Introducing the Revenue Data

Our main measure is central government revenue per capita in gold grams, which we constructed using the following steps.

1. 1. We use data on central government revenues from Mitchell.Footnote ¹⁸ We translated fiscal years into calendar years to measure each country's annual revenues in thousands of local-currency units (although in some cases revenue is listed in US dollars).

2. 2. We use population data from Mitchell to calculate revenue per capita.Footnote ¹⁹ Official population estimates are typically available only in census years, and we estimated population in noncensus years by linearly interpolating between censuses (although we drop observations for which no census occurred within two decades). For this reason, we cannot estimate revenue per capita before the date of the first census, even when earlier revenue data are available.Footnote ²⁰

3. 3. We converted all currency measures to their equivalents in British pounds to generate a common scale for revenue levels. This required constructing a new time series of historical exchange rates into pounds.Footnote ²¹ We were unable to perform this step for country-years with nonconvertible currencies, and thus we drop such observations even if revenue and population data are available.Footnote ²²

4. 4. Finally, we converted revenue per capita in British pounds into gold grams.Footnote ²³ Although unnecessary for cross-national comparisons, this step reduces problems in data visualization stemming from the volatility of the pound. It also makes our revenue series directly comparable to other historical data sets that measure revenue in gold grams.

Our revenue variable advances existing quantitative data on state revenue in both geographical and chronological coverage. The data available are extensive, going back to the early nineteenth century in Western Europe and the late nineteenth century in most of the rest of the world. Specifically, the revenue data include at least one year for eighteen Western countries and seventy-six non-Western countries. Fifteen Western countries have at least one data point in the nineteenth century, as do forty-two non-Western countries. Relative world currency prices have fluctuated violently since the Bretton Woods system ended. For this reason, we analyze data only through 1969. Appendix Figure A.1 plots revenues over time for each territory in the data set.

We are not the first to use the Mitchell revenue data for historical analysis.Footnote ²⁴ However, our approach to weighting the data points enables us to incorporate more information than in existing studies. For example, Besley and Persson compute an unweighted average over time for eighteen rich countries.Footnote ²⁵ Consequently, they do not calculate revenue collection for poorer countries in the nineteenth or early twentieth centuries, nor make time-series cross-section comparisons across a broad country and time sample. Others use Mitchell data from the nineteenth century, but only for European countries.Footnote ²⁶ Yet others use Mitchell and other sources to construct a sample that is expansive globally after 1945, but confined to Western Europe, Japan, and the Southern Cone in the nineteenth century.Footnote ²⁷

Despite clear advantages of expansive country and time coverage, our approach to measuring state revenue intake also has drawbacks. Although we follow existing work on historical revenue collection by expressing revenue in gold or silver,Footnote ²⁸ research on contemporary fiscal extraction typically examines government revenue as a percentage of GDP. Thus, comparing trends in per capita revenue extraction does not rule out the possibility that differences in revenues mostly reflect changes in societal income. We address this concern in two ways. First, we analyze patterns for taxes as a fraction of GDP using data from Andersson and Brambor,Footnote ²⁹ albeit at the cost of a restricted non-Western sample. Second, we compare the timing of revenue and income divergence to show that large increases in revenue intake among Western countries lagged large income gains by at least half a century.

Documenting the Great Revenue Divergence

Figure 1 documents the great revenue divergence. Panel A presents our main measure of per capita revenues. Panel B presents taxes as a fraction of GDP, albeit at the cost of a smaller sample in which non-Western countries are restricted to South America (plus Mexico).

FIGURE 1. The great revenue divergence

Before World War I, the West did not dwarf other areas in revenue collection. In 1913, Chile and Uruguay each collected more revenue per capita than any country in Western Europe. Denmark collected less revenue than these two, as well as Trinidad and Tobago, South Africa, Malaysia, Cuba, and Panama. The United States collected even less than that, and was slightly behind Brazil and slightly ahead of Jamaica. When compared to all non-Western countries, Western countries collected somewhat more revenue (2.5 times). However, this discrepancy is small by contemporary standards, and is mostly driven by meager revenues in many African colonies. Of the forty-nine non-Western countries in our data set in 1913, twenty-three are in Africa, and Western countries collected 5.6 times more revenue than these colonies.

These patterns changed drastically after 1913. Between 1914 and 1969, per capita revenue intake increased fifteen-fold in the average Western country. Sharp gains were not entirely confined to the West—Japan experienced a sixteen-fold increase during this period. However, other non-Western countries failed to keep pace, and on average their revenues grew four-fold. Overall, gains outside the West were 71 percent smaller than those among Western countries. The patterns are largely similar among different subsets of non-Western countries: a nearly four-fold increase in South America and the Caribbean, a four-fold increase in Africa, and a nearly six-fold increase in Asia (Middle East, South Asia, Southeast Asia, and East Asia excluding Japan).Footnote ³⁰ By 1969, the ordering of countries in terms of per capita revenue collection largely mirrors contemporary rankings, with nearly every country outside the West (except Japan) trailing every Western country. Overall, by this time, the average Western state collected 8.5 times more in per capita revenue than the average non-Western state.

The divergence is also stark when assessing taxes as a fraction of GDP. As Panel B shows, South America outpaced the West on this measure throughout the nineteenth century. In 1913, Britain lagged Brazil, Uruguay, Chile, and Argentina. However, between 1914 and 1969, taxes rose from 6.4 percent to 19.7 percent of GDP in Western countries, a three-fold difference. The gains among South American countries were smaller, rising from 7.1 percent to 12.7 percent. Overall, these gains were 41 percent smaller than those in the West.

The robustness of the main pattern to differences in GDP is unsurprising when we consider historical timing. When economic historians discuss a “great divergence,” they mean the divergence in per capita economic output between Western and non-Western countries.Footnote ³¹ Although scholars debate the timing and causes of this divergence, they agree it occurred no later than the mid-nineteenth century amid the spread of the Industrial Revolution across Europe. Figure 2 compares Western countries to non-Western countries on both revenues per capita and GDP per capita. Until World War I, Western countries typically had a larger advantage in GDP than they did in revenue collection. In the following decades, the revenue ratio increased more sharply than the GDP ratio.

