Electronic copy available at: http://ssrn.com/abstract=2602307

The Conservative Origin of Income Taxation

Isabela Mares∗ and Didac Queralt†

May 2015

Abstract

This paper examines the adoption of income taxes by Western economies sincethe 19th century. We identify two empirical regularities that challenge predictionsof existing models of taxation and redistribution: while countries with low levels ofelectoral enfranchisement and high levels of landholding inequality adopt the incometax first, countries with more extensive electoral rules lag behind in adopting these newforms of taxation. We propose an explanation of income tax adoption that accountsfor these empirical regularities. We discuss the most important economic considerationof politicians linked to owners of different factors, namely, the shift of the tax burdenbetween sectors, and examine how pre-existing electoral rules affect these politicalcalculations. The paper provides both a cross-national test of this argument and amicro-historical test that examines the economic and political determinants of supportfor the adoption of the income tax in 1842 in Britain.

∗Department of Political Science, Columbia University. im2195@columbia.edu†UC3M-Juan March Institute, Center For Advanced Studies in the Social Sciences, dqueralt@march.es

1

Electronic copy available at: http://ssrn.com/abstract=2602307

1 Introduction

The 20th century is the era of the Tax State (Schumpeter, 1918). The size of government

expanded at an unprecedented pace (Lindert, 2004). The development of massive spending

and investment programs transformed the state from a mere security provider to a central

actor in all aspects of economic life. Importantly, this transformation was only possible

after the adoption of the income tax and its unprecedented revenue generating capacity.

Such extractive capacity makes the income tax the most advanced fiscal instrument to date

(Lieberman, 2002; Tilly, 1990). In the words of Johannes Popitz, the income tax is the

“queen of taxation” (Popitz, 1926).

The income tax facilitated an increase in the extractive capacity of the state because

its adoption went hand in hand with the development of institutions that tapped into and

collected previously undetected revenues. In particular, the implementation of an income tax

required the creation of a sophisticated tax administration capable of verifying the income of

higher earning individuals in a given economy and ensuring the compliance of the latter with

their tax obligations. Thus, to understand this fundamental shift in the extractive capacity

of the state, we need to characterize the political conditions under which elites agreed to

subject their income to public scrutiny, the institutions and mechanisms set in place to verify

income, and the allocation of the tax burden across individuals who derive their income from

different sources.

Existing explanations of fiscal capacity development share one common over-arching

theme: war is the driving motor accounting for the expansion of the modern state (Dincecco,

2011; Scheve and Stasavage, 2010; Tilly, 1990). By contrast, domestic economic and political

variables that account for the decisions of political elites to invest in the development of in-

stitutions of fiscal revenue extraction have received less attention.1 While acknowledging the

importance of war, this paper turns to an examination of the latter political factors. While

the introduction of income taxes has the potential to generate higher levels of revenue, it

1An exception being Levi (1988) and Besley and Persson (2011).

2

also imposes a higher tax burden on individuals with higher levels of earnings. The adoption

of this novel form of taxation raised a range of political questions about the allocation of

this tax burden across owners that derive their income from different types of assets. Our

paper seeks to identify economic and institutional factors that facilitate the resolution of

these distributional conflicts and the adoption of income taxes.

In examining the variation in the timing at which current advanced economies adopted

income taxes, Aidt and Jensen (2009) uncover an empirical regularity that runs counter to

the predictions about the temporal sequencing between democratization and taxation derived

from canonical redistributive models (Acemoglu and Robinson, 2000; Boix, 2003): namely,

non-democratic countries with restrictive suffrage rules pioneered the permanent adoption of

income taxes. Our paper seeks to present an explanation to the puzzle posed by the sequence

of the income tax adoption in the developed world. That involves understanding the political

and economic motives of elites in non-democratic contexts to adopt a tax that imposed a

higher burden on high-income earners. Given that neither the Meltzer Richard model nor the

more recent literature on democratization and redistribution provide useful tools to address

this puzzling empirical regularity, our paper takes the task of formulating an explanation

that accounts for the non-democratic origin of the income tax. Rather than invoking elites

anticipation and fear of potential redistribution under full suffrage, we contend that one

needs to clarify to immediate economic and political advantages provided by the income tax

to incumbent elites in countries with restricted electoral rules. We argue that, in addition

to their capacity to raise new revenue, income taxes provided specific political and economic

benefits to political incumbents in limited democracies. Our theoretical explanation specifies

these advantages.

