. To this end, I include the # Years in Default between 1816-1913 of each observation. The two coefficients of interest remain virtually identical. The remaining of Table1 considers potential confounders, while making sure not to control for endogenous variables or bad controls: for instance, based on the political economy of war financing, Current Levels of Democracy, which strongly correlate with PIT, might result from tax-financed war. Similarly, Current Levels of GDP per Capita, which also correlates with PIT, are argued to result from war making in the past (Dincecco and Prado, 2012; Gennaioli and Voth, 2015). Keeping that in mind, next I consider only covariates that potentially condition war-making in the past and influence the tax structure today.21 The first potential confounder, being a Great Power in the nineteenth century, is examined in column 4. This control accounts for the idiosyncratic paths of state- and war-making in the United Kingdom, France, Germany, Italy, Austria-Hungary, and Russia.22 These countries were major military and economic powers in the nineteenth century, and could be driving the results. The coefficient of this indicator variable is positive, as one would expect, ˆ ˆ but is not statistically significant. Importantly, β1 and β2 remain the same as in column 2 and 3. Figure2 offers a visual intuition of these estimates. The left-panel shows the partial ˆ correlation between levels of PIT today and # Years at War while in Default, or β1: this

21For reference, Appendix Table A-15 reports models including bad controls. Results hold. 22In establishing which country is a Great Power in this period, I follow Flandreau and Flores(2012). Austria and Hungary are treated as two independent countries. Refer to Appendix A for details.

20 relationship is positive and statistically significant. Four cases appear as potential outliers in Figure2: Peru, Argentina, Mexico and Prussia. Appendix Table A-7 reruns the model ˆ without these observations. Once outliers are dropped, β1’s magnitudes triples. That is, ˆ ˆ Table1 offers the lower bound of β1. The right-panel in Figure2 plots β2. Consistent with the commitment problems associated with debt-financed war, this estimate is positive but does not reach standard levels of statistical significance.23

Figure2 here

War causes destruction, but damages vary greatly depending on the location of military engagement. The tax base can be badly hurt when military conflict takes place within national borders, thus inhibiting investment in fiscal capacity. Additionally, countries might invade others for extractive purposes. The location of war is thus likely to be a confounding variable. To address this logic, column 5 in Table1 controls for the location of conflict. In particular, War Location is the sum of the years at war fought abroad minus the years at war fought at home for the entire 1816-1913 period. This variable is positive when a country fight more wars abroad than at home; negative, when military disputes at home are more frequent than abroad; or zero, when countries never go to war.24 The coefficient for this variable is positive, as one would expect, but not statistically significant. Importantly, the ˆ ˆ coefficients β1 and β2 do not change despite the potential post-treatment bias induced by this control.25 All wars are not created equal. Bloodier wars might overcome resistance to taxation, while maximizing the ruler’s incentives to invest in fiscal capacity. To address this possibility, column 6 includes a control for the intensity of warfare, measured by the number of battle deaths within the period, or Casualties in 1816-1913 (Dincecco and Prado, 2012). This variable is not statistically significant, even though its presence, if only marginally, pushes

23 ˆ Appendix Table A-7 shows that, once we drop the four outliers, β2 becomes statistically signiﬁcant at ˆ 90%. Yet, its magnitude is one-order of magnitude smaller than β1. 24One case only fought the same number of military campaign abroad than at home. 25Results are virtually identical if the total number of wars fought abroad or at home are used separately.

21 ˆ down the magnitude of β1, the effect of war fought while in default. Next, I control for Ethnic Fractionalization and Civil Wars. The former might be an im- pediment to invest in fiscal capacity (Besley and Persson, 2011), while ethnically fragmented countries might also be perceived as more vulnerable to foreign military intervention. Eth- nic fractionalization is measured as of the 2000s, and is potentially endogenous to war. A long history of Civil Wars is a strong predictor of negative patterns of development (Besley and Reynal-Querol, 2014), while lack of political stability might be penalized by the credit market. Controlling for civil war, however, is far from ideal, as sometimes they result from inter-state wars (although the opposite might happen too). At the risk of incurring in post- treatment bias, columns 7 and 8 control for the level of ethnic fractionalization today and the number of years at civil wars between 1813-1916.26 The marginal effect of both controls are positive, but they are not statistically significant. Importantly, the inclusion of these variables does not change the effect of fighting wars while being in default.27 Lastly, Scheve and Stasavage(2010) show that progressive taxation, such as PIT, accelerated dramatically among countries that participated in World War I (WWI ). Including this covariate in the empirical model might lead to a post-treatment bias if countries that frequently went to war in the nineteenth century and developed higher fiscal capacity by 1914 selected into WWI. Still, one might be tempted to include a WWI indicator to check whether the coefficients of interest survive this control. As reported in column 9, they do: fighting war while being on default is positively related to higher fiscal capacity today, whereas fighting war while having access to external credit, does not. Participating in WWI is positive but does not reach conventional levels of statistical significance.28

26To minimize bias, civil wars that take place simultaneously to inter-state wars are not considered. These are a minority, nevertheless. 27Appendix Table A-9 includes a control for the Federal structure of the state as of today, which might reflect cumulated ethnic fractionalization. 28The WWI indicator takes value 1 for all countries that actively participated in WWI (i.e. suffered military casualties). This coefficient achieves conventional levels of significance in later models, when the sample size increases.

22 4.1 Alternative Endogenous Sources of War Financing

There are three additional, arguably less frequent, ways to finance war: domestic borrowing, monetary expansion (also known as printing money), and financial repression. Appendix H addresses the effect of the domestic borrowing and monetary expansion empirically. Re- sults strengthen Table1’s: war-making is even more consequential when it is fought in periods of external and domestic default and no monetary expansion. Likewise, access to domestic credit and instances of monetary expansion do not cancel the effect of fighting war while in default. Other policy, such as financial repression, office selling or confiscation introduce a downward bias, if any, on the main coefficient of interest, β1. That is, if rulers prioritize fiscal repression when they lack access to external finance, we should not expect a positive coefficient for the # Years at War while in Default, precisely because fiscal repression is implemented as to avoid fiscal capacity building.

5 Exogenous Access to Foreign Credit

Going to war and being in default are not randomly assigned. There might be unobserved elements that make states more (less) likely to go to war and more (less) likely to be in default that affect their fiscal capacity today. So far, the first source of bias is addressed by the initial wealth proxy (i.e. population density as of 1820), which accounts for who goes to war more often in the first place: richer countries. The Great Power indicator and Region fixed effect contribute to minimize bias. The second source of endogeneity, namely, who is in default in the nineteenth century, is addressed next by exploiting shocks in the international lending market since 1816. As it will become clear, crashes in the international financial market dry capital flows at a global scale, a phenomenon known as sudden-stops of credit (Calvo, 1988). Key for the identification strategy, sudden-stops ban countries from external borrowing irrespective of their (un)observed characteristics. In other words,

“Banking crises in global ﬁnancial centers (and the credit crunches that accompany

23 them) produce a ’sudden-stops’ of lending to countries at the periphery [...]. Essentially,

capital ﬂows from the ’north’ dry up in a manner unrelated to the underlying economic

fundamentals in emerging markets.” (Reinhart and Rogoﬀ, 2009, p.74).

In this section, I use sudden-stops as a form of exogenous variation of access to external credit, which structures the incentives to invest in ﬁscal capacity for countries at war.

5.1 International Financial Crashes in the Nineteenth Century

Most of the international credit in the long nineteenth century was channeled through the London Stock Exchange (LSE). London took over Amsterdam as the financial center of Europe by the turn of the eighteenth century (Neal, 1990). The LSE’s financial leadership consolidated throughout the nineteenth century, when it became the world’s leading capital exporter, far exceeding the combined capital exports of its nearest competitors, France and Germany (Feis, 1930). Table2 reports the best approximation of the market shares in lending throughout the nineteenth century. These data indicate that for the long nineteenth century, the British were “the bankers of the world” (Obstfeld and Taylor, 2004).29 At its peak, the British share of total global foreign investment was almost 80%. This contrasts with the US share of global assets in 2000, 25%, and even with the US maximum share of 50% circa 1960 (ibid.). All in all, in the nineteenth century, the LSE played an unmatched role in financing the world.

Table2 here

Conveniently enough, the LSE was not immune to crises. Table3 enumerates the onset of all banking panics and stock crashes experienced by Britain in the long-nineteenth century, as listed in Reinhart and Rogoﬀ(2009). Given Britain’s central position in the international lending market, crashes in London rapidly spread to Paris, Frankfurt and New York. Conta- gion took diﬀerent routes, including arbitrage in commodities and securities, and movement

29This conclusion is shared by many others: e.g. Reinhart and Rogoﬀ(2009), and Tomz(2007).

24 of money in various forms (specie, bank deposits, bill of exchange), cooperation among monetary authorities, and pure psychology (Kindleberger and Aliber, 2005, ch.7). One way or another, ﬁnancial crashes in London dried international lending at a global scale (Bordo, 2006).

Table3 here

Importantly, the causes of the financial collapses in the nineteenth century can be found in the British economy, not abroad. This is certainly the case for the major crises of 1825, 1847, 1857, and 1866, but less true for the 1890 panic, when a big financial imbalance in Argentina put a halt to British lending (Kindleberger and Aliber, 2005).30 More importantly, British panics did not respond to defaults by borrowers, which would cast doubts about the exogeneity of these shocks. Most of the countries that defaulted in the nineteenth century were in the periphery. Although the defaulted quantities were significant relative to their home economies, from a global prospective, they were a “sideshow” (Eichengreen, 1991, p.151). All things considered, we can safely treat the periods of sudden-stops as exogenous to every country except for Great Britain and, arguably, 1890 Argentina. For illustration purposes, Figure3 shows the evolution of British capital exports since 1865 (earlier data do not exist), while indicating the years of banking panics and stock crises as coded by Reinhart and Rogoff(2009). Figure3 reflects the boom-and-boost cycles preceding and following a banking crisis, as exemplified by the financial crisis of 1873 and 1890. Prior to each boost, lending was ferocious. Once the debt bubble exploits, international capital flows temporarily dry across the board. Precisely, it is within periods of sudden-stops that I expect rulers to have stronger incentives to finance military campaigns by means other than external borrowing: namely, taxes. But, unlike the case of default episodes (previous section), the incentive system is now structured by exogenous factors.31

30For the domestic origins of banking panics, see Neal(1998) for 1825 crisis, Dornbusch and Frenkel (1982) for 1847 crisis, and Mahate(1994) for 1866 crisis. See Marichal(1989) for an exonerative assessment of Argentina in the 1890 crisis. 31Based on Figure3, banking crises might be more damaging than stock market crises. Appendix Table A-10 includes an additional test in which only banking crises are considered.

25 Figure3 here

Crucial for the quasi-experimental setting, financial bursts are predictable only ex post, as Reinhart and Rogoff(2009) convincingly argue. But suppose that some rulers had inside information and banked external loans in anticipation to sudden-stops. If the theory advanced in Section 2 is correct and the incentives to invest in fiscal capacity are a function of credit access, anticipation implies that rulers will not invest in fiscal capacity even if the financial markets are down. That is, anticipation creates an attenuation bias on the coefficient of interest, β1 in Expression1. As Figure3 also indicates, markets do not rebound immediately after a financial crisis. Although the final resolution of financial crises varies (Chwieroth and Walter, 2013), before WWI, sudden-stops lasted four years on average (Catao, 2006). Accordingly, I establish 4-year windows (including the year at which the crisis begins) within which I assume that countries do not have access to external credit.32 For each of these lapses of time, I count the number of wars that a country was involved in. To fully test the theoretical expectation, I also compute the number of years at war that a country fights while credit flows in the international market. Importantly, the sample size grows with respect to the previous section, as it is not constrained by Reinhart and Rogoff(2009)’s default data coverage.

Table4 here

To evaluate the validity of the exogenous credit shock, Table4 compares the frequency and duration of war in periods in which access to credit is endogenous vs. periods in which it is exogenous. The statistics for the endogenous access to credit suggests that countries strategically choose when they go to war: specifically, 87% of wars coincide with periods in which countries are not in default. Likewise, wars are shorter when countries are in default: 2.0 years compared to 2.3 years when countries have access to external finance. These figures change when we evaluate scenarios of exogenous access to credit: 52% of wars now coincide

32Refer to Appendix Table A-10 for longer windows.

26 with periods in which the international lending market is down, whereas the duration of war is also balanced: 1.9 years in periods of sudden-stops compare to 2.0 years when credit flows.33 Altogether, these numbers confirm that sudden-stops are unanticipated and that they do not condition the decision to go to war, nor to retreat from it. If we judge war by its frequency and duration, Table4 suggests that we are comparing animals of the same kind when tackling with wars waged in periods in which international lending flows with wars waged in episodes of sudden-stop of credit.

5.2 The Eﬀect of Sudden-Stops

The analysis in Table5 uses the periods in which the LSE does not issue credit as a means to identify scenarios in which countries at war have stronger incentives to enhance its ﬁscal foundations. The model speciﬁcation follows Expression1. This time, however, Great Britain —the world’s banker—is dropped so as to maximize exogeneity.34

Table5 here

Column 1 establishes the benchmark model, in which the simplest, unconditional bellicist hypothesis is tested against the data. Results are now mildly positive: at a 90% confidence interval, PIT today increases in the number of years at war in the long nineteenth century, keeping everything constant. Column 1 should be compared to column 2, in which I distinguish wars fought while having exogenous access to external credit from wars fought in the midst of a sudden-stop. Column 2 suggest that the effect on current fiscal capacity critically depends on how war is financed. On the one hand, the effect is positive when it leaves the ruler out of options and pushes her to invest in fiscal capacity: a one standard deviation increase in the number of years fought while international lending stops, increases 1.3 points the average PIT today,

33For a visual intuition of the same result, refer to Appendix Figure A-6. 34Appendix Table A-13 shows results upon dropping British colonies, British military allies, even every war involving British participation.

27 equivalent to a 43.3% increase with respect to the sample mean. On the other hand, the effect of war-making when rulers have access to external lending is virtually the reverse: a one-standard deviation increase in the number of years at war when credit is available is associated with a decrease of 0.9 points in PIT as percentage of GDP. This result suggests that debt-financed war might create fiscal imbalances that are too hard to fix. These should be strongest among those states that handle over state monopoly revenues to lenders in order to regain market access after defaulting. ˆ ˆ The opposite signs of β1 and β2 suggest that the effect of war estimated in column 1 is the average of two radically different worlds. Indeed, this result advances our understanding of the conditions under which wars make states. The remaining columns in this and subsequent tables establish how robust this result is to endogeneity and measurement issues. Columns 3 to 8 in Table5 include additional controls, one at a time: the # of Years in Default, being a Great Power, War Location, War Casualties, Ethnic Fractionalization, and Civil wars, in this order. Additionally, now I control for Average War Duration, as it is no longer collinear with having access to external credit, as shown in Table4. Across different ˆ ˆ specifications, the two coefficients of interest, β1 and β2, remain virtually unchanged with respect to column 2.35 ˆ ˆ Figure4 offers an visual intuition of β1 and β2. The left panel shows the partial correlation ˆ between PIT as a Percentage of GDP and # of Years at While Credit Stops, or β1. There are arguably three outliers in this relationship: Russia, Georgia, and France. Appendix Table ˆ A-7 shows that their exclusion does not affect the size and precision of β1. The left panel ˆ plots β2. Russia and Georgia are again outliers. Appendix Table A-7 shows that they are ˆ indeed influential, as once we drop them, β2 is not statistically significant any more. This ˆ null result for β2 is consistent with some of the robustness tests that follow.

Figure4 here

35Notice that we do not have to adjust for the total number of years in which capital ﬂows or dries, as this is a common shock to every country and is picked up by the intercept.

28 Lastly, in column 9, I control for participation in WWI, despite the problems of endogeneity discussed above. Results hold. More interestingly, the coefficient for WWI sets a meaningful benchmark to compare conventional war-making in the long nineteenth century vis-à-vistotal war: the latter’s marginal effect on fiscal capacity is 1.2 points, whereas a one-standard deviation increase in the number of years at war while not having access to external credit increases PIT today by 1.3 points. Results are virtually equivalent, meaning that sufficient years of conventional war, as long as they are (at least partially) financed with domestic taxes, might exert lasting effects close to total war participation. To sum up, Table5 suggests that war does not necessarily make states. It all depends on the incentives that rulers have to invest in the fiscal capacity of the state, which, I argue, should be stronger when they lack access to external loans, and weaker when they have access to cheap credit in the international lending markets. Before discussing the implications of this result, Tables6-8 address additional measurement and endogeneity issues.