FIGURE 2. Comparing income and revenue divergence

Robustness Checks

In the appendix, we analyze the robustness of our core pattern. In Figure 1, we use taxes/GDP from Andersson and Brambor.Footnote ³² Compared to alternatives, it (a) has lesser missingness relative to our core data set (52% of country-years are missing) and (b) uses natural units. In Appendix A.1, we demonstrate qualitatively similar trends when using alternative data sets that account for differences in GDP. We analyze taxes/GDP from Beramendi, Dincecco, and Rogers,Footnote ³³ which is missing 69 percent of the country-years from our core data set, and has 47 percent fewer observations for non-Western countries than Andersson and Brambor.Footnote ³⁴ We also constructed a panel of normalized revenue data. Despite relatively better data coverage (missing 42% of country-years compared to our core sample), the units are not natural because we divide nominal revenue intake in the local currency by constant-US-dollar GDP estimates from Bolt and colleagues' update of Angus Maddison.Footnote ³⁵

One concern with our main measure of revenue per capita is that, by using nominal exchange rates, longitudinal changes in revenues may reflect changes in the foreign exchange market rather than changes in actual revenue. Appendix A.2 explains two ways in which our main measure guards against this concern. We also present intra-imperial comparisons, hence comparing territories that used the same currency or a highly stable peg. Nor can we directly account for differences in purchasing power or directly measure tax intake. However, Appendixes A.3 and A.4 explain why these shortcomings are unlikely to influence the findings. We also created a separate series that expresses central government revenue per capita in silver, rather than gold, grams. Ultimately, the choice of precious metal does not qualitatively alter the main pattern. In Appendix A.5, we discuss why we chose gold rather than silver for our primary measure. Finally, in Appendix A.6, we estimate regression coefficients for the interaction of regional location and time period to express the core pattern from Figure 1 in more precise quantitative terms.

Fiscal Demand

Demand-based theories of taxation focus on factors that create stronger preferences for central government revenues. The one most commonly studied in the literature is international warfare. Scholars broadly accept that external wars played an important role in facilitating modern European states.Footnote ³⁶ Other authors make the converse argument that less intense geopolitical competition in many ex-colonies in Sub-Saharan Africa and Latin America has undermined their state-building efforts.Footnote ³⁷

Directly, preparation for and participation in an external war raise a state's need for revenue to pay and deploy soldiers for the conflict. Indirectly, these conditions may persist in a postwar ratchet effect. States need to service debt accumulated during the conflict, and wars can also spur permanent institutional changes. Mass-mobilization wars create political consensus for egalitarian taxation systems and franchise expansion.Footnote ³⁸ These changes create pressure to sustain programs of social redistribution that emerged during the war and require high taxes to fund.Footnote ³⁹ Besley and Persson formalize the war-demand logic.Footnote ⁴⁰ The key choice in their model is government investment in future tax-collection capacity. A high valuation for public goods in “common-interest states” increases the value of future revenues, which boosts incentives for fiscal investments. External threats correspond to a high value of their public goods parameter.Footnote ⁴¹

Despite highlighting some important elements, the bellicose perspective cannot explain the great revenue divergence on its own. Why did the Western revenue advantage remain large after its former colonies gained independence, which sparked high demand outside the West? For example, between 1940 and 1975, India fought in a world war under threat of invasion (during which it raised the largest volunteer army in world history), achieved independence alongside mass franchise expansion and an ascendent political elite strongly committed to social welfare measures, and engaged in three wars with Pakistan. Yet in 1969 per capita central government revenue intake was sixty-seven times higher in Western Europe than in India. Similar international pressures in the twentieth-century Middle East and nineteenth-century South America also failed to engender sustainably large revenue collection.Footnote ⁴²

Some scholarship in the bellicose tradition incorporates additional factors that condition the effort of war on revenue collection. Examples include the presence of parliamentary institutions and the level of urbanization,Footnote ⁴³ a pre-existing political union between the military and the dominant social class,Footnote ⁴⁴ and access to international debt markets.Footnote ⁴⁵ The latter in particular is probably a factor in the great revenue divergence. However, without incorporating bureaucratic capacity into bellicose theories, we cannot simultaneously explain why a large revenue divergence arose and became permanent in the twentieth century.

Fiscal Supply

Other scholars focus on the supply of fiscal institutions that facilitate revenue collection. The core element of fiscal, or bureaucratic, capacity is information on where citizens and other producers live and how much they produce. Standardized records enable bureaucrats to determine appropriate tax quotas, to sanction nonpayers effectively, and to make society “legible.”Footnote ⁴⁶ In low-legibility societies, citizens and other producers can exit by either physically migrating or engaging in informal economic activity beyond the state's reach. Throughout history, states have needed some bureaucratic capacity to collect taxes on land and to directly tax production. Modern income and value-added taxes are even more information intensive.

Although the concept of bureaucratic capacity is inherently multifaceted, recent research measures key components of states’ information-collection abilities across broad comparative samples. Brambor and colleagues collected data on civil registration systems and state statistical offices dating back to the eighteenth century.Footnote ⁴⁷ Data on births, deaths, and marriages are essentially a precondition for effective direct taxation because otherwise bureaucrats have trouble simply identifying the citizenry. Similarly, Lee and Zhang compiled data on the effectiveness of censuses in the twentieth century, which correlates strongly with public goods provision.Footnote ⁴⁸

Fiscal capacity differs from revenue intake. States can collect information on production and life events (birth, death, marriage) without using it for taxation. States can govern a literate population capable of filling out written tax forms, without requiring them to do so. In such scenarios, fiscal capacity is latent and ready to use when the ruling group wishes.

Despite adding another important piece, analyses of fiscal supply also offer incomplete explanations for the great revenue divergence. Why did a large and permanent divergence not occur earlier? For most of the nineteenth century, Western countries outpaced others in collecting information on their citizens and educating their population. Britain imposed the world's first modern income tax during the Napoleonic Wars. However, it suspended the tax after the wars, and then did not match its 1810 per capita revenue record until 1915. Britain's high fiscal capacity remained largely latent throughout the nineteenth century. Similarly, other Western European countries improved their tax bureaucracies during the nineteenth century but underutilized their potential until the twentieth.

Sources of Fiscal Capacity

Conceptually, it is useful to distinguish war-based pressures from bureaucratic capacity. Yet empirically, these factors are not completely independent of each other. A key idea in bellicose theories of European state building is that participation in wars encouraged states to improve their bureaucracies. We agree that war is one important contributor to improvements in bureaucratic capacity over the longer run. This helps explain why latent revenue-raising capacity was higher in the West than most of the rest of the world following the Napoleonic Wars. However, bellicose theories of fiscal-capacity investments cannot answer key questions about how fiscal capacity can increase in the absence of warfare, or can fail to increase for countries within a competitive international system.

Our empirical examination begins in the early nineteenth century. At this time, historical participation in wars had likely contributed to a divergence in bureaucratic capacity between European countries and much of the rest of the world. European history provides numerous examples of states enacting bureaucratic reforms to gain a coercive edge. For example, Britain introduced the Bank of England in 1694 during the Nine Years’ War with France, which created a major financing advantage.Footnote ⁴⁹ Similar pressures during the Napoleonic Wars propelled Britain's first income tax. Mann examines several great powers (Britain, France, Prussia, Austria) and argues that warfare stimulated bureaucratic reforms before the French Revolution.Footnote ⁵⁰ These reforms introduced standards for hiring and promotion, and shifted toward salaried rather than office-owning state officials.