Consider first the economic benefits of the new tax. The incidence of the new income

tax across different sectors is rarely neutral. The design of income taxes inevitably creates

winners and losers between sectors that face a higher or lower tax burden. All non-democratic

countries adopted income taxes in a period of massive economic and political change, at a

3

time when the economic power of incumbent landowning elites was severely threatened by

the rise of a new economic elite linked to the emerging manufacturing sector. Anticipating a

future decline in economic power, politicians representing the interests of landowning elites

regarded the income tax as a tool that could rebalance some of these economic losses by

imposing a higher tax burden on the industrial sector. In adopting the new tax, incumbent

economic elites placed thus a high political priority on the creation of institutions that

monitored less visible income and also on the taxation of more mobile assets.

The above discussion implies that politicians whose interests are tied to owners of fixed

assets were more willing to support the adoption of the new income tax as compared to politi-

cians representing the interests of mobile asset holders. While the risk of capital mobility

imposed an upper bound to the tax rate levied onto capital gains, it did not prevent landed

elites from imposing, when possible, higher tax rates on industry and finance than on land.

We test for this logic by examining the consequences of differences in the bargaining power of

landed elite in the national parliaments of 19th century Western World on the probability of

income tax adoption. As common in the literature, we use concentration of land ownership

as proxy of the bargaining power of landed elites (Ansell and Samuels, 2015; Boix, 2003;

Ziblatt, 2008). Higher levels of landholding inequality were likely to enhance the probability

of adoption of the new tax, by giving landowners a higher ability to design a new tax that

imposed new fiscal burdens on the manufacturing sector. We test this proposition twofold:

cross-nationally, using a sample of 17 developed economies; and using micro-historical data,

by modeling the economic and political determinants of the vote to adopt the Income Tax

Act in the United Kingdom in 1842.

The second set of factors explaining the considerations of incumbent elites in non-

democratic settings to adopt income taxes were political in nature. Early nineteenth century

electoral rules in the Western World included a variety of provisions that used payment of

direct taxes to exclude individuals from political participation. In many early adopters, pre-

existing electoral laws conditioned voting rights on direct tax payments. These pre-existing

4

electoral provisions were an institutional mechanism that reinforced inequality in political

participation. In these countries, incumbent elites recognized that the income tax could be

layered on pre-existing electoral rules and reconfigured at times of elections as a wedge that

excluded low-income voters from parliamentary participation.

We explore the consequences of an extended restrictive electoral institution that was in

place in nineteenth century Europe prior to the adoption of universal suffrage: the vote-tax

link. This electoral provision considered payment of direct taxes as a necessary criterion

to grant electoral rights. We explore how this electoral rule that linked tax payments to

political rights modified the calculations of political insiders about the desirability of a new

income tax. In countries with a vote-tax link, the new income tax could be reconfigured

as a barrier preventing the political participation of poor voters who did not pay this tax.

These advantages of this restrictive electoral for political insiders could be shared by all high-

income voters, irrespective of their asset-ownership. As such, these considerations created

opportunities for a political alliance in support of an income class among owners of fixed

and mobile assets whose economic considerations about the adoption of the new income tax

could differ. We claim that considerations about these possible political gains prevented the

emergence of a frontal opposition of manufacturing elites to a tax which was economically

disadvantageous to this sector. To test this argument, we examine empirically the relation-

ship between vote-tax link and the probability of income tax adoption. Consistent with our

hypothesis, we find that the vote-tax link increases the probability of income tax adoption

under non-democratic settings.