5.3 Sovereign States, Secessionist War, and Alternative Depen-

dent Variable

So far, wars are attributed to the corresponding 2001 nation state irrespective of whether that territory had achieved statehood by 1913. It could be argued that wars fought in territories that were not recognized by the international system exert a different (or null) effect on the fiscal capacity of the state. For instance, colonies might not invest in their military campaigns as much as the metropolis. To address this possibility, columns 1 to 3 in Table6 rerun the analysis considering only countries that were sovereign by war time. 36 These models also control for Great Power status and War Location. Results, despite the reduction of the sample size, are similar to those reported in Table5. Sovereign states that waged war while international lending flowed are not associated with high fiscal capacity today. To the contrary, sovereign states that waged war in the midst of a sudden-stop have,

36Sovereign status is drawn from Boix, Miller and Rosato(2013).

29 on expectation, higher tax capacity today.

Table6 here

The second battery of sensitivity analyses moves in the opposite direction by considering secessionist war along inter-state wars. Secessionist wars is fought by entities that are, by definition, not sovereign. Secessionist war might or might not seek the formation of an independent modern nation-state. Next, I consider secession war that pursued this goal, as established by Wimmer and Min(2009). Current developing countries fought most of the secessionist wars in the nineteenth century. Thus, considering these wars makes the sample more representative of the universe of states at war in the period under consideration. Importantly, as reported in the bottom of Table4, once secessionist wars are considered, the frequency and duration of war in periods of sudden-stops and credit-flow are fully balanced, thus maximizing comparability. Columns 4 to 6 in Table6 suggest that waging war, either inter-state or secessionist, while international credit markets are operative, is not associated with higher fiscal capacity. On the contrary, countries that fought wars in the midst of a sudden-stop, regardless of international statehood recognition, are associated with higher levels of fiscal capacity today. The third battery of sensitivity tests addresses the choice of the outcome variable: personal income tax as a share of GDP, arguably, captures both capacity and willingness to tax. To address this potential issue, I use an alternative proxy of fiscal capacity, one that emphasizes capacity over willingness: the size of the tax administration (US AID, 2012). First, professionalized bureaucracies are necessary to assess and collect taxes, as well as to resist the natural aversion to having one’s sources of income examined and monitored (Daunton, 2001; Mares and Queralt, 2015). Second, modern tax bureaucracies were created for and by war (Acemoglu, Ticchi and Vindigni, 2011; Brewer, 1988) and tend to be fairly persistent (Fischer and Lundgreen, 1975; Schumpeter, 1918). Third, the tax administration, filled by public servants and subject to stricter controls, is relatively sheltered from spurious, fleet- ing interests of passing incumbents and, as a result, should change only slowly over time.

30 Altogether, the size of the tax apparatus captures the underlying capacity to monitor and assess private income, that is, the fiscal capacity of the state. Consistently, the size of the tax administration is a strong predictor of total tax revenue, as shown in Appending Figure A-1. In columns 7 to 9, the size of the tax administration, measured by the tax staff per thousand capita circa 2005, is regressed on the same covariates used to model income taxation.37 Results mimic previous ones: waging war while not having access to external credit (exogenized by instances of sudden-stops) is associated with a more staffed tax administration today. War waged while credit flows is not, if any, results suggest that the opposite might happen. Additionally, Appendix Table A-12 reports models of Value-Added Tax (VAT) revenue as a percentage of GDP. VAT is arguably easier to implement than the income tax (Bird and Gendron, 2007), and it may not capture accumulated investment in fiscal capacity as precisely as income taxes do. Still results for VAT are equivalent to those presented so far. Altogether, Table6 shows that results in Table5 are robust to sample change, and alternative war definitions and outcome variables.

6 Addressing War Endogeneity

The decision to go to war may be endogenous. First, countries that go to war might have greater administrative capacity to begin with (i.e. omitted variable bias). Second, the type of countries that decide to go to war when there is no access to credit may be diﬀerent in ways that are relevant to future tax collection capacity from those which choose to wait until loans are available (i.e. selection bias). I address both issues stepwise.

37Despite the smaller N, the sample still includes countries of the ﬁve continents.

31 6.1 Initial State Capacity

First, countries that are frequently at war may have greater capacity to conscript and tax. These capacities may already be captured by the Great Power indicator and the proxy of initial wealth: Population Density as of 1820. After all, we know that the income level is the strongest predictor of war participation (Gennaioli and Voth, 2015; Tilly, 1990). Never- theless, next I minimize this potential source of endogeneity by considering two covariates associated with initial state capacity: an index of state antiquity put together by Putterman (2007), and a new one on census capacity. Following Tilly’s logic, State Antiquity should correlate with accumulated military capacity: that is, older states exist because they won war in the past. Second, the capacity to successfully conduct a modern census should correlate (if not facilitate) preparation for war, if only because modern censuses tend to follow earlier enumarations that assess taxable wealth and the conscription base. To this end, I have collected information about the ﬁrst census ever conducted in every country in the sample. To control for initial administrative capacity, I create the indicator variable Modern Census by 1820, which equals 1 if a country x has conducted a modern census by 1820.38

Table7 here

Results are reported in columns 1 and 2 of Table7. The two new covariates hold a positive coefficient, as expected, but are not statistically significant.39 Importantly, once we ˆ ˆ control for both proxies of initial state capacity, the coefficients of interest, that is, β1 and β2, remain positive and negative, respectively, and statistically significant. That is, independent of observable initial capacity to prepare for war and raise taxes, only countries that fought war when the international lending market is down are associated with higher levels of fiscal capacity today.

38See Appendix for further details. 39They reach conventional level of statistical signiﬁcance when I exclude other covariates, with which they correlate.

32 6.2 Ongoing Wars

Countries that go to war despite the credit dry may be different from countries that wait for markets to lend again. Table4 (and Appendix Figure A-6) suggests that there is no much strategic timing of war making once we exogenize credit access: the frequency and duration of war inside and outside sudden-stop periods are virtually balanced. Still, we might want to address selection bias by considering only wars that are initiated while the market is still lending and, eventually, dries as a result of a financial crisis. That is, these are wars that are initiated without the expectation of a sudden-stop. Columns 3 and 4 in Table7 report the results of this test. The estimate for # of Years at War while Credit Stops decreases by 30%, approximately, with respect to Table 5, suggesting that the latter may be somewhat upward biased. Still, the new estimate is positive and statistically significant in both specifications. On the other hand, the effect of fighting wars while having access to the international market is no longer negative, but zero, which is still inconsistent with the unconditional interpretation of Tilly’s dictum. Selection bias is further addressed in a reduced-form framework in Appendix Table A-14, where war participation by country x in times of sudden-stops and credit-flow are instrumented by war-making by the immediately adjacent countries.

6.3 Involuntary War, COW and War Outcome

A potentially more interesting way to address selection into war is to analyze the effect of war-making by states that did not choose to go to war, but were dragged into it. In particular, one could argue that countries that initiate war might be different from those that are attacked in ways that shape current fiscal capacity. Based on this logic, this analysis focuses on the differential effect of war-making and credit access for countries that are attacked only, (i.e. the non-initiators). The identification assumption here is that initiators do not strike first in anticipation of a likely attack.

33 Table8 here

To conduct this test, I rely on Correlates of War (COW), which identifies the initiator of each military conflict. The COW dataset includes fewer inter-state wars than Wimmer and Min(2009), as it follows a stricter criterium about what a state is in the nineteenth century.40 Accordingly, the sample of inter-state wars is now made of 37 conflicts, and 174 country-year-wars in total. 78 of them were fought when credit flowed, and 96 when credit suddenly stopped. The average duration was 1.57 (sd = 1.04) and 1.76 (sd = 1.22) years, respectively. On the bright side, COW indicates which side eventually wins war. This is substantively compelling, as military outcomes potentially affect the incentives to invest in fiscal capacity: e.g. winners might extract from losers. To this end, Net Victory indicates the number of wars won between 1816 and 1913 by country x net of wars lost within the same period. Countries that did not fight any war have a value of 0.41 Table8 begins by replicating the differential effect of war and credit access for the entire COW sample, including initiator and non-initiators. The effects reported in columns 1 and 2 are slightly lower than those estimated in previous tables. Based on column 1, a one- standard deviation increase in the number of wars fought while not having access to credit increases PIT today by 0.8 points (equivalent to a 27% increase with respect to the sample mean). Most probably, the decrease in the predicted effect results from the sample selection of COW, which basically includes rich countries, for which additional years at war should exert a relatively smaller effect. Importantly, columns 1 and 2 imply that results are robust to sample change. In columns 3 and 4, I estimate the effect of war and credit access for countries that did not choose to go to war, but were pushed into it: that is, the non-initiators. Results are similar to the preceding ones, only bigger. Countries that were dragged into war in the midst

40Refer to Appendix Section A for further details. 41Three countries won the same number of wars that they lost: Bulgaria, Spain, and Turkey. All other zeros correspond to countries that fought no war within the period.

34 of a sudden-stop of credit present higher levels of fiscal capacity today. The effect vanishes when countries are allowed to borrow external loans to wage war. Importantly, results are robust to military conflict outcomes: winning or losing wars do not significantly modify the differential effect of war-making and credit access on long-term fiscal capacity.

7 Immediate Eﬀects of War Making and Credit Access

It has been argued that modern war exerts long-term effects on fiscal capacity, particularly when it pushes rulers to conduct institutional reforms, such as enhancing the tax bureaucracy, adopting new taxes, or building up a central bank. Consistent with this logic, we should observe effets in the short-term too, as fiscal capacity is a cumulative process. In the absence of tax data for current developing economies in the early twentieth century, I focus on two measures of state capacity that should correlate with tax capacity: the ability to conduct a modern census, and the length of open rail lines, both dated as of 1913. The first measure is clearly a requirement to adopt modern forms of direct taxation, as it establishes the potential tax base. The second measure, rail lines length, captures Mann’s (1984) notion of “infrastructural power” of the state, as it facilitates the state’s presence throughout the territory. Importantly, Dincecco, Fenske and Onorato(2016) and Queralt(2015) show that the railroad network correlates with fiscal capacity.

Table9 here

Table9 reports the eﬀects of war-making on short-term state capacity as a function of exogeneous credit access. In columns 1 and 2, a probit model regresses having a modern census by the end of the period of reference on war-making and external credit availability, plus controls. Results suggest that waging war during the long-nineteenth century in the absence of external loans increases the probability of having a modern census by 1913. On the contrary, being at war while having access to external credit does not translate into higher probability of having conducted a modern census by the end of the period of reference.

35 Columns 3 and 4 run an OLS model in which the date of adoption of the first modern census is regressed on key regressors. In this model, high values of the dependent variable imply delays in census adoption. These columns show that fighting wars in times of sudden- stop shortens delay of adoption (or, if preferred, accelerates it). Fighting wars having access to credit does not. If any, it increases delay, consistent with commitment problems of debt repayment. Results in column 4 to 6, in which I model the length of rail lines by 1913, mimic previous results. If any, columns 5 and 6 emphasize the potential perverse effects of fighting war while having access to external credit, this time on short-term state capacity. Overall, results in Table9 suggest that the effect of war-making on state capacity is cumulative. War-making in the absence of external credit in the long-nineteenth century leads to higher fiscal capacity on the verge of World War I, and this effect persists until today.

8 Discussion

Contrary to the unconditional characterization of the bellicist hypothesis, that is, more war, more state, I argue —along Tilly’s original work, that the effect of war on state-building ultimately depends on how warfare is financed. Specifically, I claim that financing war with taxes makes states for certain, whereas financing wars with loans might not, as there are commitment problems with debt repayment. The research hypothesis has been tested with cross-national historical data, including developed as well as developing economies to avoid sample selection issues, and evaluated in both the short- and long-run. To address the endogeneity of war financing, the empirical analysis exploits sudden-stops of credits, and shows that war fought while not having access to external credit is positively associated with higher fiscal capacity in the long- (and short) run. By contrast, financing wars with external capital does not translate into higher tax capacity in the long- (or short) run, and it can result detrimental, possibly due to some perverse conditions of default settlements: e.g.

36 handing over state monopolies and revenues to lenders. To address the endogeneity of war participation —unlikely conditional on sudden-stops—, I control for initial state capacity, focused on ongoing wars (that is, wars that are initiated while the market is still lending and, eventually, dries as a result of a financial crisis), and countries that dragged into war (i.e. the non-initiators). Results hold across specifications, as well as alternative dependent variables: income taxes, value-added tax (Appendix), tax administration, census capacity, and the railroad network, a measure of the infrastructural power of the state. The powers conferred by fiscal centralization to the national ruler, combined with the vested interest in self-perpetuating by the same bureaucracies that were once created to finance war, are advanced as the mechanisms of transmission. This is not the first work that qualifies the bellicist hypothesis. In evaluating its applica- bility to Latin America, Kurtz(2013) claims that this continent did not have enough societal cohesiveness to invest in fiscal capacity, nor sufficiently robust central governments to make domestic extraction profitable. Likewise, Karaman and Pamuk(2013) argue that the effect of warfare on state capacity in pre-modern Europe is conditional on the complementarities between urbanization and regime type. Interesting as they are, these works do not account for the radically different international context in which countries in the periphery are created as compared to those faced by European nations. The vast majority of states in the periphery are founded only after 1815, coinciding with the globalization of financial markets, which results from the income growth in the wake of the industrial revolution, as well as Britain’s capacity to spin off excess savings (Neal and Weidenmier, 2003; Reinhart and Rogoff, 2009). The volume of crossborder loans in the nineteenth century is unprecedented: scaled by the size of the world economy, international capital flows between 1880 and 1914 are three times as large as in the 1980s (Eichengreen, 1991, p.150). Unlike European states, from their very inception the new states in the periphery had access to unprecedented levels of inexpensive external loans, despite their weak institution-

37 alization, frequent government turnover, and lack of reputation in the international markets (Tomz, 2007).42 The “lending frenzy” (Taylor, 2006) lasted only temporarily. By the end of the nineteenth century, as a results of (inevitable) defaults in the periphery, markets did update the premia for proven lemons (Lindert and Morton, 1989; Tomz, 2007). By then, however, many wars had already been fought. Cheap external credit, I argue, unravels the relationship between war-making and state- making for three reasons: first, it allows war to be financed without raising taxes domestically, thus inhibiting political reform often associated with a well-articulated tax system: namely, representative political institutions with budgeting veto powers (Bates and Lien, 1985). Second, readily available, inexpensive external credit preempts the development of domestic credit markets, thus the formation of a corpus of domestic creditors with whom to strike bargains conducive to political reform and responsible fiscal policy (North and Wein- gast, 1989; Stasavage, 2011). Third, the globalization of lending markets exacerbates the commitment problems associated with debt servicing. It facilitates the possibility of refinancing debt with more debt instead of investing in fiscal capacity, thus heightening debt burden rather than solving it. Counterintuitively enough, countries in the periphery may have benefited from less efficient international lending markets, as that would have strength- ened the incentives to raise taxes to finance the means of war, stimulate domestic borrowing, and pass political reforms associated with responsible fiscal policy —namely, what European counterparts were pushed to do, only centuries before, when credit markets were oligopolistic and expensive (Homer and Sylla, 2005, ch.9). The perverse effect of inexpensive external credit resonates with the original Tillyian hypothesis by emphasizing the conditional effect of war on credit access: in Europe, frequent war-making and the absence of cheap external credit propelled domestic lending, and eventually political reform that addressed commitment problems of debt repayment.43 Together,

42The lending frenzy is sustained on strong information asymmetries, speculative operations, and blatant fraud (Taylor, 2006). Appendix N provides further evidence of this phenomenon. 43Domestic markets were created twofold: by lending from merchants in commercial cities (e.g. Henry IV, 1598-1610, borrowed from Paris, and marginalized increasingly-expensive Italian lenders), or by coercive

38 frequent warfare and domestic lending allowed territorial states to pursue the “coercive- capital intensive”, fiscal-military strategy that ended in the modern tax state (Tilly, 1990).44 Cheap external credit —which countries in the periphery had since their inception—breaks the causal chain of the original Tillyan hypothesis. Readily available external credit unravels the necessity to finance war with domestic debt or taxes, and ultimately, preempts the capacity to capitalize war efforts. Others have warned about the perverse effects of cheap credit: Centeno(2002), Shea (2013) and Thies(2007) claim that this form of artificial wealth distorts the incentives to tax domestically even in times of war. Building on this work, I advance the political economy of war financing, articulate the implications for long-term fiscal capacity, and test them against historical data. Interestingly, the perverse incentives associated with cheap credit are equivalent to those derived from other forms of non-tax revenue: foreign aid (de Mesquita and Smith, 2013; Moss, Pettersson Gelander and van de Walle, 2006), oil and gas revenue (Morrison, 2009; Ross, 2001), and ores from colonies (Drelichman and Voth, 2011). Altogether, the paper offers a better understanding of the conditions under which war exerts positive and lasting effects on state building. I leave for future research investigating the extent to which the mix of internal and external credit advances our understanding of the heterogeneous paths to state building in Western Europe since the fifteenth century. Conveniently enough, that analysis should benefit from richer data sources.

annexation of capital-intensive cities (Stasavage, 2011). 44To the contrary, states that kept relying on external loans to ﬁnance wars found it much harder to capitalize the eﬀect of war on state-making: e.g. Spain under Phillip II (Drelichman and Voth, 2011).