However, even for explaining historical levels of fiscal capacity in Europe, purely bellicose theories are incomplete. Battles occurred as often in China as they did in Europe between 1000 AD and 1800,Footnote ⁵¹ and other aspects of their regional state systems influenced why bureaucratic capacity grew in Europe but fell in China.Footnote ⁵² Nor was the effect of war uniform throughout European history. Wars often generated crippling debt and encouraged leaders to take irresponsible actions, such as debasing the currency (e.g., Louis XIV in France), as opposed to promoting fiscal systems that could generate consistent tax revenues over the longer term. Even in cases like Britain, where scholars largely agree that participation in wars contributed to bureaucratic development in the eighteenth century, they also point to specific scope conditions, such as early centralization, island location, and the lack of participation in wars during the early part of the Military Revolution.Footnote ⁵³

Even more pertinent for our empirical analysis, bellicose theories cannot account for two facts about fiscal capacity development in the nineteenth and twentieth centuries. First, at the beginning of the nineteenth century, no Western state (with the possible exception of Britain) had a “modern” bureaucracy. The historically unprecedented revenue increases that began during World War I would not have been possible without intensive improvements in fiscal capacity throughout the nineteenth century. Yet Western European states fought few wars with each other between 1816 and 1913. Why did fiscal capacity grow during an extended period of low warfare?Footnote ⁵⁴

Second, many ex-colonies faced high fiscal demand following independence from European powers. In some cases, the high demand stemmed from a competitive regional environment and bellicose pressures. Why did most of these states fail to develop strong bureaucracies?

Structure of the Tax System

The government chooses among three options for structuring the tax system. The first option is to exert low fiscal effort, such as relying on existing infrastructure to collect customs revenues. Collecting customs taxes requires relatively few agents at one or several major ports. These indirect taxes are easy to collect if the economy is already organized in a manner to facilitate international trade. This was true of Western states by the nineteenth century. In many colonies and ex-colonies, intervention by the colonizer restructured the economy to produce certain cash crops. Centeno notes the contrast between “administratively simple but inelastic customs taxes” and “more politically challenging, but potentially more lucrative, domestic sources of revenue,” which require greater bureaucratic capacity to collect.Footnote ⁵⁵

Low fiscal effort carries two drawbacks. First, it may not meet societal demand for revenues. Second, despite requiring low administrative effort, narrowly based taxes such as customs taxes often entail higher deadweight loss than more broadly based taxes such as an income tax.Footnote ⁵⁶ Customs taxes can also have adverse distributional consequences for the ruling elite relative to an income tax.Footnote ⁵⁷ As we discuss later, this created an impetus to reform tax systems in nineteenth-century Europe, despite the absence of strong war pressures.

These drawbacks may propel governments to choose either of two high-effort strategies. On the one hand, they can target a subset of producers to offer economic privileges (e.g., state-run monopolies or crony-owned firms) in return for revenue. This strategy can be lucrative because the government concentrates economic gains among highly legible cronies. However, what we classify as crony-favoring extraction entails high effort because it requires significant state involvement in and restructuring of the economy. Such restructuring enables only highly legible citizens to produce valuable goods, or facilitates direct government control over valuable assets. The clearest examples of crony-favoring economic interventions occur when governments construct state-owned enterprises or otherwise favor monopolies in certain industries. This creates a symbiotic political relationship whereby the government easily accesses information on the firm's production, and favored firms gain economic advantages. Collectivized agriculture in the Soviet Union provides an extreme example. More typical cases are ones like Egypt and India, in which the government actively intervenes in the economy to create a “captive tax base.”Footnote ⁵⁸ As Chaudhry describes, “In cases where the government becomes the primary employer and producer and assumes the role of setting prices, its task is simplified to monitoring the activities of corporations and agencies that it owns and manages.”Footnote ⁵⁹

On the other hand, the government can create widespread economic rights and attempt to tax this broader base. Legibility-intensive extraction requires information on the populace as a whole. Efficient collection of income taxes and value-added taxes requires detailed information on the identities and productivity of citizens and firms, as well as complex bureaucracies to collect and process this information.

Bureaucratic Capacity

One factor that influences the government's strategy for raising revenues in each period is bureaucratic (or fiscal) capacity, about which we make two assumptions. First, societal legibility is sticky in the short run. Thus, the state inherits a stock of bureaucratic capacity in each period. Second, states can boost their fiscal capacity over time. Hence, choosing legibility-intensive extraction in period 1 can bolster fiscal capacity in period 2.

The idea that bureaucratic capacity is persistent is widespread in the literature. Stickiness in bureaucratic capacity implies that states facing a demand shock are rarely able to rapidly and dramatically improve societal legibility. Dincecco's discussion of the historical origins of state capacity in Europe starts with the fall of the Carolingian Empire in the 800s and extends into the twentieth century.Footnote ⁶⁰ Conversely, once created, bureaucracies tend to self-perpetuate even when fiscal demand is temporarily low. Some legibility reforms can persist without any spending at all. For example, introducing last names and addresses enables states to find and distinguish citizens.Footnote ⁶¹ This information facilitates revenue extraction, even if not used immediately.

States can also take concerted efforts to improve societal legibility over time. Although this can occur through various channels, we focus on learning-by-doing effects. Thus, choosing legibility-intensive extraction early on can bolster future bureaucratic capacity, similar to fiscal-capacity investments in Besley and Persson's model.Footnote ⁶²

Setup

We model an interaction between a government and citizens, which unfolds over two periods. Time is denoted by t ∈ {1, 2}, and players value consumption equally across the two periods. The sequence of moves within each period is: (1) the government chooses how to structure the tax system, (2) the government proposes a specific tax rate to each citizen, (3) each citizen either complies with the tax and produces in the formal sector, or exits to the informal sector.

Society consists of a continuum of atomless citizens with mass N, which has a lower bound that strictly exceeds 1 (defined later) and an upper bound of $\bar{N}$ (defined in Appendix C). The set of citizens is ${\cal N}$, and each citizen is indexed by i. In both periods, each citizen produces output worth Y _i, the value of which is determined by the government's actions (presented later). Each citizen also has an exit option that yields consumption of a fraction e _i ∈ (0, 1) of that output.Footnote ⁶⁴ The exit option is individual-specific and, in each period, is independently drawn for each citizen from a smooth density function H(e _i) with positive support on [0, 1].Footnote ⁶⁵

In each period, the government decides how to structure the tax system. It begins each period with an endowment of customs revenues worth R ^cus > 0. One option for structuring the tax system is to exert low effort and rely solely on customs revenues. Alternatively, the government can choose either of two high-effort strategies to collect additional taxes.