Our paper contributes to the literature that examines the origin and development of fiscal

institutions in several important ways. First, to account for the puzzling adoption of the

income tax in non-democratic settings Aidt and Jensen (2009), we formulate a new micro-

logic that accounts for the motifs of elites in countries with restrictive suffrage to support

the development of extractive fiscal institutions. Our explanation stresses the importance

of inter-sectoral competition among landed and manufacturing elites over the allocation of

5

the tax burden, and considerations about likely gains in terms of political representation

associated with the adoption of the new tax. Our sectoral model stressing elite competition

dovetails with other recent work on democratization and redistribution, such as work by

Ansell and Samuels (2015), but diverges from models by Acemoglu and Robinson (2000)

and Boix (2003), which posit the existence of strong opposition of landed elites to new forms

of taxation and redistribution. We also show that the political conflict at the time of income

tax adoption was primarily a conflict along sectoral lines that differed from later class-based

conflicts over the progressivity of the tax system. While the most divisive political issues at

the time of income tax adoption concerned horizontal redistribution across sectors, the most

significant political conflict around World War I concerned vertical redistribution between

high and low income earners as well as issues concerning the fairness in the distribution of

the burden of war (Scheve and Stasavage, 2010).

To develop and test these arguments, the remaining part of the paper will be organized as

follows. We begin with a descriptive exploration of cross-national patterns in the adoption

of income taxes in the Western World. We then formulate a number of arguments that

seek to identify the economic and political considerations of political elites in countries with

restrictive suffrage about the desirability of income tax adoption. We first present a number

of hypotheses about the economic considerations of the owners of different assets with respect

to the advantages of the new income taxes. Next, we explore how different electoral rules

modified these economic calculations. Following the cross-national tests, we turn to the

analysis of one case to further test our theoretical hypotheses. We examine the adoption

of the Income Tax in Great Britain in 1842, by documenting the Parliamentary discussions

that preceded the adoption of this legislation and by modeling the determinants of support

of the new tax. Our roll-call vote analysis examines the relationship between district-level

economic conditions, biographic characteristics and partisanship of MPs on parliamentary

support for the new legislation. We conclude by offering evidence of the tax incidence of the

income tax across sectors. Some final remarks follow.

6

2 The Adoption of the Income Tax

Beginning with the first two decades of the nineteenth century, governments began to

take on an increased number of obligations, which included investment in the development

of infrastructure, the improvement in the provision of public health and the expansion of

the provision for education (Lindert, 2004). To finance these new responsibilities, politicians

sought to find ways to diversify the sources of revenue that could be enlisted for tax purposes.

The list of possible sources for fiscal revenue considered at the time was, from a contemporary

perspective, heterogeneous. Possible taxes that were under consideration at the time included

taxes on windows and buildings, taxes on luxuries, taxes on matches, taxes on business and

so on. Nevertheless, proposals to impose taxes on income were only one among the many

alternatives that was regarded as a source of possible tax revenue.

Recommendations to adopt a tax on income met with strong skepticism and opposition

(Daunton, 2001). Opponents of the new tax invoked both political and administrative dif-

ficulties associated with its implementation. First, considerable difficulties existed about

ways in which “income”, the new category that was at the basis of the new tax, could be

ascertained and what its different components that could be assessed for tax purposes should

be. Considerable uncertainty existed as to whether the different sources of income were to

be taxed separately, in separate schedules, or whether they were to be taxed jointly. Coun-

tries that pioneered income tax adoption chose very different solutions to this question, with

Prussia and Britain establishing the two possible extremes with respect to the technology of

taxation of the new income tax. In Britain, all different components of income were taxed

in different schedules in an effort to strengthen the monitoring capacity of the state (Mus-

grave, 1969). Prussia, by contrast adopted what was known as a “synthetic” income tax

which required all taxpayers to sum up their disparate sources of revenues for tax purposes

(Popitz, 1926). Given the difficulties in defining income, the unit of the new tax, consider-

able uncertainty existed about the designation of the authority that had the responsibility

to verify the tax obligations of different individuals (Daunton, 2001). Finally critics of the

7

new tax objected that the incentives of individuals to comply with the new tax obligations

would be very low.