39 References

Accominotti, Olivier, Marc Flandreau and Riad Rezzik. 2011. “The spread of empire: Clio and the measurement of colonial borrowing costs.” The Economic History Review 64(2):385–407.

Acemoglu, Daron, Davide Ticchi and Andrea Vindigni. 2011. “Emergence And Persistence Of Ineﬃcient States.” Journal of the European Economic Association 9(2):177–208.

Acemoglu, Daron, Simon Johnson and James Robinson. 2005. “The Rise of Europe: At- lantic Trade, Institutional Change, and Economic Growth.” American Economic Review 95(3):546–579.

Ardant, Gabriel. 1975. Financial Policy and Economic Infrastructure of Modern States and Nations. In The Formation of National States in Western Europe, ed. Charles Tilly. New Jersey: Princeton University Press.

Barkey, Karen and Sunita Parikh. 1991. “Comparative Perspectives on The State.” Annual Review of Sociology 17:523–549.

Barro, Robert J. 1979. “On the determination of the public debt.” Journal of Political Economy 87(5):940–971.

Bates, Robert H. and Da-Hsiang D. Lien. 1985. “A Note on Taxation, Development, and Representative Government.” Politics & Society 14(1):53–70.

Besley, Timothy and Marta Reynal-Querol. 2014. “The Legacy of Historical Conﬂict: Evi- dence from Africa.” American Political Science Review 108:319–336.

Besley, Timothy. and Torsten Persson. 2011. Pillars of Prosperity: the Political Economics of Development Clusters. Princenton: Princeton University Press.

Bird, Richard M. and Pierre-Pascal Gendron. 2007. The VAT in Developing and Transitional Countries. New York: Cambridge University Press.

Boix, Carles, Michael K. Miller and Sebastian Rosato. 2013. “A Complete Dataset of Political Regimes, 1800-2007.” Comparative Political Studies 46(12):1523–1554.

Bonney, Richard. 1999. The Rise of the Fiscal State in Europe, c.1200-1815. Oxford: Oxford University Press.

Bordo, Michael D. 2006. “Sudden Stops, Financial Crises, and Original Sin in Emerging Countries: Dj vu?” NBER Working Papers (12393).

Braun, Juan Ll, Mat´ıasLl Braun, Ignacio Briones, JoséD´ıazB, LüdersRolf S and Gert Wagner H. 2000. “Economia Chilena 1810-1995. Cuentas.” Instituto de Econom´ıade la Universidad Católica de Chile, Documento de Trabajo 187.

Brewer, John. 1988. The Sinews of Power: War, Money and the English State: 1688-1783. Cambridge: Harvard University Press.

40 Bulow, Jeremy and Kenneth Rogoﬀ. 1989. “Sovereign Debt: Is to Forgive to Forget?” The American Economic Review 79(1):pp. 43–50.

Butcher, Charles and Ryan Griﬃths. 2015. “Alternative international systems? System structure and violent conﬂict in nineteenth-century West Africa, Southeast Asia, and South Asia.” Review of International Studies 41:715–737.

Calomiris, Charles and Stephen Haber. 2014. Fragile by Design: The Political Origins of Banking Crises and Scarce Credit. Princeton, NJ: Princeton University Press.

Calvo, Guillermo A. 1988. “Servicing the Public Debt: The Role of Expectations.” American Economic Review 78(4):647–661.

Catao, Luis. 2006. “Sudden Stops and Currency Drops: A Historical Look.” IMF Working Paper pp. 1–61.

Centeno, Miguel Angel. 2002. Blood and Debt. War and the Nation-State in Latin America. Pennsylvania University Press.

Chwieroth, Jeﬀrey M. and Andrew Walter. 2013. “Financial Crises and Political Turnover: A Long Run Panoramic View.” Mimeo .

Collier, Simon and William Sater. 2004. A History of Chile. Cambridge University Press.

Comin, Diego and Bart Hobijn. 2010. “An Exploration of Technology Diﬀusion.” American Economic Review 100(5):2031–59.

Daunton, Martin. 2001. Trusting Leviathan: the Politics of Taxation in Britain, 1799-1914. New York: Cambridge University Press. de Mesquita, Bruce Bueno and Alastair Smith. 2013. “Aid: Blame It All on “Easy Money”.” Journal of Conﬂict Resolution 57(3):524–537. della Paolera, Gerardo and Alan M. Taylor. 2013. “Sovereign debt in Latin America, 1820- 1913.” Revista de Historia Econmica / Journal of Iberian and Latin American Economic History (New Series) 31:173–217.

Desch, Michael C. 1996. “War and Strong States, Peace and Weak States?” International Organization 50(2):237–268.

Dincecco, Mark. 2011. Political Transformations and Public Finances: Europe, 1650-1913. Cambridge University Press.

Dincecco, Mark, Giovanni Federico and Andrea Vindigini. 2011. “Warfare, Taxation, and Political Change: Evidence from the Italian Risorgimento.” Journal of Economic History 71:887–914.

Dincecco, Mark, James Fenske and Massimiliano Gaetano Onorato. 2016. “Is Africa Diﬀer- ent? Historical Conﬂict and State Development.” Mimeo .

41 Dincecco, Mark and Mauricio Prado. 2012. “Warfare, ﬁscal capacity, and performance.” Journal of Economic Growth 17(3):171–203.

Dornbusch, Rudiger and Jacob A. Frenkel. 1982. “The Gold Standard and the Bank of England in the Crisis of 1847.” NBER Working Papers (1039).

Drelichman, Mauricio and Hans-Joachim Voth. 2011. “Lending to the Borrower from Hell: Debt and Default in the Age of Philip II.” The Economic Journal 121:1205–1227.

Eichengreen, Barry. 1991. “Historical Research on International Lending and Debt.” Journal of Economic Perspectives 5(2):149–169.

Ekelund, Robert B. and Robert D. Tollison. 1981. Mercantilism as a Rent-Seeking Society. College Station: Texas A&M.

Ertman, Thomas. 1997. Birth of the Leviathan. Cambridge: Cambridge University Press.

Feis, Herbert. 1930. Europe, the World’s Banker 1870-1914. New Haven: Yale University Press.

Findlay, Ronald and Kevin O’Rourke. 2007. Power and Plenty. Trade, War and the World Economy in the Second Millenium. Princeton: Princeton University Press.

Fischer, Wolfram and Peter Lundgreen. 1975. The Recruitment and Training of Adminis- trative and Technical Personnel. In The Formation of Nation States in Western Europe, ed. Charles Tilly. Princeton University Press.

Flandreau, Marc and Juan H. Flores. 2012. “The Peaceful Conspiracy: Bond Markets and International Relations During the Pax Britannica.” International Organization 66:211– 241.

Flores-Macias, Gustavo and Sarah E. Kreps. 2013. “Political Parties at War: A Study of American War Finance, 17892010.” American Political Science Review 107:833–848.

Frankema, Ewout. 2011. “Colonial taxation and government spending in British Africa, 1880-1940: Maximizing revenue or minimizing eﬀort?” Explorations in Economic History 48(1):136 – 149.

Gardner, Leigh A. 2012. Taxing Colonial Africa. The Political Economy of British Imperi- alism. Oxford: Oxford University Press.

Gennaioli, Nicola and Hans-Joachim Voth. 2015. “State Capacity and Military Conﬂict.” Review of Economic Studies .

Goyer, Doreen S. and Gera E. Draaijer. 1992a. The Handbook of National Population Cen- suses: Africa and Aisa. Wesport, CT: Greenwood Press.

Goyer, Doreen S. and Gera E. Draaijer. 1992b. The Handbook of National Population Cen- suses: Europe. Westport, CT: Greenwood Press.

42 Goyer, Doreen S. and Gera E. Draaijer. 1992c. The Handbook of National Population Cen- suses: Latin America and the Caribbean, North America, and Oceania. Wesport, CT: Greenwood Press.

Griﬃths, Ryan D. and Charles R. Butcher. 2013. “Introducing the International System(s) Dataset (ISD), 18162011.” International Interactions 39(5):748–768.

He, Wenkai. 2013. Paths toward the Modern Fiscal State. Rhode Island: Harvard University Press.

Hendrix, Cullen S. 2010. “Measuring state capacity: Theoretical and empirical implications for the study of civil conﬂict.” Journal of Peace Research 47(3):273–285.

Herbst, Jeﬀrey. 2000. States and Power in Africa: Comparative Lessons in Authority and Control. New Jersey: Princeton University Press.

Hintze, Otto. 1975. The historical essays of Otto Hintz. New York: Oxford University Press.

Hoﬀman, Philip and Jean-Laurent Rosenthal. 2000. “Divided we fall: The political economy of warfare and taxation.” Mimeo .

Hoﬀman, Philip T. 1994. Early Modern France, 1450-1700. In Fiscal Crises, Liberty, and Representative Government, 1450-1789, ed. Philip T. Hoﬀman and Kathryn Norberg. Stanford: Standford University Press pp. 226–252.

Homer, Sidney and Richard Sylla. 2005. A History of Interest Rates. Fourth ed. New Jersey: Wiley Finance.

Jorgensen, Erika and Jeﬀrey Sachs. 1988. “Default and Renegotiation of Latin American Foreign Bonds in the Interwar Period.” NBER Working Papers 2636.

Karaman, K. Kivan¸cand S¸evket Pamuk. 2013. “Diﬀerent Paths to the Modern State in Europe: The Interaction Between Warfare, Economic Structure, and Political Regime.” American Political Science Review 107:603–626.

Kindleberger, Charles P. and Robert Z. Aliber. 2005. Manias, Panics, and Crashes: A History of Financial Crisis. New Jersey: Wiley.

Kurtz, Marcus J. 2013. Latin American State Building in Comparative Perspective. Social Foundations of Institutional Order. New York: Cambridge University Press.

Levi, Margaret. 1988. Of Rule and Revenue. Berkeley: University of California Press.

Lindert, Peter H. 2004. Growing Public: Social Spending and Economic Growth since the Eighteenth Century. New York: Cambridge University Press.

Lindert, Peter H. and Peter J. Morton. 1989. How Sovereign Debt has worked. In Developing Country Debt and Economic Performance, ed. Jeﬀrey Sachs. Vol. 1 Chicago: Chicago University Press.

43 Mahate, Ashraf A. 1994. “Contagion Eﬀects of Three Late Nineteenth-Century British Bank Failures.” Business and Economic History 12(1):102–115.

Mann, Michael. 1980. “State and Society, 1130-1815: An Analysis of English state ﬁnances.” 1:165–208.

Mann, Michael. 1984. “The autonomous power of the state : its origins, mechanisms and results.” European Journal of Sociology / Archives Européennesde Sociologie / Europäisches Archiv fürSoziologie 25(2):185–213.

Mares, Isabela and Didac Queralt. 2015. “The Non-Democratic Origins of Income Taxation.” Comparative Political Studies 48(14):1974–2009.

Marichal, Carlos. 1989. A century of debt crisis in Latin America: From Independence to the Great Depression, 1820-1930. Princeton University Press.

Morrison, Kevin M. 2009. “Oil, Nontax Revenue, and the Redistributional Foundations of Regime Stability.” International Organization 63:107–138.

Moss, Todd J., Gunilla Pettersson Gelander and Nicolas van de Walle. 2006. “An Aid- Institutions Paradox? A Review Essay on Aid Dependency and State Building in Sub- Saharan Africa.” Center for Global Development Working Paper 74.

Musgrave, Richard A. 1969. Fiscal Systems. New Haven: Yale University Press.

Neal, Larry. 1990. The Rise of Financial Capitalism. New York: Cambridge University Press.

Neal, Larry. 1998. “The Financial Crisis of 1825 and the Restructuring of the British Finan- cial Systems.” The Federal Reserve Bank of St. Louis Review 80:53–76.

Neal, Larry D. and Marc D. Weidenmier. 2003. Crises in the Global Economy from Tulips to Today. In Globalization in Historical Perspective, ed. Michael D. Bordo, Alan M. Taylor and Jeﬀrey G. Williamson. Chicago: University of Chicago Press pp. 473–514.

North, Douglass C. and Barry R. Weingast. 1989. “Constitutions and Commitment: The Evolution of Institutional Governing Public Choice in Seventeenth-Century England.” The Journal of Economic History 49(4):pp. 803–832.

Nunn, Nathan and Diego Puga. 2012. “Ruggedness: The Blessing of Bad Geography in Africa.” Review of Economics and Statistics 94(1):20–36.

O’Brien, Patrick. 2001. “Fiscal Exceptionalism: Great Britain and its European Rivals From Civil War to Triumph at Trafalgar and Waterloo.” Department of Economic History, LSE, Working Paper 65.

Obstfeld, Maurice and Alan M. Taylor. 2003. “Sovereign risk, credibility and the gold standard: 18701913 versus 192531*.” The Economic Journal 113(487):241–275.

44 Obstfeld, Maurice and Alan M. Taylor. 2004. Global Capitalism: Integration, Crisis, and Growth. New York: Cambridge University Press. Persson, Torsten and Guido Tabellini. 2003. The Economic Effects of Constitutions. Cam- bridge: MIT. Putterman, Louis. 2007. “State antiquity index, version 3. Mimeo, Brown University.”. Queralt, Didac. 2015. “From Mercantilism to Free Trade: A History of Fiscal Capacity Building.” Quarterly Journal of Political Science 10(2):221–273. Rasler, Karen A. and William R. Thompson. 1985. “War Making and State Making: Govern- mental Expenditures, Tax Revenues, and Global Wars.” The American Political Science Review 79(2):491–507. Reid, Richard. 2012. Warfare in African History. New York: Cambridge University Press. Reinhart, Carmen M. and Kenneth S. Rogoff. 2009. This Time is Different. Eight Centuries of Financial Folly. New Jersey: Princeton University Press. Rosenthal, Jean-Laurent and R. Bin Wong. 2011. Before and Beyond Divergence: The Politics of Economic Change in China and Europe. Cambridge: Harvar. Ross, Michael L. 2001. “Does Oil Hinder Democracy?” World Politics 53(3):325–361. Sarkees, Meredith Reid and Frank Wayman. 2010. Resort To War. CQ Press. Scheve, Kenneth and David Stasavage. 2010. “The Conscription of Wealth: Mass Warfare and the Demand for Progressive Taxation.” International Organization 64(4):529–561. Schumpeter, Joseph A. 1918. The Crisis of the Tax States. In Joseph A. Schumpeter: The Economics and Sociology of Capitalism, ed. Swedberg Richard. 1991 ed. Princeton, NJ: Princeton University Press. Shea, Patrick E. 2013. “Financing Victory: Sovereign Credit, Democracy, and War.” Journal of Conflict Resolution . Slantchev, Branislav L. 2012. “Borrowed Power: Debt Finance and the Resort to Arms.” American Political Science Review 106:787–809. Sprague, O. M. W. 1917. “Loans and Taxes in War Finance.” The American Economic Review 7(1):199–213. Stasavage, David. 2011. States of Credit. Size, Power, and the Development of European Polities. New Jersey: Princeton University Press. Stone, Irving. 1999. The Global Export of Capital from Great Britain, 1865-1914: A Statis- tical Survey. New York: MacMillan. Stubbs, Richard. 1999. “War and Economic Development: Export-Oriented Industrialization in East and Southeast Asia.” Comparative Politics 31(3):337–355.

45 Suter, Christian. 1992. Debt cycles in the World-Economy: foreign Loans, ﬁnancial crises, and debt settlements, 18201990. Boulder, Co:: Westview Press.