Under a legibility-intensive strategy, the government grants legal rights to all citizens to participate in the formal economy. This choice yields output Y _i = 1 for each citizen, which is subject to taxation. Due to limitations in bureaucratic capacity, the government does not know the value of the exit option for each citizen. A fraction L _t ∈ (0, 1) of citizens are legible, and the government perfectly knows the value of e _i for each legible citizen. The remaining 1 − L _t citizens are illegible. The government knows only the prior distribution of possible values of e _i for such citizens. In any period that the government chooses legibility-intensive extraction, it pays a fixed cost $F\in ( 0, \;\bar{F}) $, with an upper bound $\bar{F} > 0$ defined later (see notes 70 and 74). We interpret F as a relative cost. When political elites perceive low-effort taxes such as customs taxes as creating extreme economic inefficiencies or unfavorable distributional consequences for themselves, in effect, the cost F of shifting the tax base is lower.

Under a crony-favoring strategy, the government favors a subset of legible citizens, normalized to mass 1.Footnote ⁶⁶ This could involve limiting economic production to specific cronies, or putting economic production directly under state ownership. Restructuring the economy to reduce competition enables each favored citizen to produce Y _i = Y, which is subject to taxation, but pushes any production by the mass N − 1 of non-cronies outside the reach of the state, and hence Y _i = 0. To make the trade-offs nontrivial, we assume (a) the crony-favoring strategy bolsters the production of favored citizens relative to their production under legibility-intensive extraction and (b) crony-favoring extraction diminishes total output. Formally, 1 ≤ Y < N, which also forms a lower bound for N. Structuring the tax system to favor cronies also incurs a fixed cost. Despite not requiring a similar bureaucratic effort as the legibility-intensive strategy, subsidies paid to favored firms and the difficulty of displacing vested economic interests create costs for a government to actively intervene to distort market competition. To isolate the role of bureaucratic capacity in distinguishing the two high-effort strategies, we assume that the government pays the same fixed cost F in any period it chooses crony-favoring extraction.

The fraction of legible citizens, L _t, reflects bureaucratic capacity (equivalently, fiscal supply). We assume L ₁ is an exogenous parameter. However, L ₂ depends in part on the revenue-collection strategy in period 1. If the government chooses legibility-intensive extraction in period 1, then L ₂ grows at a rate Δ > 1, which yields L ₂ = min{ΔL ₁, 1}. Higher values of Δ indicate stronger learning-by-doing effects, and hence higher potential bureaucratic growth. By contrast, if the government chooses low-effort or crony-favoring extraction in period 1, then L ₂ = L ₁.Footnote ⁶⁷ The endogeneity of L ₂ to choices in period 1 is the only way in which decisions in period 1 affect those in period 2.

If, in a particular period, the government chooses a high-effort strategy for structuring the tax system, it then proposes an individual-specific tax rate τ _i ∈ [0, 1] to each citizen. For every legible citizen, the value of e _i is revealed before the government chooses τ _i; and for every illegible citizen, this value is revealed immediately afterwards. Each citizen knows its individual-specific values of τ _i, e _i, and Y _i when moving. Each simultaneously responds to its tax proposal either by complying and consuming (1 − τ _i)Y _i, or exiting and consuming e _iY _i.

The government's consumption in each period depends on the revenues raised. All revenues are assumed to be spent on public goods. The government is rewarded for increasing revenue to get closer to the amount of expenditures demanded by society, denoted as $R_t^{{\rm dem}} $ for fiscal demand, and penalized for raising taxes such that total revenue exceeds the socially desired amount. The rationale here is that citizens want to fund desired public goods, but do not want to contribute taxes for undesired projects or private rents. Formally, total revenues R _t equal the customs endowment R ^cus plus any additional taxes collected through a high-effort extraction strategy. The government gains a marginal benefit of 1 from any (endogenously raised) taxes raised such that total revenues do not exceed $R_t^{{\rm dem}} $, and a marginal benefit of −1 from any taxes that push total revenues above $R_t^{{\rm dem}} $. Thus, if $R_t^{{\rm dem}} < R^{{\rm cus}}$, then the government gains negative utility from raising any taxes beyond the customs endowment. If instead $R_t^{{\rm dem}} > R^{{\rm cus}}$, then the government receives positive utility from any taxes raised up to $R_t^{{\rm dem}} -R^{{\rm cus}}$, and negative utility from raising any additional taxes. We assume $R_1^{{\rm dem}} > 0$ is a fixed parameter and that Nature draws $R_2^{{\rm dem}} > 0$ in between periods according to a distribution described later. To parallel our assumption about how endogenous tax revenue affects the government's utility, we assume the government's marginal benefit from any exogenous customs receipts is 1. However, the precise assumption here is immaterial, because the government consumes customs revenues regardless of its strategic choices. Finally, in any period in which the government chooses a high-effort strategy for structuring the tax system, it pays the fixed cost F. This subtracts from its consumption.

Short-Term Revenue Intake: Analysis of Period 2

We solve backwards to characterize subgame perfect strategies, and all proofs are in Appendix C. In period 2, the government cares solely about short-term revenue intake. We first derive maximum possible revenues, and then we explain the government's optimal approach.

Each legible citizen will comply with a tax proposal that satisfies τ _i ≤ 1 − e _i. Thus, maximizing revenue requires the government to set the individual-specific tax rate to make each legible citizen indifferent between complying and exiting: $\tau _i^\ast{ = } 1-e_i$. This strategy induces every legible citizen to comply.Footnote ⁶⁸ By contrast, a lack of discriminating information forces the government to set the same tax rate for each illegible citizen. The optimal rate balances two considerations. A higher tax rate yields a higher fraction of income, τ, from each illegible citizen who complies, but decreases the fraction of illegible citizens who comply. Only illegible citizens with low-valued exit options (relative to the tax proposal) comply, e _i ≤ 1 − τ. The revenue-maximizing tax proposal for illegible citizens solvesFootnote ⁶⁹

(1)$$\hat{\tau }\equiv \arg \mathop {\max }\limits_{\tau \in [ 0, 1] } \int_0^{1-\tau } \tau \,dH( e_i) $$

Structuring the tax system to favor cronies yields a unit mass of favored citizens who each produce Y _i = Y and are perfectly legible, and the remaining mass of N − 1 citizens produce Y _i = 0. Legibility-intensive extraction enables all citizens to legally produce Y _i = 1, but only a fraction L ₂ are legible. If the government makes the revenue-maximizing tax proposal to each citizen, expected revenues under each way to structure the tax system are

(2)$$R^{{\rm crony}} = \underbrace{{Y\int _0^1 ( 1-e_i) dH( e_i) }}_{{{\rm All\;favored\;citizens\;are\;legible}}}$$

(3)$$R^{{\rm leg}}( L_2) = N\underbrace{{\left[{L_2\int_0^1 ( 1-e_i) dH( e_i) } \right.}}_{{{\rm Legible\;citizens}}} + \underbrace{{\left. {( 1-L_2) \int_0^{1-\hat{\tau }} \hat{\tau }\,dH( e_i) } \right]}}_{{{\rm Illegible\;citizens}}}$$

Comparing these two terms shows that legibility-intensive revenue extraction yields higher revenues than crony-favoring policies if and only if bureaucratic capacity is high enough.