By invoking a combination of these arguments, in some countries opponents of the new

tax succeeded in blocking its adoption. In other countries, proponents of the new income tax

triumphed against considerable political opposition. Examples of the latter are conservative

Prime Minister Robert Peel who succeeded in forging a political coalition in favor of an

income tax, or Robert Miquel, the architect of the Prussian income tax who engineered

a political coalition in favor of the new tax. Miquel’s proposal for an income tax was

adopted with a considerable political majority of 322 votes in the Prussian lower house,

with only 40 deputies opposing the law. Support for the new tax came from Conservatives,

National Liberals and Free Conservatives, three of the largest political parties on the right.

Similarly, Peel built an overarching conservative majority to adopt the new income tax in

1842. Appealing to the “superior goal” of fiscal responsibility, Peel reintroduced the income

tax in Britain with 97.5 percent favorable conservative vote.

The Prussian and British experiences were not exceptional. Figure 1 presents descrip-

tive information about the timing of the adoption of income taxes across seventeen Western

economies during the period between 1842 —the year when Britain introduced the first per-

manent income tax—and 1939, the year when Switzerland introduced a permanent income

tax at the federal level. The horizontal axis in each graph displays time. The vertical axis

represents the scope of suffrage. For each country, the solid line presents year-to-year infor-

mation on the scope of suffrage. The vertical dashed line indicates the year of the permanent

adoption date of the income tax.2

Figure 1 Here

The descriptive information presented in Figure 1 reveals a number of surprising patterns

concerning the relationship between the adoption of income taxes and democratization. First

2Both Franchise and the Date of Adoption of the income tax are drawn from Aidt and Jensen (2009).Franchise refers to the electorate (for parliamentary elections) in percentage of the enfranchised age and sexgroup, before women’s suffrage, male population only. Coded 0 if no elections took place.

8

and surprisingly, we find that a significant number of countries adopted the income tax at

a time of very restrictive suffrage. These countries include Britain, which pioneered the

modern income tax in 1842, Austria (1849), Italy (1864), Japan (1887), Norway (1892),

Netherlands (1893), and Sweden (1902). In Britain, the earliest adopter, only 13 percent

of the population was eligible to vote in 1842. In Austria-Hungary, the immediate follower,

curia representation, which guaranteed disproportioned political leverage to the landed elites,

was still in place. In Italy, only 8 percent of adult population was entitled to vote in 1864, the

year when income tax was adopted. By contrast, countries that had established extensive

suffrage rules early on during the period of democratization delayed the adoption of income

taxes by several decades.3 In France proposals to adopt income taxes that were advanced on

repeated occasions during the 19th century encountered political opposition, and an income

tax was adopted only in 1911. These descriptive patterns suggest that early democratizers

faced more difficulty in reaching a political consensus about the new tax.

How effective were these early taxes in raising fiscal revenue? Were these early income

taxes just scraps of paper adopted for purely ceremonial purposes alone? To examine these

questions, we explore the consequences of the adoption of income taxes on direct tax revenue.

If income taxes were purely ceremonial, we should see no relationship between income taxes

and the level of fiscal revenue.

Figure 2 Here

Figure 2 plots the evolution of direct taxation as a share of central government revenue

for the early adopters in Western Europe: the United Kingdom, Austria-Hungary, Italy,

Norway, Netherlands and Sweden. All these six countries adopted the income tax under

very restrictive conditions of franchise. Certainly, the initial tax rates adopted were low

for modern standards. The top rates never exceeded 5% (Kennan, 1910; Popitz, 1926).

Yet these taxes were progressive from the very beginning, imposing a higher burden on

3New Zealand is the only exception to this rule: this country early adopted the income tax while alreadyhaving extended franchise.