Taylor, Alan M. 2006. Foreign Capital Flows. In The Cambridge Economic History of Latin America Volume 2: The Long Twentieth Century, ed. Victor Bulmer-Thomas, John Coatsworth and Roberto Cortes-Conde. Cambridge: Cambridge University Press pp. 57– 100.

Taylor, Alan M. and Jeﬀrey G. Williamson. 1994. “Capital Flows to the New World as an Intergenerational Transfer.” Journal of Political Economy 102(2):348–371.

Taylor, Brian D. and Roxana Botea. 2008. “Tilly Tally: War-Making and State-Making in the Contemporary Third Worlds.” International Studies Review 10(1):27–56.

Thies, Cameron. 2005. “War, Rivalry, and State Building in Latin America.” American Journal of Political Science 49:451–465.

Thies, Cameron G. 2007. “The Political Economy of State Building in Sub-Saharan Africa.” The Journal of Politics 69(3):716–731.

Tilly, Charles. 1990. Coercion, Capital, and European States. Cambridge: Basil Blackwell.

Tilly, Charles. 1993. European Revolutions: 1492-1992. Oxford: Blackwell.

Tin-bor Hui, Victoria. 2004. “Toward a Dynamic Theory of International Politics: Insights from Comparing Ancient China and Early Modern Europe.” International Organization 58(1):175–205.

Tomz, Michael. 2007. Reputation and International Cooperation: Sovereign Debt across Three Centuries. Cambridge: Cambridge University Press.

US AID. 2012. Fiscal Reform and Economic Governance Project, 2004-2010. US AID.

Vizcarra, Catalina. 2009. “Guano, Credible Commitments, and Sovereign Debt Repayment in Nineteenth-Century Peru.” The Journal of Economic History 69(02):358–387.

Webber, Carolyn and Aaron Wildavsky. 1986. A history of taxation and expenditure in the Western world. Berkeley: University of California Press.

Wimmer, Andreas and Brian Min. 2009. “The Location and Purpose of Wars Around the World: A New Global Dataset, 18162001.” International Interactions 35(4):390–417.

46 Figure 1: Chile at War. An example of war financing as a function of access to the international financial market. The light-gray area identifies years at war without access to international lending markets. The dark-gray area identifies years at war with access to international lending markets. Debt and tax data drawn from Braun et al.(2000). 30 20 10 0 1820 1840 1860 1880 1900 1920 year

Tax Revenue as % of GDP Public Foreign Debt as % of GDP

47 Figure 2: Partial Correlations of Personal Income Tax and Endogenous War- Financing. Estimates are drawn from column 4 in Table1.

Denmark Denmark South Africa South Africa 4 4 BelgiumTurkey BelgiumTurkey Zimbabwe Zimbabwe

Italy Austria Italy Chile 2 2 Austria Brazil United Kingdom Chile Thailand ThailandIndonesiaFinland Peru IndonesiaFinland Brazil United Kingdom Spain Argentina VenezuelaColombia France UnitedNew ZealandStates of America UnitedNewSpain BoliviaStatesZealand of America ColombiaVenezuela BoliviaHungaryMalaysia Ecuador HondurasIcelandZambia MalaysiaPeruKenyaHonduras 0

Kenya 0 Ecuador TunisiaMyanmar Mexico TunisiaZambiaIcelandMyanmarArgentinaParaguay China ParaguayUruguayFranceGuatemalaPhilippinesIreland PortugalUruguayGuatemalaIreland PortugalCanada Hungary Canada ChinaAustraliaNicaragua Philippines AustraliaNicaraguaGreece GreeceNorwayNetherlands NorwayNetherlands DominicanIndiaPolandMorocco Republic IndiaMexicoDominicanPolandMorocco Republic PanamaEl SalvadorJapan Panama El Salvador Sweden Sri Lanka Japan Costa Rica SwedenCostaSri Lanka Rica -2 GermanyIvory Coast -2 Residuals from Regressing Residuals SwitzerlandEgyptSouth Korea from Regressing Residuals Ivory CoastEgypt

PIT as % of GDP Controls on PIT as % of GDP Controls on Switzerland Germany South Korea NigeriaRomania NigeriaRomania -4 Russia -4 Russia -5 0 5 10 -20 0 20 40 Residuals from Regressing Residuals from Regressing # Years at War while in Default on Controls # Years at War while Having Access to Credit on Controls

coef = .167, (robust) se = .075, t = 2.21 coef = .028, (robust) se = .027, t = 1.03 (a) War in Default (b) War having Credit Access

48 Figure 3: British Capital Exports from 1865 to 1914. In light-gray: Banking panics of 1865, 1873, and 1890. In dark-gray: the stock crisis of 1907. Source: Stone(1999). 200 150 100 50 Milions of current Pounds Sterling of Pounds current Milions 0 1865 1870 1875 1880 1885 1890 1895 1900 1905 1910 1915

49 Figure 4: Partial Correlations of Personal Income Tax and Exogenous War- Financing. Estimates are drawn from column 4 in Table5. 10 10

Belgium Belgium Namibia Namibia Denmark Denmark 5 5 ZimbabweIsraelTurkey IsraelZimbabwe South AfricaFinland FinlandTurkey South Africa AustriaItaly LesothoItaly Spain Lesotho AustriaThailand France SpainBhutan PeruBhutan NetherlandsBrazil IndonesiaThailand UnitedIndonesiaSwazilandIrelandIran StatesIceland of America YemenIranNepalIcelandPortugalSwazilandUnitedPeru States of America Chile NewPolandPortugal ZealandZambiaChina Paraguay Argentina Brazil NetherlandsLithuaniaPolandNewIreland ZealandChile BoliviaNepalLithuaniaUruguayYemenMacedonia HungaryParaguayZambiaUruguayVenezuelaMacedoniaColombiaBolivia ColombiaVenezuelaMongolia Morocco Argentina Mongolia 0 Norway 0 MoroccoKenyaNorwayTunisia Japan Japan NicaraguaHondurasTunisiaSwitzerlandKenyaGreece PhilippinesGreeceSwedenSwitzerlandEcuadorHondurasGuatemalaNicaraguaChina CambodiaEcuadorGuatemalaHungarySlovakiaPhilippinesSwedenArmeniaSloveniaEstoniaMexicoEthiopia EstoniaArmeniaSloveniaSouthMexicoSlovakiaEthiopia Korea Cambodia SouthCzechIvoryLatvia RomaniaKorea Republic CyprusCoast FranceCyprusIvoryAzerbaijanCroatiaRomaniaLatvia CoastCzech Republic AzerbaijanCroatiaAustraliaChadCanadaMaliSenegalVietnam CanadaChadCongoAustralia ElBurundi SalvadorMyanmarCongoIndiaBulgaria GermanyMalaysiaMaliSenegalKazakhstanIndiaDemocraticTajikistanPanamaBulgariaCostaBurundiMyanmarRwanda RicaEl Republic Salvador of the Congo Residuals from Regressing Residuals Rwanda from Regressing Residuals CostaDemocraticTajikistanKazakhstanGuineaPakistan RicaMalaysia Republic of the Congo PakistanGuinea

PIT as % of GDP Controls on Panama PIT as % of GDP Controls on Nigeria Georgia NigeriaGermanySriBosnia Lanka and Herzegovina DominicanBosniaSri Lanka Republicand Herzegovina MoldovaBelarusLebanonDominican Republic LebanonBelarusMoldovaVietnam Georgia Egypt Ukraine Ukraine Madagascar MadagascarEgypt Russia BangladeshAlbania Albania Bangladesh Russia -5 -5 -10 -5 0 5 10 15 -5 0 5 10 15 Residuals from Regressing Residuals from Regressing # Years at War while Credit Stops on Controls # Years at War while Credit Flows on Controls

coef = .251, (robust) se = .055, t = 4.54 coef = -.252, (robust) se = .069, t = -3.65 (a) War while Credit Stops (b) War while Credit Flows

50 Table 1: Personal Income Tax Today (as % of GDP) as a Function of War and Endogenous Credit Access in the Long-Nineteenth Century

(1) (2) (3) (4) (5) (6) (7) (8) (9) # Years at War 1816-1913 0.037 (0.024) # Years at War while in Default 0.150** 0.167** 0.168** 0.186** 0.137* 0.159* 0.157** 0.171** (0.071) (0.074) (0.076) (0.078) (0.072) (0.081) (0.075) (0.077) # Years at War with Access to Credit 0.034 0.032 0.028 -0.005 0.020 0.031 0.020 0.027 (0.025) (0.026) (0.027) (0.052) (0.043) (0.026) (0.029) (0.025) Population Density in 1820 3.389** 3.493** 3.420** 3.378* 3.384** 3.172* 3.473** 3.324* 2.983 (1.578) (1.597) (1.656) (1.723) (1.593) (1.839) (1.707) (1.658) (1.831) Oil Producer -0.822 -0.945 -0.919 -0.928 -1.016 -0.632 -0.944 -1.013 -0.874 (0.640) (0.659) (0.673) (0.687) (0.714) (0.913) (0.749) (0.682) (0.667) Sea Access 0.021** 0.021** 0.022** 0.022** 0.020** 0.022* 0.022** 0.024*** 0.024*** (0.008) (0.008) (0.008) (0.008) (0.009) (0.011) (0.008) (0.009) (0.008) Desert -0.051 -0.057 -0.060 -0.059 -0.061 -0.078 -0.053 -0.064 -0.050 (0.062) (0.060) (0.060) (0.060) (0.065) (0.085) (0.074) (0.062) (0.055) # Years in default -0.009 -0.008 -0.017 -0.008 -0.008 -0.015 -0.011

51 (0.013) (0.014) (0.014) (0.014) (0.013) (0.014) (0.014) Great power 0.317 (1.549) War Location 0.052 (0.058) War Casualties 1816-1913 0.906 (1.868) Ethnic fractionalization 0.325 (1.546) # years at Civil War 1816-1913 0.066 (0.051) WWI Participant 0.752 (0.906) Intercept 3.269** 3.390** 3.496** 3.458** 3.568** 3.633** 3.338* 3.478** 2.813* (1.361) (1.359) (1.418) (1.448) (1.390) (1.696) (1.822) (1.416) (1.604) Colonial Origins FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Region FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 63 63 63 63 63 54 62 63 63 R-squared 0.756 0.759 0.760 0.761 0.766 0.723 0.759 0.766 0.764 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 2: External capital Stock by Country in the Long-Nineteenth Century

1825 1855 1870 1890 1914 United Kingdom 0.5 0.7 4.9 12.1 19.5 France 0.1 - 2.5 5.2 8.6 Germany - - - 4.8 6.7 Netherlands 0.3 0.2 0.3 1.1. 1.2 United States 0.0 0.0 0.0 0.5 2.5 Canada - s - 0.1 0.2 All 0.9 0.9 7.7 23.8 38.7 UK/all 0.56 0.78 0.64 0.51 0.50 World GDP - - 111 128 221 Values represent gross foreign assets in current USD billion. Source: Table 2.1 in Obstfeld and Taylor (2004).

52 Table 3: Banking Crises and Stock Market Crashes in London, 1816-1913

Banking Crises Stock Market Crises 1825 1865 1837 1866 1838 1867 1839 1910 1840 1911 1847 1912 1848 1913 1849 1850 1857 1866 1873 1890 Source: Reinhart and Rogoﬀ (2009). 1873 banking panic added. Results robust to its exclusion (see Appendix Table A-10).

53 Table 4: Frequency and Duration of War as a function of Credit Access. Refer to Appendix Figure A-6 for a Visual Illustration.

Endogenous access to credit Countries in sample: 63 Wars in sample: 111 Total war-year-country: 399, out of which spa - 348 were fought while having Access to Credit, with avg. duration of 2.34 years (1.88)a spa - 51 were fought while in Default, with avg. duration of 2.05 years (1.52)

Exogenous access to credit Countries in sample: 107 Wars in sample: 114 Total war-year-country: 466, out of which spa - 222 were fought while Credit Flows, with avg. duration of 2.00 years (1.60) spa - 244 were fought while Sudden-Stop, with avg. duration of 1.90 years (1.24)

Exogenous access to credit including Secessionist Wars Countries in sample: 107 Wars in sample: 147 Total war-year-country: 624, out of which spa - 314 were fought while Credit Flows, with avg. duration of 2.02 years (1.54) spa - 349 were fought while Sudden-Stop, with avg. duration of 2.02 years (1.37) The difference in war-year-country between the endogenous and exogenous credit cases results from countries not included in Reinhart and Rogoff(2009) fighting wars against states included in that sample. a Standard deviation of duration in parenthesis.

54 Table 5: Personal Income Tax Today (as % of GDP) as a Function of War and Exogenous Credit Access in the Long-Nineteenth Century

(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) # Years at War 1816-1913 0.052* (0.028) # Years at War while Credit Stops 0.273*** 0.275*** 0.251*** 0.221*** 0.303*** 0.269*** 0.241*** 0.261*** 0.273*** (0.056) (0.056) (0.055) (0.074) (0.081) (0.055) (0.055) (0.053) (0.060) # Years at War while Credit Flows -0.200*** -0.200*** -0.252*** -0.191*** -0.206*** -0.198*** -0.186*** -0.214*** -0.201*** (0.057) (0.057) (0.069) (0.059) (0.068) (0.058) (0.059) (0.056) (0.052) Population Density in 1820 1.623 1.238 1.247 0.788 1.159 2.314 1.220 1.221 0.799 1.243 (1.365) (1.318) (1.332) (1.396) (1.311) (1.485) (1.572) (1.318) (1.246) (1.336) Oil Producer 0.098 0.127 0.108 0.130 0.043 0.156 0.218 0.022 -0.006 0.125 (0.474) (0.468) (0.474) (0.464) (0.472) (0.679) (0.498) (0.477) (0.459) (0.466) Sea Access 0.027*** 0.028*** 0.028*** 0.029*** 0.027*** 0.028*** 0.026*** 0.029*** 0.029*** 0.028*** (0.007) (0.007) (0.007) (0.007) (0.008) (0.010) (0.008) (0.007) (0.007) (0.007) Desert Territory 0.004 0.013 0.014 0.015 0.008 0.028 0.006 0.012 0.011 0.013 (0.045) (0.045) (0.046) (0.045) (0.046) (0.067) (0.048) (0.045) (0.045) (0.045) # Years in Default 1816-1913 0.008

55 (0.010) Great Power 2.712** (1.166) War Location 0.054 (0.040) War Casualties 1816-1913 -0.481 (0.880) Ethnic fractionalization -0.306 (1.254) # Years at Civil War 1816-1913 0.066* (0.037) Average War duration 1816-1913 0.008 (0.124) WWI participant 1.261** (0.533) Intercept 1.250 1.331 1.295 1.279 1.345 1.347 1.591 1.281 0.739 1.327 (0.862) (0.829) (0.843) (0.811) (0.819) (1.131) (1.263) (0.826) (0.810) (0.843) Colonial Origins FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Region FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 106 106 106 106 106 87 105 106 106 106 R-squared 0.551 0.587 0.588 0.610 0.594 0.554 0.585 0.592 0.609 0.587 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Table 6: Models of Personal Income Tax Today (as % of GDP) using Conservative Deﬁnition of Statehood, Liberal Deﬁnition of Warfare, and Alternative Outcome Variable

Dependent Variable → PIT 2000s Tax Staﬀ 2000s

Sample → Sovereign States Only Secessionist War Included All States (1) (2) (3) (3) (4) (5) (6) (3) (7) (8) (9) # Years at War while Credit Stops 0.150*** 0.161*** 0.114 0.181*** 0.161*** 0.180*** 0.036** 0.035** 0.034** (0.052) (0.057) (0.067) (0.050) (0.054) (0.055) (0.015) (0.014) (0.014) # Years at War while Credit Flows -0.146** -0.191** -0.140** -0.069 -0.091 -0.069 -0.018 -0.021 -0.017 (0.060) (0.085) (0.065) (0.074) (0.085) (0.075) (0.019) (0.019) (0.019) Population Density in 1820 4.399* 3.859 4.542* 1.458 1.128 1.456 0.217 0.188 0.216 (2.419) (2.845) (2.484) (1.349) (1.437) (1.359) (0.239) (0.255) (0.242) Oil Producer 0.311 0.302 0.328 0.015 0.029 0.017 -0.106 -0.104 -0.103 (0.589) (0.620) (0.597) (0.471) (0.472) (0.475) (0.097) (0.097) (0.100) 56 Sea Access 0.027** 0.029** 0.025** 0.027*** 0.028*** 0.027*** 0.002 0.002 0.002 (0.011) (0.011) (0.012) (0.007) (0.007) (0.007) (0.001) (0.001) (0.001) Desert Territory 0.044 0.060 0.080 0.013 0.013 0.013 0.000 0.001 0.001 (0.064) (0.057) (0.048) (0.045) (0.045) (0.045) (0.005) (0.005) (0.005) Great Power 1.432 1.964 0.136 (1.552) (1.257) (0.237) War Location 0.037 0.001 0.001 (0.031) (0.026) (0.005) Intercept 1.999* 1.842* 2.272* 1.158 1.111 1.159 -0.116 -0.128 -0.118 (1.173) (1.062) (1.121) (0.851) (0.842) (0.858) (0.136) (0.140) (0.138) Colonial Origins FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Region FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Observations 49 49 49 106 106 106 79 79 79 R-squared 0.825 0.831 0.831 0.584 0.597 0.584 0.669 0.672 0.669 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 Table 7: Personal Income Tax Today (as % of GDP) as a Function of War and En- dogenous Credit Access in the Long-Nineteenth Century, with Special Attention to Omitted Variable Bias and Selection into War