Lemma 1 (Maximum tax extraction)

A unique threshold $\bar{L}\in ( 0, \;1) $ exists such that $R^{{\rm leg}}( \bar{L}) = R^{{\rm crony}}$. Given this threshold, we can express maximum tax revenues in period 2 as

(4)$$R_2^{{\rm max}} = \left\{{\matrix{ {R^{{\rm crony}}\quad } & {{\rm \;if\;}L_2 < \bar{L}} \cr {\;\;\;\quad } & {} \cr {R^{{\rm leg}}( L_2) \quad } & {{\rm \;if\;}L_2 \ge \bar{L}} \cr } } \right.$$

This result establishes the importance of fiscal supply. The next question is whether it is optimal for the government to maximize revenues, which depends on fiscal demand. If $R_2^{{\rm dem}} < R^{{\rm cus}}$, then the government can fund all desired expenditures while exerting low effort at tax collection. At the other extreme, if $R_2^{{\rm dem}} > R^{{\rm cus}} + R_2^{{\rm max}} $, the government taxes maximally. Within these lower and upper bounds, if fiscal demand is close to R ^cus, then the fixed cost deters the government from pursuing high-effort extraction. If instead fiscal demand is close to $R^{{\rm cus}} + R_2^{{\rm max}} $, then the government pays the fixed cost for either legibility-intensive or crony-favoring extraction, but intentionally sets taxes to collect less-than-maximum revenues. Proposition 1 presents a subgame perfect Nash equilibrium strategy profile.Footnote ⁷⁰

Investments in Fiscal Capacity: Analysis of Period 1

In period 1, the government cares not only about how the structure of the tax system affects contemporaneous revenues, but also about revenue intake in period 2. The core implication is qualitatively unaltered: only governments with high fiscal demand and (the potential for) high fiscal supply choose legibility-intensive extraction. However, the threshold for “high” fiscal supply (i.e., bureaucratic capacity) is lowered because the shadow of the future heightens incentives to pursue legibility-intensive extraction. To make the analysis parallel with the preceding section, we characterize threshold values of initial bureaucratic capacity that determine optimal actions. We then take comparative statics on other variables that influence the government's calculus for investing in fiscal capacity.

If the starting level of bureaucratic capacity takes an extreme value, then the government's choice for how to structure the tax system in period 1 does not influence its choices in period 2. Suppose initial societal legibility is very high, $L_1 > \bar{L}$ (see Lemma 1 for this threshold). This guarantees that legibility is high enough in period 2 that, regardless of the government's actions in period 1, legibility-intensive extraction yields more revenues than crony-favoring extraction. If instead initial legibility is very low, $L_1 < \underline{L} \equiv {{\bar{L}} \over {\rm \Delta }}$, then the converse implication is true. Even if the government gets a learning-by-doing boost to bureaucratic capacity in period 2, crony-favoring extraction would yield more revenue than legibility-intensive extraction.

Only if initial legibility is between these two thresholds does the shadow of the future yield new insights. To illuminate the substantively important insight, we focus on a specific case within the range $L_1\in ( \underline{L} , \;\bar{L}) $. Fiscal demand is low in period 1, $R_1^{{\rm dem}} < R^{{\rm cus}}$. This ensures that the government will not seek additional revenues in period 1 beyond the customs endowment, even if it chooses to invest in fiscal capacity. This is the interesting case illuminated by the dynamic analysis because, in a single-shot game, the government would never choose legibility-intensive extraction if fiscal demand is low. We also assume that Nature draws fiscal demand for period 2 $( {R_2^{{\rm dem}} ) $ from a Bernoulli distribution that takes value $R_{{\rm high}}^{{\rm dem}} $ with probability p _high ∈ (0, 1), and $R_{{\rm low}}^{{\rm dem}} $ with complementary probability. We set these values so that for a low draw of fiscal demand, the government chooses a low-effort tax structure and does not seek additional revenues in period 2 beyond the customs endowment; and for a high draw of fiscal demand, the government seeks maximum extraction. The formalization of these thresholds follows directly from Proposition 1: $R_{{\rm low}}^{{\rm dem}} < R^{{\rm cus}} + F$, and $R_{{\rm high}}^{{\rm dem}} > R^{{\rm cus}} + R^{{\rm leg}}( 1) $.Footnote ⁷³

If the government chooses legibility-intensive extraction in period 1, then its total expected consumption across the two periods is

(5)$$\underbrace{{R^{{\rm cus}}-F}}_{{{\rm Period\;1}}} + \underbrace{{R^{{\rm cus}} + p_{{\rm high}}[ {R^{{\rm leg}}( \min { {\rm \Delta }L_1, \;1} ) -F} ] }}_{{{\rm Period\;2}}}$$

If instead the government does not invest in bureaucratic development in period 1, then its expected utility is

(6)$$\underbrace{{R^{{\rm cus}}}}_{{{\rm Period\;1}}} + \underbrace{{R^{{\rm cus}} + p_{{\rm high}}( R^{{\rm crony}}-F) }}_{{{\rm Period\;2}}}$$

Comparing these two terms uncovers the conditions under which the government invests in fiscal capacity for period 2: legibility-intensive extraction (after getting the learning-by-doing boost) must yield sufficiently more revenues than crony-favoring extraction. Equating the two preceding expressions enables us to define an implicit threshold $\hat{L}$ that determines whether the government makes this investment:Footnote ⁷⁴

(7)$$R^{{\rm leg}}( {\rm \Delta }\hat{L}) = R^{{\rm crony}} + \displaystyle{F \over {\,p_{{\rm high}}}}$$

As in the period 2 analysis, the stock of bureaucratic capacity influences the government's optimal revenue-raising strategy. However, in period 1, the government may choose legibility-intensive extraction even if fiscal demand is low and crony-favoring extraction yields higher maximum revenues. The future gains created by bureaucratic growth change the government's calculus. Proposition 2 formalizes this intuition.Footnote ⁷⁵

Implications for Revenue Divergence

The model explains why increases in fiscal demand over time can create a revenue divergence between states that vary in bureaucratic capacity. High-capacity states distinguish themselves in revenue collection only when fiscal demand is high.Footnote ⁷⁷

In Figure 3, we depict hypothetical revenue trajectories for two countries that are identical except in their initial level of bureaucratic capacity, one with low $L_1 < \hat{L}$ and one with intermediate $L_1\in ( \hat{L}, \;\bar{L}) $. We assume that fiscal demand is low for both in period 1. Consequently, neither government raises revenue beyond its endowed customs taxes. However, the government with intermediate initial bureaucratic capacity (solid line) nonetheless will sink the fixed cost to create future foundations for legibility-intensive extraction. By contrast, the low-legibility state (dashed line) has no incentive to invest in future fiscal capacity despite anticipating an identical probability of high fiscal demand in period 2. The learning-by-doing effects from investing in bureaucratic capacity are sufficiently small that this government would continue to respond to high fiscal demand in the future with crony-favoring extraction.