Initial Capacity Ongoing War (1) (2) (3) (4) # Years at War while Credit Stops 0.233*** 0.239*** 0.135* 0.166** (0.062) (0.079) (0.070) (0.070) # Years at War while Credit Flows -0.247*** -0.241*** -0.085 -0.077 (0.068) (0.091) (0.078) (0.078) Population Density in 1820 0.770 0.696 1.019 0.897 (1.425) (0.863) (1.483) (1.408) Oil Producer 0.146 0.156 0.225 0.178 (0.463) (0.517) (0.470) (0.459) Sea Access 0.027*** 0.030*** 0.026*** 0.030*** (0.007) (0.007) (0.008) (0.007) Desert Territory 0.012 -0.016 0.004 -0.024 (0.046) (0.045) (0.046) (0.033) Great Power 2.814** 2.672** 3.268** 3.101** (1.195) (1.127) (1.260) (1.189) Modern Census 0.801 1.349 (1.239) (1.270) State Antiquity 0.001 0.002 (0.001) (0.001) Intercept 1.339 0.564 1.461* 0.423 (0.832) (0.922) (0.864) (1.035) Colonial Origins FE Yes Yes Yes Yes Region FE Yes Yes Yes Yes Observations 106 103 106 103 R-squared 0.613 0.646 0.586 0.617 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

57 Table 8: Personal Income Tax Today (as % of GDP) as a Function of War and En- dogenous Credit Access in the Long-Nineteenth Century, with War Data drawn from COW’s Inter-State Military Conﬂict Database

All countries Non-Initiators (1) (2)3 (3) (4) # Years at War while Credit Stops 0.379*** 0.396*** 0.453*** 0.473** (0.099) (0.107) (0.152) (0.183) # Years at War while Credit Flows 0.075 0.062 0.162 0.121 (0.173) (0.175) (0.207) (0.251) Population Density in 1820 1.206 1.242 1.260 1.284 (1.467) (1.489) (1.480) (1.506) Oil Producer -0.065 -0.059 -0.163 -0.153 (0.445) (0.452) (0.449) (0.455) Sea Access 0.028*** 0.028*** 0.027*** 0.028*** (0.007) (0.007) (0.007) (0.007) Desert Territory -0.020 -0.018 -0.021 -0.019 (0.031) (0.032) (0.031) (0.032) Great Power 0.602 0.699 1.374 1.481 (1.374) (1.404) (1.275) (1.274) Modern Census by 1820 0.879 0.922 1.186 1.220 (1.206) (1.220) (1.090) (1.136) Net Victory -0.048 -0.035 (0.098) (0.116) Intercept 0.748 0.711 0.810 0.797 (0.814) (0.829) (0.815) (0.825) Colonial Origins FE Yes Yes Yes Yes Region FE Yes Yes Yes Yes Observations 102 102 102 102 R-squared 0.651 0.652 0.647 0.647 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

58 Table 9: Short-term Eﬀects of War Making on State Capacity as a function of War-Financing Instruments

Modern Census Log(Rail Lines) By 1913 By 1913 Delay Delay By 1913 By 1913 By 1913 (1) (2) (3) (4) xxx (5) (6) (7) Probit Probit OLS OLS OLS OLS OLS # Years at War while Credit Stops 0.146** 0.151** -2.716** -2.606* 0.100** 0.092* 0.087* (0.069) (0.071) (1.311) (1.323) (0.047) (0.048) (0.049) # Years at War while Credit Flows -0.029 -0.034 1.680 1.726 -0.070 -0.117* -0.121* (0.051) (0.052) (1.458) (1.474) (0.061) (0.069) (0.065) Population Density in 1820 1.284* 1.270* -23.522 -21.260 1.493 1.081 1.062 (0.691) (0.687) (23.825) (24.037) (1.135) (1.065) (1.047) Oil Producer -0.150 -0.069 -1.485 -0.973 -0.131 -0.080 -0.120 (0.319) (0.339) (10.788) (11.896) (0.511) (0.596) (0.534) Sea Access 0.009** 0.008* -0.502*** -0.477*** -0.001 0.001 -0.001 (0.005) (0.005) (0.168) (0.178) (0.007) (0.007) (0.006)

59 Desert Territoru -0.015 -0.012 -1.348 -1.229 0.084* 0.076 0.079 (0.031) (0.032) (0.917) (0.979) (0.050) (0.052) (0.051) State Antiquity -0.000 -0.019 -0.000 (0.001) (0.023) (0.001) Great Power -15.279 -14.595 1.748*** 1.809*** (16.825) (16.317) (0.609) (0.605) Modeern Census by 1820 0.342 (0.555) Intercept -1.415*** -1.310* 151.254*** 157.767*** 5.719*** 5.974*** 5.938*** (0.542) (0.683) (15.411) (18.265) (1.167) (1.819) (1.186) Topographical controls No No No No Yes Yes Yes Colonial Origins FE Yes Yes Yes Yes Yes Yes Yes Region FE Yes Yes Yes Yes Yes Yes Yes Observations 102 99 106 103 62 61 62 R-squared . . 0.521 0.515 0.610 0.633 0.634 The Great Power indicator in columns 1 and 2 cannot be estimated due to perfect collinearity. To fully account for the topographical characteristics of rail line building, models include additional controls: land area, tropical weather, and terrain ruggedness. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 **NOT FOR PUBLICATION**

Supplementary Online Appendices

The Legacy of War on Fiscal Capacity

These appendices contain materials, results and robustness checks that supplement the main text.

Appendices

A. Data Details

B. Cross-Sectional Distribution of Warfare and Acces to Credit

C. Sensibility Tests involving the Main Dependent Variable

D. Inﬂuence of Outliers

E. Inﬂuence of Fixed Eﬀects

F. Sensibility Tests addressing Measurement and Sample Decisions

G. Sensibility Test involving Crises Timing and Length of Sudden-Stops

H. Alternative Sources of War Financing

I. Value-Added Tax as Outcome Variable

J. Military Alliances, British Colonies, and British Wars

K. Instrumenting for War-Making

L. Including Bad Controls

M. Further Evidence of the Exogeneity of Sudden-stops

N. Further Evidence of the Lending Frenzy

O. Appendix-Speciﬁc References

i A. Data Details

Personal Income Tax. The IMF Global Financial Statistics (GFS) PIT data that I use refers to cash-accounts (as recommended by the IMF). For the few cases that these data is not available, I use non-cash values, which correlate at .97 with cash-accounts. To minimize influence of abnormal values, I compute average PIT values as a percentage of GDP for the years 1995- 2005. This decade maximizes the sample size compared to earlier and later decades. Personal Income Tax data is scarce, even for the IMF. Scarcity is indeed indicative of the sophistication of this tax. Missing values are filled in with various sources. Crucially, column 1 in Table A-6 shows that data augmentation does not change the point estimates of interest. That is, models that only use GFS PIT data yield the same results than models that augment GFS data with additional sources. Cases not covered by the GFS are filled as follows: for Chile, Nicaragua, Ecuador and Guatemala, data are drawn from the Inter-American Development Bank Dataset;45 for Nepal, data are drawn from the Ministry of Finance;46 For Sri Lanka, data are drawn from the Department of Fiscal Policy;47 for Lebanon, data are available from the Ministry of Fi- nance for the 2000-5 period;48 For Zambia, data are for 2005 only, and are drawn from CMI Report;49 For Guinea, Rwanda, Chad, Namibia and Yemen, Kenya, Mali, Nigeria, Philip- pines, Senegal and Vietnam only 2004 data are drawn from the pilot study of the USAID Fiscal Reform and Economic Governance Project, 2004-2010. Again, results do not change as a result of the data augmentation (refer to column 1 in Table A-6).

45IDB (Inter-American Development Bank) and CIAT (Inter- American Center of Tax Administrations). 2012. Latin America and the Caribbean Fiscal Burden Database, 1990-2010. Database n. IDB-DB-101. Washington, DC. 46Nepal Rastra Bank, Research Department Government Finance Division. 2014. A Handbook of Gov- ernment Finance Statistics. 47Available at http://www.treasury.gov.lk/fiscal-operations/fiscal-data.html. Accessed, March 31, 2015. 48Available at ttp://www.finance.gov.lb/EN-US/FINANCE/REPORTSPUBLICATIONS/DOCUMENTSANDREPORTSISSUEDBYMOF/Pages/PublicFinanceReports.aspx. Accessed on March 31, 2015. 49Odd-Helge Fjeldstad and Kari K. Heggstad. 2011. The tax systems in Mozambique, Tanzania and Zambia: capacity and constraints Bergen: Chr. Michelsen Institute (CMI Report R 2011:3) 124 p.

ii Tax staﬀ. The size of the tax administration is drawn from the USAID Fiscal Reform and Economic Governance Project, 2004-2010. To maximize the sample size, I combine the values for 2004, 2007-10. This variable is a strong predictor of total tax revenue to GDP, as Figure A-1 shows.

Figure A-1: Total Tax Revenue vs. Size of the Tax Administration 50 40 30 20 Tax as % of GDP Revenue 10 0 0 1 2 3 Tax Staff per Thousand Capita

Default. Default data is drawn from Reinhart and Rogoff(2009). Default episodes account for outright defaults (partial or complete) and rescheduling, with which the debtor forces its creditors to accept longer repayment schedules, interest rate concessions, or both. Defaults come to an end when defaulters offer creditor(s) an acceptable payoff: e.g. refinancing the loan, some form of state monopoly revenue, even land. External debt in Reinhart and Rogoff(2009, p.11) is indistinctively issued by official (public) and private entities. I work with 63 out the 68 countries in their sample, all for which complete data is available. The five countries excluded due to tax-data limitations are: Algeria, Angola, Central African Republic, Ghana and Taiwan. Strict default periods, as those used in the analysis, are a conservative estimate of the length of the spans in which countries do not have access to credit. Defaulters live by their reputation: even today, when third party institutions veil for investors’ interests, it takes between 2.9 and 4.7 years for an average defaulter to regain access markets following the end of a default, declining to 2.9 years only recently (Gelos et

iii al. 2011). And half of defaulting countries do not full regain market access within seven years of the end of the default (Richmond and Dias 2009).

Census. A modern census involves periodicity, universality, and individual enumeration by means of house-to-house visitation. The date of the ﬁrst modern census is coded from three sources: Goyer and Draaijer(1992 b,c,a).

Military Conﬂict. The main source is Wimmer and Min(2009). Most wars can be easily matched to current states thanks to the geographical location provided in Wimmer and Min (2009). For non-obvious matches, I make the following assumptions:

1. Splits. This refers to wars attributed by Wimmer and Min(2009) to former political entities that eventually split. Countries aﬀected are: Austria-Hungary, Czechoslovakia, and Korea. To facilitate matching, entries have been duplicated and attributed evenly across current political units: Austria and Hungary, Czech Republic and Slovakia, and North Korea and South Korea, respectively. Example: suppose that Austria-Hungary fought 10 wars within 1816-1913, then I assign 10 wars to Austria and 10 wars to Hungary. The assumption is that both entities inherit evenly the ﬁscal burden and consequences of warfare.

2. Secessionist wars. Wimmer and Min’s (2009) data attribute war participation to the colonial power only. I extend this code by attributing war participation to the territory that seeks independence too.

3. Subnational to National Units. Table A-2 lists political units that were eventually incorporated to a larger unit (or merged into one). These are non-state and sub-state actors that can be easily matched to current nation-states.

4. Tentative matching. Table A-3 lists political units that cannot be matched to current units without making too many assumptions. These cases are not considered in the

iv main analysis. However, results do not change if they are (refer to column 5 in Table A-9).

5. Unmatched Units. A series of former polities map into various modern states. They are not considered in the analysis: Bornu (modern Chad, Niger and Cameroon), Khanate of Kokand (Kazakhstan and Uzbekistan), Mandingo Empire (eleven states in West Africa), Oyo (various states in West Africa), Zuku, Tukolor Empire (Mali, Nigeria and Guinea), Bambara Empire (Guinea and Bali), and Principality of Jammu (China, Tibet, Pakistan).

Civil wars are excluded from the analysis because their contribution to state building is yet to be established. Porter (1994) argues that civil war was positive for state-building in early-modern Europe. Similarly, Balcells and Kalyvas (2014) suggest that irregular warfare might serve to state building. However, others ﬁnd opposite evidence in Africa (Herbst, 2000) and Latin America (Cardenas, 2010; Centeno, 2002). Civil wars are only considered as a control.

A note on COW vs Wimmer-Min: To enter the COW inter-state war dataset prior to 1920, territorial units must possess diplomatic relations with both Britain and France. A considerable large number of states that went to war during the nineteenth century —mainly outside Europe—had not yet established sufficient relations with both of these states (Grif- fiths and Butcher, 2013). As a result, they are excluded from the COW inter-state datase. The COW offers three additional datasets: extra-state, non-state, and civil wars. Wars against or between colonies and other non-internationally recognized states entities enter these three auxiliary COW datasets. But, unlike Wimmer and Min(2009), those wars are not mapped onto current state borders, preventing a clear match between past warfare and current nation-states.

Table A-1 reports the summary statistics of these and the remaining variables. Right

v after it, Table A-2 lists quasi-direct matches of sub-national to national units, and Table A-3 lists all tentative matches.

vi Table A-1: Summary Statistics

Variable Mean Std. Dev. Min. Max. N Source

Personal Income Tax as % of GDP 1995-2005 2.999 3.258 0 15.058 107 Various Sources, see above Tax Staff per 1000 capita 2004-10 0.702 0.557 0.03 2.398 80 Various Sources, see above # Years at War with Access to Credit 7.603 12.679 0 62 63 Wimmer and Min(2009) and Reinhart and Rogoff(2009) # Years at War while in Default 1.095 2.763 0 11 63 Wimmer and Min(2009) and Reinhart and Rogoff(2009) # Years at War 1816-1913 (full sample) 4.346 9.851 0 60 107 Wimmer and Min(2009) # Years at War while Credit Flows 2.075 4.718 0 27 107 Wimmer and Min(2009) and Reinhart and Rogoff(2009) # Years at War while Credit Stops 2.271 5.485 0 36 107 Wimmer and Min(2009) and Reinhart and Rogoff(2009) # Years at War while Credit Flows COW 0.757 1.612 0 9 103 Sarkees and Wayman(2010) and Reinhart and Rogoff(2009) # Years at War while Credit Stops COW 0.913 2.054 0 8 103 Sarkees and Wayman(2010) and Reinhart and Rogoff(2009) Modern Census by 1820 0.093 0.292 0 1 107 coded from Goyer and Draaijer(1992 b,c,a) Modern Census by 1914 0.607 0.491 0 1 107 coded from Goyer and Draaijer(1992 b,c,a) vii First Modern Census Date 1888.963 57.413 1666 1984 107 coded from Goyer and Draaijer(1992 b,c,a) Oil Producer 0.692 0.464 0 1 107 calculated from Wimmer and Min(2009) # Years at Civil War 1816-1913 1.794 4.48 0 26 107 calculated from Wimmer and Min(2009) War Location 1816-1913 0.028 9.743 -31 58 107 calculated from Wimmer and Min(2009) Population Density in 1820 0.205 0.289 0 1.635 107 World Mapper www.worldmapper.org Great Power 0.065 0.248 0 1 107 Flandreau and Flores(2012) War Casualties 1816-1913 0.111 0.275 0 1.512 88 Dincecco and Prado(2012) ln(Rail Lines) 7.804 2.125 0 12.908 63 Comin and Hobijn(2010) Ethnic Fractionalization 0.37 0.273 0.004 0.9 106 Wimmer and Min(2009) Sea Access 36.57 35.594 0 100 107 Nunn and Puga(2012) Desert 1.862 5.016 0 26.132 107 Nunn and Puga(2012) State Antiquity 445.054 210.295 25 860.975 104 Putterman(2007) Region 2.636 1.152 1 6 107 coded by author British Colony 0.187 0.392 0 1 107 coded by author Iberian Colony 0.187 0.392 0 1 107 coded by author Other Colony 0.327 0.471 0 1 107 coded by author WWI Participant 0.374 0.486 0 1 107 coded by author Table A-2: Quasi-Direct Matches between Political Units listed in Wimmer-Min 2009 and Modern Nation-States

Original unit → Matched to space Original unit → Matched to Hanover Germany Syria (Arab Kingdom of Syria) Syria Hesse Electoral Germany Algeria (Barbary states) Algeria Hesse Grand Ducal Germany Afghanistan (Durrani Kingdom) Afghanistan Baden Germany Benin (Kingdom of Dahomey) Benin Bavaria Germany Benin Empire Benin viii Wuerttemburg Germany Argentina (United Provinces of Rio de la Plata) Argentina Saxony Germany Georgia (Kingdom of Kartli-Kakheti) Georgia Mecklenburg Schwerin Germany Madagascar (Merina Kingdom) Madagascar Modena Italy Mali (Tukulor Empire) Mali Papal States Italy Yugoslavia (Kingdom of Serbia) Serbia Tuscany Italy South African Republic South Africa Two Sicilies Italy Orange Free State South Africa Senegal (Kingdoms of Jolof and Futa Toro) Senegal Tibet China Sri Lanka (Kingdom of Kandy) Sri Lanka Transvaal South Africa Sudan (Kingdom of Sinnar) Sudan Xhosa South Africa Sudan (Mahdiyya state) Sudan Republic of Vietnam Vietnam Table A-3: Tentative Matches. These are political units listed in Wimmer-Min that cannot be directly matched to current states. They are not used in the main text analysis, but results are robust to their inclusion, as shown in column 4 in Appendix Table A-9.

Original unit Matched to Aceh Sultanate Indonesia Ashanti Kingdom Ghana Buganda Uganda Emirates of Kano Nigeria Khanate of Kiva Uzbekistan Kingdom of Bharatpur India Kingdom of Lahore Pakistan Balinese Kingdom of Lombok Indonesia Maratha Empire India Sanusi Empire Lybia Sokoto Nigeria Zulu South Africa Zulu Kingdom South Africa Ndebele Kingdom Zimbabwe Kingdom of Sindh Pakistan Kingdom of Waalo Senegal

ix B. Cross-Sectional Distribution of Warfare and Acces to Credit

First, Table A-4 reports the distribution of war and endogenous access to credit. This first sample is upper bounded by default data coverage in Reinhart and Rogoff(2009). Second, Table A-5 reports the breakdown of war participation and exogenous access to credit. This sample is larger and is upper bounded by data availability of the dependent variable. Third, the actual location of warfare is plotted in Figure A-2. Darker areas indicate higher frequency of war in territory x. Fourth, Figure A-3 plots the distribution of war participants regardless of war location. Again, darker areas indicate higher rates of participation (regardless of the location of conflict). These two figures show that most wars were fought outside Europe but involved at least one European power.

Table A-4: Endogenous access to Credit and War Participation: This table lists the # Years at War having Access to International Markets (WA), and # Years at War while being in Default (WD). N = 63

WA WD WA WD WA WD Argentina 7 9 Hungary 3 0 Poland 0 0 Australia 0 0 Iceland 0 0 Portugal 0 0 Austria 3 0 India 0 0 Romania 1 0 Belgium 1 0 Indonesia 0 0 Russia 39 2 Bolivia 10 1 Ireland 0 0 South Africa 4 0 Brazil 15 0 Italy 13 0 South Korea 0 0 Canada 0 0 Ivory Coast 0 0 Spain 6 4 Chile 3 5 Japan 5 0 Sri Lanka 2 0 China 27 0 Kenya 0 0 Sweden 0 0 Colombia 1 0 Malaysia 0 0 Switzerland 0 0 Costa Rica 0 0 Mexico 1 8 Thailand 10 0 Denmark 3 0 Morocco 5 0 Tunisia 2 0 Dominican Republic 0 0 Myanmar 6 0 Turkey 17 2 Ecuador 1 0 Netherlands 7 0 United Kingdom 58 0 Egypt 8 1 New Zealand 0 0 United States of America 5 0 El Salvador 4 0 Nicaragua 1 2 Uruguay 1 0 Finland 0 0 Nigeria 0 0 Venezuela 0 0 France 60 0 Norway 0 0 Zambia 0 0 Germany 8 0 Panama 0 0 Zimbabwe 0 0 Greece 2 1 Paraguay 7 0 Guatemala 0 3 Peru 2 11 Honduras 0 2 Philippines 0 0

x Table A-5: Exogenous access to Credit and War Participation: This table lists the # Years at War while Credit Flows (WF), and # Years at War while Credit Stops (WS). N = 107

WF WS WF WS WF WS Albania 0 0 Germany 3 5 Norway 0 0 Argentina 3 13 Greece 1 2 Pakistan 0 0 Armenia 0 0 Guatemala 2 1 Panama 0 0 Australia 0 0 Guinea 0 0 Paraguay 1 6 Austria 1 2 Honduras 2 0 Peru 6 7 Azerbaijan 0 0 Hungary 1 2 Philippines 0 0 Bangladesh 0 0 Iceland 0 0 Poland 0 0 Belarus 0 0 India 0 0 Portugal 0 0 Belgium 0 1 Indonesia 0 0 Romania 0 1 Bhutan 0 1 Iran 4 5 Russia 27 14 Bolivia 6 5 Ireland 0 0 Rwanda 0 0 Brazil 3 12 Israel 0 0 Senegal 0 2 Bulgaria 1 2 Italy 5 8 Slovakia 0 0 Burundi 0 0 Ivory Coast 0 0 Slovenia 0 0 Cambodia 4 0 Japan 4 1 South Africa 4 0 Canada 0 0 Kazakhstan 0 0 South Korea 0 0 Chad 0 0 Kenya 0 0 Spain 3 7 Chile 5 3 Latvia 0 0 Sri Lanka 2 0 China 13 14 Lebanon 0 0 Swaziland 0 0 Colombia 1 0 Lesotho 0 0 Sweden 0 0 Congo 0 0 Lithuania 0 0 Switzerland 0 0 Costa Rica 0 0 Macedonia 0 0 Tajikistan 0 0 Croatia 0 0 Madagascar 4 1 Thailand 5 5 Cyprus 0 0 Malaysia 0 0 Tunisia 2 0 Czech Republic 0 0 Mali 0 1 Turkey 9 10 Democratic Republic of the Congo 0 0 Mexico 4 5 Ukraine 0 0 Denmark 1 2 Moldova 0 0 United Kingdom 26 32 Dominican Republic 0 0 Mongolia 0 0 United States of America 3 2 Ecuador 1 0 Morocco 1 4 Uruguay 0 1 Egypt 7 2 Myanmar 4 2 Venezuela 0 0 El Salvador 3 1 Namibia 0 0 Vietnam 10 13 Estonia 0 0 Nepal 0 0 Yemen 0 0 Ethiopia 4 4 Netherlands 1 6 Zambia 0 0 Finland 0 0 New Zealand 0 0 Zimbabwe 0 0 France 24 36 Nicaragua 2 1 Georgia 9 1 Nigeria 0 0

xi Figure A-2: The Geography of Military Conﬂict in the Long-Nineteenth Century. Colors indicate the total number of years at war. Source: Wimmer and Min(2009).

(20,47] (10,20] (6,10] (3,6] [1,3] no war

xii Figure A-3: Frequency of War Participation in the Long-Nineteenth Century. Colors indicate the total number of years at war. Source: Wimmer and Min(2009).

(19,62] (10,19] (7,10] (6,7] (4,6] (2,4] [1,2] no war

xiii C. Sensibility Tests involving the Main Dependent Variable

Appendix A explains that the IMF GFS Data is augmented with various sources. Column 1 shows that when the sample is limited to countries for which there is GFS data, results hold despite that the N decreases by over 20%. Columns 2 and 3 replace average Personal Income Tax as % of GDP between 1995 and 2005 for its equivalent in the period 1990-2000, and 2000-2010 (both of which are also augmented when possible). These columns show that results are the same as in the main text, but the sample size marginally decreases. All in all, results do not hinge on small variations in the dependent variable.

Table A-6: Personal Income Tax Today (as % of GDP) as a Function of War and Endogenous Credit Access in the Long-Nineteenth Century, with Special Attention to Measurement Decisions in the Dependent Variable

(1) (2) (3) Non-Augmented Dep Variable Dep Variable Dep Variable, dated as of dated as of 1995-2005 1990-2000 2000-10

# Years at War while Credit Stops 0.264*** 0.184** 0.196*** (0.076) (0.081) (0.059) # Years at War while Credit Flows -0.251*** -0.181** -0.220*** (0.071) (0.075) (0.071) Population Density in 1820 0.737 2.267 1.031 (1.627) (1.572) (1.474) Oil Producer 0.012 0.180 0.156 (0.535) (0.709) (0.420) Sea Access 0.029*** 0.033*** 0.026*** (0.008) (0.010) (0.007) Desert Territory -0.060 -0.043 0.021 (0.059) (0.058) (0.039) Great Power 2.657** 2.090 3.016** (1.250) (1.329) (1.152) Modern Census by 1820 0.565 0.265 1.325 (1.345) (1.451) (1.347) Intercept 1.495 0.465 1.436* (0.925) (1.168) (0.734) Region FE Yes Yes Yes Colonial Origins FE Yes Yes Yes Observations 87 83 104 R-squared 0.658 0.602 0.634 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

xiv D. Inﬂuence of Outliers

Based on the partial correlation Figures2 and4, I identify potential outliers in each relationship: Peru, Argentina, Mexico and Prussia, in the case of endogenous access to credit, and Russia, Georgia, and France, in the case of exogenous access to credit. Table A-7 reports models without these outliers, and based on the new estimates, I rerun the partial correlation plots, shown in Figures A-4 and A-5. Both sets of results suggest that outliers ˆ downward bias the main coefficient of interest β1, especially in the case of endogenous access ˆ to credit. By contrast, β2 turns positive in the endogenous model (although it is barely ˆ significant and the magnitude of this coefficient is one order of magnitude smaller than β1) and turns non-significant in the exogenous access to credit sample, consistent with other results in that section. Results are similar if outliers are identified by standard measures of influence: e.g. Cook’s distance. Altogether, this exercise confirms the positive effect of war fought in the absence of external credit, and the null (at best mixed) effect of war fought while having access to external credit.

xv Table A-7: Checking for Inﬂuential Outliers. PIT as % of GDP Today as a Function of War and Exogenous/Exogenous Access to Credit in the Long Nineteenth Century once Outliers are excluded.

Dependent Variable: PIT as % of GDP Endogenous Access Exogenous Access To Credita To Creditb (1) (2)

# Years at War while in Default 0.517** (0.230) # Years at War with Access to Credit 0.041* (0.023) # Years at War while Credit Stops 0.248*** (0.090) # Years at War while Credit Flows -0.109 (0.150) # Years in Default -0.015 (0.014) Population Density in 1820 2.583 0.712 (1.761) (1.391) Oil Producer -0.861 -0.053 (0.670) (0.450) Sea Access 0.020** 0.029*** (0.008) (0.007) Desert -0.090 0.010 (0.061) (0.046) Great Power 1.047 3.148*** (1.438) (1.183) Intercept 3.402** 1.037 (1.427) (0.829) Region FE Yes Yes Colonial Origins FE Yes Yes Observations 59 103 R-squared 0.788 0.610 a Models with Endogenous Credit Access exclude Argentina, Mexico, Peru and Russia. b Models of Exogenous Credit Access exclude France, Georgia, and Russia. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

xvi Figure A-4: Partial Correlations of Personal Income Tax and Endogenous War-Financing once Outliers are dropped: Argentina, Mexico, Peru, Russia. These estimates are drawn from column 1 in Appendix Table A-7, and should be compared to those in Figure2.

Denmark Belgium DenmarkBelgium South Africa 4 South Africa 4 Zimbabwe Turkey Zimbabwe Turkey

Italy Chile 2 Austria 2 Brazil Italy Thailand ThailandIndonesia Venezuela Brazil Finland Austria IndonesiaColombiaFinland United Kingdom Venezuela New Zealand Spain Ecuador Colombia United States of America ChileTunisiaKenyaUnitedNew StatesZealand of America IrelandHondurasBolivia Ireland China Ecuador Kenya HondurasIcelandBoliviaParaguay France 0 Zambia Tunisia Malaysia Guatemala 0 IcelandPhilippinesUnitedUruguay Kingdom MalaysiaPortugalUruguay Portugal ParaguayHungary ZambiaNetherlandsMyanmar NetherlandsMyanmarCanada Nicaragua Philippines AustraliaCanadaGreeceSpain IndiaAustralia India ChinaGreece Japan JapanGuatemalaDominicanNicaraguaEl Salvador Republic El SalvadorDominicanPanamaNorwayPoland RepublicMorocco Hungary NorwayPolandMorocco France PanamaCosta Rica Costa Rica SwedenSri LankaSouth Korea SwedenSri Lanka -2

-2 South Korea Residuals from Regressing Residuals Ivory Coast from Regressing Residuals Egypt PIT as % of GDP Controls on Switzerland PIT as % of GDP Controls on Switzerland EgyptGermany Ivory Coast Nigeria Nigeria Romania Romania Germany -4 -4 -1 0 1 2 3 -20 0 20 40 Residuals from Regressing Residuals from Regressing # Years at War while in Default on Controls # Years at War while Having Access to Credit on Controls

coef = .516, (robust) se = .230, t = 2.25 coef = .041, (robust) se = .023, t = 1.75 (a) War in Default (b) War while having Credit Access

Figure A-5: Partial Correlations of Personal Income Tax and Exogenous War-Financing once Outliers are dropped: Russia, Georgia, and Mexico. These estimates are drawn from column 2 in Appendix Table A-7, and should be compared to those in Figure4. 10 10

Belgium Belgium Namibia Namibia Denmark Denmark

5 Israel 5 Zimbabwe IsraelZimbabwe FinlandTurkey Finland South Africa South Africa Turkey Lesotho Lesotho Spain IndonesiaBhutan Italy BhutanIndonesiaAustria AustriaThailandYemenIreland Netherlands Brazil Yemen Nepal Thailand NepalPortugalSwazilandIranIceland Spain PortugalIrelandItalySwazilandUnited States Chileof America Chile UnitedPolandNewPeruUruguay StatesZealandZambia of America Paraguay Argentina UruguayVenezuelaIranLithuaniaIcelandPolandNewColombia ZealandPeru VenezuelaLithuaniaMacedonia Netherlands ZambiaMongolia Macedonia Japan Bolivia ColombiaBolivia Mongolia Morocco Brazil ParaguayPhilippines SwitzerlandTunisiaEcuadorHonduras 0 Japan HondurasSwitzerlandTunisiaPhilippinesKenya 0 SouthKenya Korea GuatemalaNicaragua EcuadorNicaraguaSouthNorway GreeceKorea Morocco GreeceHungarySwedenNorwaySlovakia Cambodia GuatemalaHungarySlovakiaSwedenEstoniaSloveniaMexicoChina Argentina Ivory CoastSloveniaEstoniaArmeniaCzechMexico Republic Cambodia AzerbaijanCzechArmeniaIvoryEthiopia Republic CoastChadCyprus ChadCyprusCongoCroatiaBurundiAzerbaijanLatvia Ethiopia CroatiaLatviaCongoRwandaBurundiRomaniaIndiaAustraliaCanadaSenegalMali IndiaCanadaAustraliaDemocraticPanamaRomaniaCostaRwanda RepublicRica El of Salvador the Congo China ElPanama SalvadorCostaDemocratic RicaPakistanMalaysia Republic of the Congo MaliPakistanMalaysiaSenegal KazakhstanTajikistan Myanmar Residuals from Regressing Residuals MyanmarTajikistanKazakhstanGuineaBulgaria from Regressing Residuals GuineaNigeriaBulgaria PIT as % of GDP Controls on Dominican Republic PIT as % of GDP Controls on Dominican Republic NigeriaLebanon Lebanon BosniaMoldovaSri and Lanka Herzegovina MoldovaSriBosnia LankaGermany and HerzegovinaVietnam Belarus BelarusUkraine GermanyUkraine Bangladesh Bangladesh Madagascar Vietnam MadagascarEgypt Egypt Albania Albania -5 -5 -5 0 5 10 -4 -2 0 2 4 Residuals from Regressing Residuals from Regressing # Years at War while Credit Stops on Controls # Years at War while Credit Flows on Controls

coef = .248, (robust) se = .089, t = 2.78 coef = -.108, (robust) se = .150, t = -.72 (a) War while Credit Stops (b) War while Credit Flows

xvii E. Inﬂuence of Fixed Eﬀects

Region and colonial origins fixed effects (6 and 4 categories, respectively) are intended to minimize unobserved heterogeneity across countries. However, if covariates are highly correlated within region/colonial origins, adding fixed effects might induce high multicollinearity and outliers. Based on the simplest specification of the exogenous access to credit model, I stepwise drop region fixed effects and colonial origins fixed effects.