FIGURE 3. Hypothetical revenue trajectories

If Nature draws high fiscal demand for both governments in period 2, then revenue divergence occurs. Although both extract maximally, the government with higher bureaucratic capacity chooses legibility-intensive extraction and gains higher revenue intake.Footnote ⁷⁸ By contrast, the low-legibility state chooses crony-favoring extraction. Existing models, such as Besley and Persson's,Footnote ⁷⁹ cannot account for this divergence. In their model, any state that anticipates high fiscal demand in the future is a “common value state” that will invest in fiscal capacity to capitalize on demand shocks. However, in our model, bureaucratic capacity conditions the effect of demand shocks. In the short term, low fiscal supply pushes high-demand states toward crony-favoring rather than legibility-intensive extraction. In the long term, states with poor prospects for bureaucratic growth will have lower incentives to invest in fiscal capacity even if they anticipate high fiscal demand in the future.

Low Fiscal Demand in the Nineteenth-Century West

Between the conclusion of the Napoleonic Wars and the start of World War I, revenue intake was low in Western countries. Bellicose theories anticipate this pattern because infrequent intra-European warfare lowered fiscal demand. However, despite minimal pressures from war, fiscal capacity grew throughout the century. Our theory accounts for nonbellicose stimulants to bureaucratic reform.

In Figure 4, we highlight low revenue intake in important states. Britain imposed the world's first modern income tax during the Napoleonic Wars,Footnote ⁸⁰ a period we highlight in gray. However, per capita revenue intake declined afterwards, and this decline is even more pronounced when accounting for Britain's strong economic growth. Even when Britain reimposed an income tax, in 1842, the marginal rate began at 2.9 percent, and it remained low into the twentieth century. France imposed a new set of direct taxes starting with its Revolution, but did not implement an income tax until World War I. The United States experienced a brief spike in revenues when it imposed an income tax during its Civil War and Reconstruction (1861–72), which we also highlight in gray. However, for most of the nineteenth century, the US government was a “state of courts and parties.” In the 1870s, customs revenues from the Port of New York accounted for more than half of all federal revenues.Footnote ⁸¹ Customs revenues were sufficient to cover the small federal budget, and during the century they constituted on average 72.1 percent of US revenues. The major outlier among Western countries was New Zealand, which we omit from the Other Western average in the figure to not obscure the main pattern. Throughout the nineteenth century, New Zealand consistently collected high levels of customs taxes relative to the small white population.

FIGURE 4. Western revenue intake before WWI

Low revenue intake reflected low fiscal demand. The long nineteenth century was more peaceful than the preceding or subsequent periods. Britain, for instance, participated in a major war against at least one other European power for 76 of the 150 years from 1665 to 1815, but in only three years between 1816 to 1913. Wars of imperial conquest or within established colonies occurred more frequently (61 years for Britain). However, these conflicts were much less costly than intra-European conflicts. For example, the First Anglo–Burmese war and the Anglo–Zulu war each cost £5 million, and the First Anglo–Afghan War cost £14 million. By contrast, Britain's participation in World War I cost £3.25 billion. The main reason for the lower costs was that imperial wars required small commitments from metropolitan troops.Footnote ⁸² To assess this claim systematically, we analyzed whether war years correlate with a larger mobilization of domestic British troops.Footnote ⁸³ When defining “wars” as intra-European conflicts, the correlation is positive and statistically significant. By contrast, when defining “wars” as imperial, the correlation is null.

Similarly, until the very end of this period, a limited franchise dampened domestic incentives for social provision and redistributive taxation. Britain did not provide old-age pensions until 1908, unemployment insurance until 1911, or universal secondary education until 1918. Britain's upper-class political leadership would not personally benefit from heavy taxation, and they faced minimal demand from their middle-class electorate to boost expenditures. Across Western Europe, demand for welfare provisions was low throughout the nineteenth century.Footnote ⁸⁴

Given low fiscal demand, bellicose theories anticipate retrenchment in Britain and many continental powers. However, these theories cannot account for why, nonetheless, fiscal capacity tended to increase during this period. All ten countries that introduced civil registration systems for births and deaths before 1850 are in Western Europe or its offshoots.Footnote ⁸⁵ Age heaping in US censuses declined by 62 percent among native whites between 1850 and 1900, with even larger improvements among other racial groups.Footnote ⁸⁶ These gains were essential for enabling Western states to collect unprecedented levels of revenues when fiscal demand spiked starting in 1914.

Our model helps explain why Western states invested in fiscal capacity despite low fiscal demand (and hence they underutilized their fiscal capacity). One relevant factor was that European countries began the period with greater fiscal capacity than most non-Western countries, that is, L ₁ was high. This was, in part, a legacy of prior wars, as discussed earlier. Yet two other key conditions from the model that explain bureaucratic growth were independent of war.

First, every Western country experienced industrialization before World War I.Footnote ⁸⁷ This increased the potential for bureaucratic growth, which corresponds with higher Δ in the model. Industrialization reshaped citizens in ways that made it easier for states to control and tax them. Residents of cities are easier to monitor than villages, and it is easier to impose income taxes on cash wages than on harvests consumed as subsistence. The basic literacy, numeracy, and awareness of time required by industrial firms created positive spillovers for states. Zhang and Lee provide quantitative evidence that literacy is strongly associated with state capacity, and industrialization and urbanization facilitated increases in literacy.Footnote ⁸⁸

Second, European states inherited irrational and inefficient state structures from the eighteenth century. This, in effect, reduced the net costs of reforming the tax system, hence lowering the fixed cost (F in the model) of implementing legibility-intensive extraction. Reforming the bureaucracy enabled political elites to reduce economic deadweight losses. Even the early-nineteenth-century English bureaucracy, often cited as a global model, was recruited through patronage networks and purchase. Many so-called bureaucrats had sinecures and were recompensed by fees rather than salaries.Footnote ⁸⁹ Statesmen took actions throughout the nineteenth century to eliminate these abuses and to recruit bureaucrats by examination.Footnote ⁹⁰