Table A-8: Personal Income Tax Today (as % of GDP) as a Function of War and Endogenous Credit Access in the Long-Nineteenth Century, with Special Attention to the Inﬂuence of Fixed Eﬀects

(1) (2) (3)

# Years at War while Credit Stops 0.227*** 0.283*** 0.157* (0.056) (0.068) (0.092) # Years at War while Credit Flows -0.181*** -0.265*** -0.185** (0.060) (0.077) (0.082) Population Density in 1820 1.335 0.511 1.466 (1.386) (1.545) (1.539) Oil Producer 0.214 0.851 0.784 (0.508) (0.521) (0.615) Sea Access 0.031*** 0.020** 0.020** (0.007) (0.009) (0.010) Desert Territory 0.012 0.018 0.056 (0.046) (0.055) (0.057) Intercept 2.290*** 1.101* 1.310** (0.781) (0.615) (0.605) Region FE Yes No No Colonial Origins FE No Yes No Observations 106 106 106 R-squared 0.533 0.298 0.118 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

xviii F. Sensibility Tests addressing Measurement and Sample Decisions

Table A-9 investigates the extent to which results hinge on particular cases, regions, assumption about error correlation, matching decisions, or territorial configuration of the state. The bellicist hypothesis has been tested in Europe with positive result. More importantly, European countries are overrepresented in inter-state warfare. Column 1 in Table A-9 drops all European countries from the sample to assess their influence. Great Powers (Austria, France, Germany, Hungary, Italy, Russia and the United King- dom) follow distinct trajectories from the rest of the world: they developed domestic financial markets early enough and also had high military capacity. For these reasons too, they are more prone to participate in war. Column 2 drops them from the sample. Arguably, wars happening in country x might affect the likelihood of war-making among its neighbors. To account for contemporaneous error, Column 3 fits clustered standard errors to address residual correlation at the regional level. I consider 6 regions in the world. Thus, cluster standard errors create smaller standard errors than the Huber-White standard error counterparts, which are used in the main text. Appendix Table A-3 lists past polities that cannot be matched with current state-borders without making various assumptions. The analyses in the main text do not consider these polities, but column 4 in Table A-9 does in order to minimize any potential attrition bias. Results hold. A federal structure might condition central government tax yields as well as correlate with past warfare if a non-unitary state results from ethnical or ideological civil war. Column 5 includes this variable as a control. Data for federal structure is drawn from Treisman (2014).

xix Table A-9: Personal Income Tax Today (as % of GDP) as a Function of War and Endogenous Credit Access in the Long-Nineteenth Century, with Special Attention to Measurement and Sample Decisions

(1) (2) (3) (4) (5) Europe Great Power Cluster Tentative Federal Dropped Dropped Std. Error Matches Structure # Years at War while Credit Stops 0.190** 0.260*** 0.233*** 0.233*** 0.237*** (0.075) (0.082) (0.041) (0.062) (0.063) # Years at War while Credit Flows -0.140** -0.146 -0.247*** -0.247*** -0.252*** (0.054) (0.116) (0.058) (0.068) (0.066) Population Density in 1820 -0.952 0.800 0.770 0.770 0.802 (0.742) (1.433) (1.850) (1.425) (1.423) Oil Producer -0.005 -0.053 0.146 0.146 0.183 (0.422) (0.441) (0.387) (0.463) (0.475) Sea Access 0.016** 0.026*** 0.027** 0.027*** 0.026*** (0.006) (0.007) (0.008) (0.007) (0.007) Desert Territory 0.011 0.009 0.012 0.012 0.012 (0.049) (0.046) (0.027) (0.046) (0.046) Great Power 2.814*** 2.814** 2.898** (0.252) (1.195) (1.283) Modern Census by 1820 0.108 1.146 0.801 0.801 0.761 (1.820) (1.306) (0.524) (1.239) (1.232) Federal Structure -0.315 (0.793) Intercept 1.380* 1.105 1.339 1.339 1.371 (0.763) (0.838) (0.669) (0.832) (0.862) Region FE Yes Yes Yes Yes Yes Colonial Origins FE Yes Yes Yes Yes Yes Observations 88 100 106 106 106 R-squared 0.560 0.597 0.613 0.613 0.614 In column 1, Great Power is dropped because all of them were European. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

xx G. Sensibility Test involving Crises Timing and Length of Sudden-

Stops

The 1910 crisis is a stock-market crash, not a banking crash. Based on Figure3, the stock-market crash might not cause comparable capital dry shocks. Accordingly, column 1 in Table A-10 treats the 1910 stock-market crisis as a non-crisis, and investigates whether this has any impact. Reinhart and Rogoff(2009) do not list the 1873 banking crisis for Great Britain, despite it being a major crisis in the nineteenth century (Kindleberger and Aliber, 2005). Technically, the 1873 crisis originated in Austria and Germany. But, it was only a matter of months that the crisis reached London, causing a sudden-stop of credit (Bordo 1986), as Figure3 reflects. Based on the relevance of this crisis, I include it in the main analysis. For the sake of robustness, column 2 excludes the 1873 banking crisis as a cause of sudden-stop. Columns 3 and 4 allow for longer spells of sudden-stops. Specifically, columns 3 and 4 replace the four-year rule of credit stop based on Catao(2006) for five and six years spells, respectively. The effect of fighting war during these longer periods is still positive, although, consistently with the actual length of sudden-stops, the effect becomes weaker.

xxi Table A-10: Personal Income Tax Today (as % of GDP) as a Function of War and Endogenous Credit Access in the Long-Nineteenth Century, with Special Attention to Timing Issues

(1) (2) (3) (4) 1910 Crisis 1873 Crisis 5-year Sudden-Stop 6-year Sudden-Stop Dropped Dropped Windows Windows

# Years at War while Credit Stops 0.189** 0.243*** 0.170*** 0.159*** (0.077) (0.073) (0.056) (0.054) # Years at War while Credit Flows -0.175* -0.213*** -0.248*** -0.305*** (0.102) (0.081) (0.080) (0.087) Population density in 1820 0.853 1.319 0.780 0.765 (1.419) (1.450) (1.437) (1.423) Oil Producer 0.178 0.149 0.168 0.185 (0.464) (0.469) (0.462) (0.461) Sea Access 0.028*** 0.028*** 0.026*** 0.026*** (0.008) (0.007) (0.007) (0.007) Desert Territory 0.007 0.016 0.014 0.012 (0.046) (0.047) (0.046) (0.046) Great Power 2.702** 2.076 2.774** 2.681** (1.305) (1.370) (1.169) (1.116) Modern Census by 1820 0.782 1.296 0.809 0.801 (1.255) (1.236) (1.273) (1.277) Intercept 1.339 1.368 1.330 1.333 (0.844) (0.837) (0.840) (0.834) Region FE Yes Yes Yes Yes Colonial Origins FE Yes Yes Yes Yes Observations 106 106 106 106 R-squared 0.596 0.606 0.607 0.612 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

xxii H. Alternative Sources of War Financing

There are three additional ways to finance war: domestic borrowing, printing money, and financial repression. The former strategy should not be of major concern: Domestic borrowing requires a developed financial market, something that, in the period under consideration, was only guaranteed in a few European countries (Reinhart and Rogoff, 2009, ch.7). The pool of domestic investors in the periphery tended to be small, and loans to government represented a large share of their portfolio. This implied expensive credit relative to other options overseas (della Paolera and Taylor, 2013; Flandreau and Flores, 2012).50 Not surprisingly, countries in the periphery resorted to international markets for financing. Columns 1-3 in Table A-11 adress the possibility of fighting wars while having access to either domestic or external credit and the lack thereof. The first row shows the coefficient of not having access to the domestic or international markets (i.e. domestic and external default), while the fourth row shows the effect of having access to either to the domestic or international markets. In the former case, I expect the incentives to invest in fiscal institutions to be maximum. Consistent with this expectation, the magnitude of the coefficients grows with respect to those reported in Table1. Column 2 adds a Great Power indicator to control for differences in domestic credit markets, and column 3 controls for the War Location, as it could influence the capacity to mobilize resources domestically. The point ˆ ˆ estimate of the two coefficients of interest, β1 and β2, remain fairly stable. A second means to financing war is expanding the money supply (also known as printing money). Except as an extreme measure of last resort, printing money has came to occupy a “subordinate position” in pre-1913 war finance (Sprague, 1917). The reason is that expanding the money supply has inflationary consequences. A sudden expansion of the money supply gives the government a temporary relief with which to purchase additional

50An example might be illustrative: domestic lenders in Mexico would apply rates in the range of 300%- 500% (Centeno, 2002).

xxiii Table A-11: PIT as % of GDP Today as a Function of War and Endogenous Credit Access in the Long Nineteenth Century, with Special Attention to Domestic Default Episodes and Money Printing

Accounting for Accounting for Domestic Default Money Printing (1) (2) (3) (3) (4) (5) (6) # Years at war while in external and domestic default 0.171** 0.172** 0.187** (0.073) (0.075) (0.076) # Years at war while in external default but no money printing 0.171* 0.172* 0.203** (0.092) (0.094) (0.097) # Years at war while in external default and money printing 0.154*** 0.157*** 0.142** (0.055) (0.056) (0.057) # Years at war with access to credita 0.032 0.028 -0.005 0.032 0.028 -0.006 (0.026) (0.027) (0.052) (0.026) (0.028) (0.053) Population Density in 1820 3.421** 3.377* 3.383** 3.422** 3.380* 3.389** (1.655) (1.723) (1.591) (1.673) (1.741) (1.610) xxiv Oil Producer -0.917 -0.926 -1.007 -0.919 -0.927 -1.016 (0.671) (0.685) (0.709) (0.680) (0.695) (0.722) Sea Access 0.022** 0.022** 0.020** 0.022** 0.022** 0.020** (0.008) (0.008) (0.009) (0.008) (0.008) (0.010) Desert Territory -0.060 -0.059 -0.062 -0.061 -0.059 -0.063 (0.060) (0.060) (0.065) (0.060) (0.061) (0.066) # Years in defaultb -0.009 -0.008 -0.017 -0.009 -0.008 -0.017 (0.013) (0.014) (0.014) (0.013) (0.014) (0.014) Great Power 0.332 0.314 (1.551) (1.567) War Location 1816-1913 0.051 0.053 (0.058) (0.060) Intercept 3.464** 3.423** 3.517** 3.497** 3.459** 3.574** (1.403) (1.434) (1.373) (1.433) (1.464) (1.403) Colonial Origins FE Yes Yes Yes Yes Yes Yes Region FE Yes Yes Yes Yes Yes Yes Observations 63 63 63 63 63 63 R-squared 0.761 0.761 0.767 0.760 0.761 0.767 a In columns 1-3, access to credit refers to either domestic or international markets, or both. b Years in default refer to external default. Robust Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 weapons or pay bills, but this gain is rapidly dissipated by the costs of inflation (Rockoff 1998, Schumpeter 1938). Nevertheless, it is worth checking what is the effect of printing money on long-term fiscal capacity. In the absence of direct data of instances of money printing, I use episodes of inflationary crises, as coded by Reinhart and Rogoff(2009), to explore how inflation weakens the incentives to invest in fiscal capacity while being at war despite not having access to external credit. In other words, this test assumes that inflationary crises are related to episodes of money supply expansions. Inflation does not dissipate soon. To account for these lags, I add four leads to the onset of an inflationary crisis. Suppose, for instance, that the onset of an inflation crisis dates as of 1900 in country x. Then, I assume that inflation stays around until 1904, five years in total. Based on that, I estimate the effect of being at war and in default in the presence and absence of an inflationary crisis. Again, I expect inflationary crisis (i.e. the proxy of money printing) to dissipate the incentives to invest in fiscal capacity despite being in default. The results in columns 4-6 in Table A-11 reinforce and qualify earlier findings. First, they confirm that waging war while being in default is related to higher fiscal capacity in the long-run regardless of money printing: both coefficients are positive. However, based on the coefficients’ magnitude, if inflation is kept under control (i.e. the ruler does not print money), fiscal capacity might be even higher in the long-run. This result implies that incumbents that are not tempted to print money while being at war and in default are those investing more decisively in the fiscal capacity of the state, holding everything else constant. A third way to finance war is financial repression. Calomiris and Haber (2014) and Reinhart (2012) show that, if anything, financial repression is a substitute of fiscal capacity building.51 I lack systematic data about instances of financial repression, and cannot test this proposition here. However, if fiscal repression is negatively correlated to efforts of building tax capacity, the omission of financial repression, if any, biases downwards the main coefficient

51See Menaldo (2016) for recent evidence about it.

xxv of interest, β1. That is, if rulers prioritize fiscal repression when they lack access to external finance, we should not expect a positive coefficient for the # Years at War while in Default, precisely because fiscal repression is implemented as to avoid fiscal capacity building. The same applies to alternative ways to finance war, such as office selling (Hoffman, 1994).

xxvi I. Value-Added Tax as Outcome Variable

Column 1 regresses average VAT revenue as percentage of GDP between 1995 and 2005 on the benchmark regressors. VAT data is drawn from IMF Government Financial Statis- tics. The sample size is limited by data availability. We can extend the dependent variable by replacing missing values for those reported in USAID Fiscal Reform and Economic Gov- ernance Project, 2004-10, as we did with PIT data. Results with the extended dependent variable are reported in column 2. Columns 3 and 4 adds controls for initial state capacity.

xxvii Table A-12: Value-Added Tax (VAT) as % of GDP Today as a Function of Years at War and Exogenous Access to External Credit in the Long Nineteenth Century

(1) (2) (3) (4) # Years at War while Credit Stops 0.229* 0.097 0.126** 0.097* (0.124) (0.059) (0.060) (0.057) # Years at War while Credit Flows 0.065 0.047 0.040 0.037 (0.104) (0.079) (0.081) (0.077) Population Density in 1820 0.326 -0.260 -0.237 -0.371 (1.098) (0.784) (0.778) (0.839) Oil Producer -1.165 -1.018 -1.042 -1.188 (0.761) (0.684) (0.697) (0.733) Sea Access 0.005 0.008 0.011 0.008 (0.013) (0.008) (0.009) (0.008) Dessert Territory 0.097* 0.029 0.034 0.022 (0.051) (0.054) (0.054) (0.058) Great Power -3.416** -0.420 -0.574 -0.309 (1.355) (1.375) (1.417) (1.364) Modern Census by 1820 -1.223 (0.896) State Antiquity 0.000 (0.002) Intercept 1.285 2.207** 2.112** 2.202** (1.182) (0.861) (0.845) (0.958) Extended Dep Var No Yes Yes Yes Region FE Yes Yes Yes Yes Colonial Origins FE Yes Yes Yes Yes Observations 65 105 105 102 R-squared 0.439 0.388 0.394 0.381 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

xxviii J. Military Alliances, British Colonies, and British Wars

Appendix Table A-13 examines the eﬀect of (1) military alliances in the international system, (2) the eﬀect of being a British colony, and (3) British active participation at war. Do results change when we account for these potential confounders?