Similarly, many governing elites considered the income tax more equitable than the easily collected tariffs and excise taxes it replaced, which also bolstered quasi-voluntary compliance.Footnote ⁹¹ Sir Robert Peel's speech proposing the reintroduction of the British income tax in 1842 stated that the resulting surplus would be used in “making of great improvements in the commercial tariff in England; in addition to these improvements to abate the duties on some great articles of consumption,”Footnote ⁹² and in fact a general reduction in tariffs occurred around that time. Previously, state officials viewed the income tax as appropriate only at times of serious fiscal crises, such as wars. However, reformers like Peel argued that an income tax with low marginal rates offered a welfare-improving means to pay the ordinary costs of peacetime administration.Footnote ⁹³ Across the continent, landed elites often pushed for the introduction of income taxes, not to finance greater social expenditures but to ensure that the rising capitalist elites incurred a greater share of the total tax burden.Footnote ⁹⁴

Primary Product Exporters and Empires in the Nineteenth Century

Customs revenues in primary product exporters

Even with low demand for public expenditures in the West, we might still expect these states to collect more revenue than states then (or recently) under Western colonial rule. European colonial rule was typically based on predatory extraction, which perpetuated low fiscal supply. Colonizers shaped the fiscal systems of dependencies to reflect metropolitan objectives. Most colonies made extensive use of either coercive labor institutions or local intermediaries. Spanish administrators in the Americas plundered their colonies for gold and silver, often using indigenous forced labor for mining and other production purposes. Elsewhere, Europeans forcibly imported millions of Africans to work as slaves on plantations throughout the West Indies and other areas where the climate permitted the production of sugar and other valued commodities. African colonies were characterized by high labor coercion and financial decentralization.Footnote ⁹⁵ Britain collected head, hut, and other direct taxes in Africa through Native Authorities acting on the state's behalf.Footnote ⁹⁶ Throughout Africa, colonists co-opted local indigenous institutions and aimed simply to collect enough taxes to balance the budget. Similarly, in most areas of colonial India, colonial officials delegated the collection of the land revenue tax to princes or large landlords, and the government was usually unable to raise the rates they assessed.Footnote ⁹⁷

Despite not boosting legibility, European colonizers structured local economies to facilitate easily taxable primary product exports. This enabled some dependencies to keep pace with the West without high fiscal effort. Europe's dominance in this period was based on superior military technology, scientific innovations and economic development, and epidemiological advantages, rather than high levels of taxation. In Figure 5, we compare revenue intake in Britain and the average among other Western countries to four baskets of non-Western countries between 1800 and 1913.

FIGURE 5. Comparative revenue intake before World War I

Countries in the Southern Cone of South America gained independence in the first half of the nineteenth century. On average, their revenue intake was relatively high. Between 1900 and 1913, they collected 33 percent more in revenue per capita than Britain, and more than twice the revenue of other Western countries, on average. Measuring taxes as a percentage of GDP reveals similar discrepancies. Whereas Britain and the average of all other Western countries each raised 6.4 percent of GDP in taxes, the corresponding figure for Southern Cone countries was 9.2 percent. Revenues were particularly high in Chile, which reflected a boom in nitrate mining.Footnote ⁹⁸ Between 1900 and 1913, customs taxes constituted, on average, 71.3 percent of Chile's total revenues. In addition to how easy it was for primary product exporters to collect customs taxes from a handful of ports, high demand also contributed to revenue extraction in Chile. Victory in the War of the Pacific (1879–83) cemented the influence of domestic coalitions that favored an expansive, extractive state.Footnote ⁹⁹

In Panels (b) and (c) in the figure, we plot revenues from territories that were, at the time, subjected to colonial occupation. We distinguish between two types: plantation colonies, in which a high fraction of the population was forced migrants engaged in production of cash crops on plantations; and colonies of occupation, with largely indigenous populations. Plantation colonies, with more direct rule and high levels of trade, collected somewhat less revenue than European countries: from 1900 to 1913, 55 percent less than Britain, and 31 percent less than other Western European countries.Footnote ¹⁰⁰ However, these gaps are strikingly small compared to modern discrepancies or when considering the vastly superior bureaucratic institutions in the West. Furthermore, when normalizing by GDP, the advantage flips. Plantation colonies collected 72 percent more than Britain, and more than two times the average of other Western European states.Footnote ¹⁰¹

Western countries were clearly distinguished from occupation colonies in revenue intake, even before World War I. Between 1900 and 1913, Britain collected nearly thirteen times more in revenue per capita than occupation colonies, and other Western countries collected eight times more. Yet once again, these magnitudes were small by modern standards, and differences in GDP account for most of the discrepancy. Normalizing by income, Britain collected only twice as much revenue as occupation colonies, and the rest of the West collected only 65 percent more than occupation colonies.Footnote ¹⁰²

One possible concern is that comparing sovereign and nonsovereign polities is problematic. Perhaps colonizers exploited their colonies to fund expenditures at home, which would enable them to keep domestic taxes low. However, this alternative explanation cannot account for the patterns presented here. It cannot explain why independent states in the Southern Cone extracted large amounts of revenue, nor why occupation colonies in Africa and Asia extracted minimal revenue. Research by economic historians shows that in the largest empires (Britain and France), colonial subsidies and defense expenditures exceeded in magnitude any revenue intake, which departed from the goal of financial self-sufficiency in the colonies. Analyzing Britain in the half-century preceding World War I, Davis and Huttenback argue that the empire is better characterized as “a redistribution of income within the United Kingdom than as a transfer from the empire to the mother country.”Footnote ¹⁰³ Although many European investors benefited from colonial rule, this was possible because of the security environment funded by metropolitan taxpayers. Only in the small empires with one or several profitable colonies (Dutch, Belgian, Portuguese) did the empire contribute a significant net inflow to the metropole, mirroring patterns from imperial Spain in earlier centuries.Footnote ¹⁰⁴ These authors also stress that “colonial revenues were first and foremost needed to secure internal order … [and] to pay the salaries of government officials who administered the government departments.”Footnote ¹⁰⁵

Surging Revenues in the Twentieth-Century West

Starting with World War I, Western governments experienced permanently high demand for revenues. This change, in conjunction with their bureaucratic foundations for effective legibility-intensive extraction, yielded sustainably large revenue intake, as seen in Figure 1.

The two world wars required unprecedented mobilization of troops, reorganization and management of the economy to supply the war effort, and financing needs. European states overhauled their tax systems,Footnote ¹¹³ and experienced pressure to expand the franchise and provide citizens with a broad array of social welfare benefits to reward their sacrifices,Footnote ¹¹⁴ even in countries that did not directly participate in the wars.Footnote ¹¹⁵ The Great Depression also stimulated demand for activist states, as did geopolitical competition during the Cold War.

The legacy of high bureaucratic capacity enabled Western states to capitalize on these demand shocks. Improvements in fiscal capacity during the nineteenth century, which we described earlier, were a precondition for unprecedented increases in revenue intake. Legibility-intensive extraction—income taxes in particular—replaced narrowly based taxes as the primary revenue source in Western states. Heavily reliant on bureaucratic competency and societal legibility, income taxes represented a major technological breakthrough in taxation capacity. Mares and Queralt praise the “unprecedented revenue generating capacity” of “the most advanced fiscal instrument to date.”Footnote ¹¹⁶ This is also true of advanced consumption taxes, such as the value-added tax that became common in Western Europe.Footnote ¹¹⁷ Each tax requires high social legibility for efficient collection. Income taxes are hard to evade if citizens receive monetary income by check or transfer, and value-added taxes are difficult to evade if firms routinely provide and receive invoices for sales and purchases.

In Figure 6, we present the fraction of revenue deriving from either customs or income taxes for Western states. Customs taxes once constituted the main source of revenues in Western offshoots and were also sizable in Western Europe. However, by the second half of the twentieth century, they were largely unimportant. In 1969, customs taxes provided only 6 percent of revenues in Western offshoots, and 10 percent in Western Europe. By this time, income taxes were the main source of revenues for Western offshoots (69%). Income taxes were less important in Western Europe (34%) because these countries relied more heavily on advanced consumption taxes. In 1969, all direct taxes plus advanced consumption taxes constituted, on average, 54 percent of revenues in Western European countries.Footnote ¹¹⁸

FIGURE 6. Sources of Western revenues: customs and income taxes

Low Fiscal Capacity in Former Western Colonies

The revenue gap between the West and the rest of the world increased exponentially after World War II, when most of the colonized world gained independence, as seen in Figure 1. Insufficient bureaucratic reforms during the colonial period offer a more compelling explanation than bellicose theories for the general inability of non-Western states to converge toward Western revenue intake after gaining independence. Despite generally high demand in the postcolonial world, inadequate bureaucratic capacity hindered revenue intake.

Under colonial rule, the predominant strategies of taxing cash-crop exports and relying on local intermediaries did not require advanced bureaucracies. For example, for direct taxation, British administrators in Africa largely relied on head or hut taxes because they were “unable to collect information on individual taxpayers and their incomes.” They varied the rate of taxation by district based on the crude assumption that Africans “living in areas close to the railways or opportunities for wage labour could afford to pay a higher rate than those living in more remote regions.”Footnote ¹¹⁹ Easy-revenue sources often satisfied the limited needs of colonial states before World War II, at least relative to the costs of constructing more intensive systems.

Independent countries in South America in the nineteenth century also had an alternative to building intensive bureaucracies, even when they competed in wars. Low global interest rates enabled states to pay for wars with debt, which they often renegotiated after the wars, rather than to develop intensive forms of domestic tax collection.Footnote ¹²⁰ The availability of international capital reduced the elasticity problems inherent in low-legibility taxation. Instead, states could mortgage future customs and mineral revenues to gain short-term financing.

When fiscal demand rose after independence in the twentieth century, the neglect of bureaucratic reforms during the critical juncture of colonial rule became problematic. Most former colonies lacked a civil registration system at independence (which we use as a proxy for bureaucratic development in Appendix B), although most early-independence South American countries established a civil registration system in the late nineteenth century. At independence, India had forty-six times as much census-age misreporting as the United States.Footnote ¹²¹ Given the low supply, we anticipate that the heightened fiscal demand after gaining independence should not discernibly boost revenue collection. Lee and Paine provide quantitative support for this contention by demonstrating null differences in countries’ revenue intake before and after independence.Footnote ¹²²

In many postcolonial countries, low legibility has persisted long after independence. Many still lack extensive written or electronic records to monitor activity, or banking intermediaries that reduce the need for government agents to meet in person to collect taxes. In some African and Asian countries, customs revenues became more important in the mid-twentieth century. Governments gained freedom to set tariff rates, and older land or labor taxes declined in importance or were abolished by post-independence governments intent on reforming the colonial system. Bates notes that after independence many African rulers used funds from agricultural marketing boards (which serve the ostensible purpose of stabilizing prices for and revenues from primary products) to raise revenues by exploiting farmers.Footnote ¹²³ Even when non-Western states have tried to impose modern legibility-intensive taxes, a lack of bureaucratic capacity has often impeded collection. In 1969, the average non-Western country collected 28 percent of its revenues from income taxes, and 20 percent from customs taxes.Footnote ¹²⁴ One exception was South Africa (51% of revenues from income taxes), which was highly effective at raising taxes within the white community.Footnote ¹²⁵

Egypt and India provide striking contrast cases for bellicose theories. Despite frequent participation in international warfare, these states collect low levels of revenue. A state needs high bureaucratic capacity to make its society legible. When this condition fails, rulers turn to crony-favoring extraction. Egypt and India each faced high demand for revenue given their participation in prolonged international rivalries (with Israel and Pakistan, respectively) that on several occasions flared into war. Yet both developed large and inefficient public sectors, as opposed to cultivating more sustainable sources of revenues. “The SOE [state-owned enterprise] sector does represent a captive tax base, and even as the SOEs run at a loss and seek financing abroad, they still generate a predictable source of taxes and compulsory payments to various fiscal agencies.”Footnote ¹²⁶ Egypt's attempt to implement a broad land reform in the 1950s and 1960s, which would have cut out large landowners as intermediaries in the tax-collection process, failed due to basic problems of bureaucratic information on land titles and related issues.Footnote ¹²⁷ In India, the proportion of revenue collected through direct taxes fell during the twentieth century, from 28 percent in 1900 to 15 percent in 2000, with a low of 6.5 percent in 1987.Footnote ¹²⁸ Tax-avoidance rates remained high, and the government chose to raise import duties and nationalize large sectors of the economy. In 1969, each country collected less revenue per capita than the average non-Western country, and a small share of their revenues came from income taxes (15% in Egypt and 17% in India).

The main exceptions to the general pattern of fiscal weakness in the non-Western world are the “developmental states” of East Asia. Our theory anticipates these exceptions, which combined high supply and high demand. Japan, Taiwan, and South Korea all had long traditions of professionalized bureaucracies—longer than those in the West. These countries experienced high demand for revenue to fund participation in World War II, their subsequent recovery, Cold War rivalries (including the Korean War), and ambitious programs of infrastructural development and public service provision. In 1969, Japan ranked ninth globally in per capita revenue collection, and exceeded the Western average. Japan collected 62 percent of its revenue from income taxes, which was in line with Western offshoots. As late as 1964, Taiwan collected less revenue per capita than the average non-Western country—but by 1969, nearly twice as much. South Korea collected 33 percent of its revenues from income taxes in 1969, in line with the Western European average.Footnote ¹²⁹ Taiwan and South Korea further converged to Western patterns in subsequent decades.