1. Accounting for Military Alliances. Military alliances might change the incentives to go to war and also facilitate access to external credit. To account for this, I control for military alliance with the four credit capitals in the long-nineteenth century: the British, the French, the German, and the USA. Despite having uneven weight in global finances (refer to Table2), one might argue that any of these four countries had both the capacity to finance third countries and coordinate military interventions with them. To code military alliances, I rely on Gibler (2009). This dataset offers dyads of military alliances between independent countries since 1816. Some of these alliances were short-lived while others were enduring. To account for this heterogeneity, I compute the share of years between 1816-1913 under which a given country has any form of military alliance (defense, neutrality, non-aggression, and entente) with any of these credit capitals. For instance, Portugal had at least one military alliance with Britain for the whole period. Accordingly, it takes the maximum value: 100%. Other countries (e.g. Belgium) stroke no military alliance with Britain during the long- nineteenth century. Accordingly, the value for Belgium is zero. Since the total number of years in sample are 98, these shares may be interpreted as the total number of years under which a given country had a military alliance with any of the four financial capitals. Importantly, colonies are assumed to have military alliances by default, and this should be picked up by the colonial origins fixed effect.

2. Excluding British Colonies. It is proved that British colonies had access to external credit in more favorable conditions than other colonies (Accominotti, Flandreau and Rezzik, 2011). Since Britain is the credit capital and the military superpower of the long-nineteenth

xxix century, one might suspect that the decision to go to war for British colonies is different from other countries’. The British colonial origins fixed effect might not address this source of heterogeneity well enough. To address this issue, columns 3 and 4 re-run Expression1 excluding all British colonies and using the exogenous variation of credit access.

3. Excluding Wars Fought by Britain. Having already addressed strategic considera- tions with respect to British colonies, we might wonder whether wars involving British direct participation are comparable to other wars. To address this issue, columns 5 and 6 report models excluding all wars in which the British explicitly participated.

xxx Table A-13: PIT as % of GDP Today as a Function of War and Exogenous Access to Credit in the Long Nineteenth Century, with Special Attention Military Alliances, Favorable Access to Credit by British Colonies, and Wars Involving British Participation.

All Countries Included British Colonies Excluded British Wars Excluded (1) (2) (3) (4) (5) (6) # Years at War while Credit Stops 0.289*** 0.298*** 0.207*** 0.197*** 0.332*** 0.339*** (0.076) (0.061) (0.051) (0.037) (0.080) (0.072) # Years at War while Credit Flows -0.298*** -0.290*** -0.236*** -0.238*** -0.320*** -0.313*** (0.089) (0.087) (0.049) (0.042) (0.055) (0.053) Population Density in 1820 0.753 0.648 2.395 1.928 0.741 0.699 (1.462) (1.415) (1.823) (1.619) (1.408) (1.374) Oil Producer 0.007 0.025 0.228 -0.009 0.084 0.118 (0.477) (0.454) (0.464) (0.419) (0.454) (0.433) Sea Access 0.025*** 0.029*** 0.025*** 0.029*** 0.027*** 0.030*** (0.007) (0.007) (0.008) (0.007) (0.007) (0.007) Desert Territory 0.017 -0.010 0.081 0.042 0.015 -0.013 (0.049) (0.034) (0.050) (0.027) (0.046) (0.032) Alliance with Britain 0.001 -0.000 xxxi (0.007) (0.008) Alliance with France 0.158** 0.138** (0.068) (0.068) Alliance with Germany -0.011 -0.007 (0.020) (0.021) Alliance with USA 0.630 0.811 (0.878) (0.625) Great Power 0.823 0.806 2.503** 2.525** 2.789** 2.665** (1.209) (1.155) (1.246) (1.115) (1.161) (1.115) Modern Census by 1820 0.895 0.699 0.752 (1.558) (1.302) (1.213) State Antiquity 0.001 0.004*** 0.001 (0.001) (0.001) (0.001) Intercept 1.201 0.468 0.471 -1.624* 1.307 0.644 (0.835) (0.975) (0.820) (0.835) (0.840) (0.986) Colonial Origins FE Yes Yes Yes Yes Yes Yes Region FE Yes Yes Yes Yes Yes Yes Observations 106 103 86 83 106 103 R-squared 0.635 0.668 0.557 0.656 0.625 0.658 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 K. Instrumenting for War-Making

This section addresses the endogeneity of war in a reduced-form framework. In analyzing the effect of war in Europe, Gennaioli and Voth(2015) instrument war frequency of country i based on war participation by adjacent countries against third countries. The logic behind this instrument is that contextual circumstances that lead neighboring countries to war might increase the likelihood of country i going to war against a third country. The exclusion restriction is that there is no effect of war in neighboring countries on fiscal capacity that is not the result of the risk of war (ibid.). Here I follow a similar strategy. However, instead of running a pure IV model, I stick to a reduced-form set up, in which I replace inter-state war fought by country i while credit stops (flows) for inter-state wars fought by immediately adjacent neighbors while credit stops (flows). For reference, wars of i against adjacent countries are excluded to maximize exogeneity. I fit the reduced-form instead of a fully-fledged IV model with two endogenous variables (# of years at war while credit dries, and # of years at war while credit flows), as that requires too many untestable assumptions. A ratio of the two endogenous variables would simplify everything, except that there are zeros in both variables, leading to indeterminate form (0/0) or infinite values (n/0) in the key explanatory variable. In light of both limitations, I opt for the reduced-form model, in which I replace the two potentially endogenous variables for their two instruments. Accordingly, Expression1 becomes

PITi,1995−2005 = α + β1(#years at war by i’s-adjacent neighbors between 1816-1913 | external lending stops)

+β2(#years at war by i’s-adjacent neighbors between 1816-1913 | external lending ﬂows)

+Xiδ + γ + ρ + i (2)

where controls and ﬁxed eﬀect batteries remain the same, and common episodes of suddens- stops are used to exogenize access to external credit. In Gennaioli and Voth(2015) all countries have adjacent neighbors. However, some

xxxii islands in my sample have no adjacent neighbor whatsoever: Australia, Iceland, Madagascar, Philippines, and New Zealand. Column 1 shows the result for every country except these cases. The exclusion restriction requires the instrument not to be directly related with the outcome or unobservables affecting the outcome. The latter assumption can be addressed to the best extent by controlling for further covariates. Accordingly, column 2 includes all controls for which I have full data. Columns 3 and 4 rerun the same tests including islands. ˆ ˆ In every model, the coefficients of interest, β1 and β2, hold the expected sign: that is, the instrumented-version of waging war while having access to external credit does not increase (nor decrease) fiscal capacity today, whereas the instrumented-version of waging war while not having access to external loans is associated with higher fiscal capacity today. The main difference with Table7 results is the size of the effects: these are now attenuated as a result of the imperfect match between war-making by country i and its adjacent countries.

xxxiii Table A-14: Reduced-Form Model of War. Personal Income Tax as % of GDP Today as a Function of War and Exogenous Access to Credit in the Long Nineteenth Century, with War Participation of Country i Instrumented by War Participation by Adjacent Countries

(1) (2) (3) (4) years at war by i’s-adjacent neighbors between 1816-1913 while external lending stops 0.112* 0.119* 0.111* 0.112* (0.062) (0.067) (0.062) (0.067) years at war by i’s-adjacent neighbors between 1816-1913 while external lending ﬂows -0.069 -0.079* -0.071 -0.074 (0.044) (0.045) (0.044) (0.045) Population Density in 1820 1.201 0.310 1.129 0.473 (1.129) (1.375) (1.128) (1.379) Oil Producer 0.232 -0.142 0.194 -0.081 (0.526) (0.511) (0.493) (0.492) Sea Access 0.030*** 0.029*** 0.029*** 0.029*** (0.009) (0.009) (0.008) (0.008) Desert Territory -0.015 -0.047 -0.004 -0.043 (0.047) (0.033) (0.045) (0.034)

xxxiv War Location 0.052 0.056 (0.052) (0.052) Great Power 1.446 1.435 (1.519) (1.530) Modern Census by 1820 1.050 1.041 (1.478) (1.496) State Antiquity 0.000 0.000 (0.002) (0.001) Ethnic Fractionalization -0.827 -0.464 (1.226) (1.181) # Years at Civil War 1816-1913 0.073 0.070 (0.055) (0.053) Intercept 1.841** 1.530 1.523* 1.075 (0.877) (1.380) (0.867) (1.384) Islands Included No No Yes Yes Region FE Yes Yes Yes Yes Colonial Origins FE Yes Yes Yes Yes Observations 101 98 106 102 R-squared 0.445 0.564 0.556 0.652 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 L.Including Bad Controls

Covariates that result from treatment are known as bad controls. Their inclusion in empirical models bias the estimate of interest, in this case β1 and β2. This is also known as post-treatment bias. Here I consider four potential bad controls: democracy, preferences for redistribution, GDP per capita, and trade openness. Bates and Lien(1985) claim that democratic institutions result from tax-financed war participation. Scheve and Stasavage (2010) suggest that preferences for the size of government is endogenous to war participation. Dincecco and Prado(2012) show that long-term GDP is a function of participation in war in the past. Queralt(2015) suggest that trade openness follows fiscal capacity building, which is here argued to result from war participation. Table A-15 corroborates that the inclusion ˆ ˆ of bad controls impact the size of the coefficients of interest. Still, both β1 and β2 hold the expected sign and achieve statistical significance within conventional levels.

xxxv Table A-15: Models of PIT as % of GDP Today as a Function of Exogenous Credit Access and War-Making in the Long Nineteenth Century including Bad Controls.

(1) (2) (3) (4) # Years at War while Credit Stops 0.238*** 0.225*** 0.147*** 0.232*** (0.059) (0.061) (0.054) (0.063) # Years at War while Credit Flows -0.230*** -0.238*** -0.139* -0.243*** (0.069) (0.068) (0.075) (0.070) Population Density in 1820 0.419 0.676 0.711 0.830 (1.444) (1.419) (1.090) (1.482) Oil Producer 0.280 0.008 -0.386 0.171 (0.494) (0.505) (0.405) (0.471) Sea Access 0.022*** 0.024*** 0.010 0.027*** (0.007) (0.007) (0.007) (0.007) Desert Territory 0.009 0.026 -0.013 0.014 (0.049) (0.051) (0.036) (0.047) Modern Census by 1820 0.337 0.828 -0.060 0.838 (1.347) (1.228) (1.042) (1.251) Great Power 2.133* 2.685** 1.331 2.809** (1.215) (1.190) (1.222) (1.208) Polity IV Score 1995-2005 0.134*** (0.051) Government Size 1995-2005 -4.540* (2.427) ln(Per Capita GDP) 1995-2005 1.136*** (0.217) Trade Openness 1995-2005 0.003 (0.007) Intercept 1.631* 2.091** -5.095*** 1.133 (0.826) (1.023) (1.402) (1.019) Region FE Yes Yes Yes Yes Colonial Origins FE Yes Yes Yes Yes Observations 104 104 106 106 R-squared 0.641 0.623 0.732 0.613 Sources of bad controls: Democracy: Marshall and Jaggers (2000); Per Capita GDP and Trade Openness: World Bank Indicators; Government Size: Feenstra et al. (2013). Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

xxxvi M. Further Evidence of Exogeneity of Sudden-stops

Table4 suggests that the frequency and length of war in and outside sudden-stop periods are virtually identical (or balanced). Figure A-6 shows this diﬀerently. In particular, I plot the Total Number of Wars per Year in the sample, and identify the onset of sudden-stops. Financial crises that begin within four years of the last sudden-stop (the average duration) are not plotted. If sudden-stops are anticipated, we should observe a systematic increase in the frequency of war just preceding the ﬁnancial collapse onset. However, this is not the case. Wars take place before and after sudden-stops, almost evenly, as Table4 descriptive statistics shows.

Figure A-6: Total Number of Wars per Year vs. Sudden-Stops Onset (vertical line) 15 10 # # wars 5 0 1815 1835 1855 1875 1895 1915

xxxvii N. Details of the Lending Frenzy

Although a full characterization of the lending frenzy goes beyond the possibilities of this paper, one can easily elucidate the favorable terms of credit faced by countries in the periphery twofold: by comparing their bond yields with those paid by the European powers in the nineteenth century, and with those paid by the latter in pre-modern times, when their state capacity was still developing. First, between 1850 and 1914, the largest Latin American countries barely paid a 2% pre- mium relative to the European core (Lindert and Morton, 1989), despite their radically different levels of institutional consolidation. Similarly, colonies borrowed at similar prices than their metropolises, despite having entirely different economic fundamentals (Accominotti et al. 2011, Ferguson and Schularick 2006). Second, European powers paid higher interests in pre-modern times than countries in the periphery in the nineteenth century. The critical period of European state formation goes from the fifteenth to the seventeenth century Tilly(1990, p.81). This is a period in which royal power begins to reassert itself, monopolize violence, and settle the first permanent systems of tax collection at a national-scale, which matches to a great extent the challenges faced by the rulers of the newly created states in the periphery. The average nominal yield in the 15th-17th century in Castile, France and, the UK were 8.75, 7.25, and 7.78, respectively (calculations based on Stasavage 2011). These are actually conservative estimates: Homer and Sylla(2005, Table 8) show that bond yields could be significantly higher than these, reaching rates as high as of 100%. In stark constrast, in Latin American, only Honduras and Paraguay paid higher yields than these in the nineteenth century (Marichal, 1989, Appendix A and B). Specifically, by the turn of the century, no Latin American economy paid nominal interests above 6% (ibid.). Arguably, the underwriters played a crucial role too: in return for low interest rates, countries in the periphery would grant financial intermediaries monopoly over sovereign bond trading, which would sell in secondary markets at higher rates. Country-specific examples

xxxviii of this exchange can be found in: Flandreau and Flores(2012) for Brazil, Suzuki (1994) for Japan, and Weller (2015) for Porfirian Mexico. All in all, despite common challenges, countries in the periphery were treated in a more generous way by international markets than their European counterparts had been centuries before. This is due to the very different international context in which states were created. The European countries were built in times in which the financial markets were underde- veloped and oligopolistic, whereas states in the periphery were created in times of financial boom and cheap credit caused by excess savings in the European core resulting from the industrial revolution.

xxxix O. Appendix-Speciﬁc References

Balcells, Laia and Stathis N. Kalyvas. 2014. “Does Warfare Matter? Severity, Duration, and Outcomes of Civil Wars.” Journal of Conﬂict Resolution 58(8): 1390-1418

Bordo, Michael D. 1986. “Financial Crises, Banking Crises, Stock Market Crashes and the Money Supply: Some International Evidence, 1870-1933.” In Financial Crises and the World Banking System, ed. Forrest Capie and Geoﬀrey E. Wood. London: MacMillan.

C´ardenas,Mauricio. 2010. “State Capacity in Latin America.” Econom´ıa10(2):1-45.

Feenstra, Robert C., Robert Inklaar and Marcel P. Timmer. 2013. “The Next Generation of the Penn World Table.” URL: www.ggdc.net/pwt

Ferguson, Nial and Moritz Schularick. 2006. “The Empire Eﬀect: The Determinants of Country Risk in the First Age of Globalization, 1880-1913.” The Journal of Economic His- tory 66(2):283-312.

Gelos, R. Gaston, Ratna Sahay and Guido Sandleris. 2011. “Sovereign borrowing by developing countries: What determines market access?” Journal of International Economics 83(2):243-254.

Gibler, Douglas M. 2009. International Military Alliances, 1648-2008. Washington D.C.: CQ.

Marshall, Monty G. and Keith Jaggers. 2000. Polity IV Project: Political Regime Char- acteristics and Transitions, 1800-2010. Center for International Development and Conﬂict Management. University of Maryland.

Menaldo, Victor. 2016.“The Fiscal Roots of Financial Underdevelopment.” American Jour- nal of Political Science 60(2):1540-5907.

Porter, Bruce D. 1994. War and the Rise of the State: The Military Foundations of Modern Politics. New York: Free Press.

Reinhart, Carmen. 2012. “The Return of Financial Repression” Centre for Economic Policy Research.

Richmond, Christine and Daniel A Dias. 2009. “Duration of Capital Market Exclusion: An Empirical Investigation.” Available at SSRN: http://ssrn.com/abstract=1027844.

Rockoﬀ, Hugh. 1998. “The United States: from ploughshares to swords.” In The Economics of World War II, ed. Mark Harrison. New York: Cambridge University Press pp. 81-121.

xl Schumpeter, Elizabeth Boody. 1938.“English Prices and Public Finance, 1660-1822.” Re- view of Economics and Statistics 20(1):21-37.

Suzuki, Toshio. 1994. Japanese Government Loan Issues in the London Capital Market 1870-1913. London: London.

Treisman, Daniel. 2014. “What Does Cross-National Empirical Research Reveal about the Causes of Corruption?” in Handbook of Political Corruption, ed. Paul Heywood. New York: Routledge.

Weller, Leonardo. 2015. “Government versus Bankers: Sovereign Debt Negotiations in Porﬁrian Mexico, 18881910.” The Journal of Economic History 75: 1030-1057.

xli