9

higher income individuals, who were very remarkably concentrated at the time. To get a

better understanding of their fiscal impact, we plot the total evolution of direct taxation 15

years before and 15 years after the adoption of the income tax. Year 0 denotes the year in

which the income tax was adopted by each country. Figure 2 suggests the existence of a

structural break in direct taxation revenue of approximately 7 points precisely at the year of

adoption. Certainly, the effect we observe in Figure 2 is not trivial, and poses the question

of why this tax was adopted in the first place.4 To solve this question we need to unravel

why a consensus about the adoption in income taxes and investment in the development

of fiscal capacity emerged in some countries, but not in others. Next, we formulate two

complementary hypotheses that account for the adoption of income taxes under conditions

of limited political suffrage.

Our first hypothesis is that the adoption of the income tax results from a political bargain

in which owners of different assets attempt to minimize their own tax burden, while shifting

a higher burden of taxation onto the competing sector. This is, in essence, an argument of

inter-sectoral redistribution through taxation. Specifically, we conjecture that in economic

conditions that increased the bargaining power of landed elites (or old elites), owners of fixed

assets were more successful in imposing the new tax on owners of mobile assets (urban elites

or new elites). One such factor that increased the bargaining power of landed elites is the

level of landholding inequality. Our second hypothesis is that the calculations of political

elites about the desirability of the new tax were influenced by pre-existing electoral rules.

Some of the electoral provisions that were in place in the countries with limited suffrage

conditioned political participation on payment of direct taxes. The adoption of income

taxes and its addition to the menu of existing direct taxes had the potential to increase the

4We do not have comparable data for Japan and New Zealand, also early adopters. For these countries,we only have post-adoption data. However, these suggest that the early income taxes in both countries werealso consequential. In 1890 Japan, five years after the income tax was adopted, this tax represented 2.3% ofdirect taxation (Ranis, 1959). This number grew to 10.7% only ten years later (ibid.). Actually, if we countthe corporate tax as an income tax (as it was in the UK), these figures are 3% and 22.6%, respectively. InNew Zealand, the income tax represented 17.6% of total direct taxation only two years after its adoption(Statistics New Zealand, 1893). These figures suggest that the income tax was neither scraps of paper amongits early adopters outside Europe.

10

existing inequalities of political representation, by linking electoral rights to the payment

of taxes. Incumbent elites (either landed or urban) elected under these restrictive electoral

rules recognized the political advantages of a new tax that could raise the barriers to political

participation of low income voters, and, accordingly, supported its adoption.

2.1 Economic considerations for income tax adoption: sectoral

incidence

Prior to the adoption of the income taxes, governments relied on very heterogenous taxes

to collect the fiscal revenue that was necessary for the financing of public and military

expenditures. These taxes included poll taxes, property taxes, and a variety of consumption

taxes (Popitz, 1926; Seligman, 1911). The income tax promised to bring two changes to

this policy landscape. The first innovation was that of uniformity. The income tax sought

to replace a variety of different sources of revenue with one single category: income. The

income tax also carried the promise of bringing in higher levels of fiscal revenues, by enlisting

resources that had previously evaded tax authorities. These previously untaxed resources

were distributed unequally across sectors. Precisely this unevenness in the location of the

source of the new tax revenue became a source of distributional conflict at the time of income

tax adoption.

Two questions were at the center of the inter-sectoral conflict. The first was the desir-

ability of the taxation of income, as compared to the taxation of other sources of revenue.

The second question concerned the inter-sectoral allocation of the tax burden (i.e. the in-

cidence of taxation), which was particularly relevant in a period of major economic change

associated with the industrialization of Western economies. We hypothesize that owners of

fixed assets considered the taxation of income as a much more attractive fiscal instrument

as compared to owners of mobile assets. First, owners of fixed assets favored the reliance on

income as the category that defined the tax liability of the individuals. That allowed fiscal

authorities to draw on and enlist new sources of revenue, such as profits, that had previ-

11

ously escaped taxation and that were more likely to be found in the newly rising industrial

sector. A brief examination of the provisions of income taxes adopted by these countries

lend support to the economic hypothesis. The 1842 income tax adopted in Britain did not

tax capital gains, not even from real estate, whose major owners were landowners, but did

levy income stemming from trade, thus targeting the new industrial elites (Daunton, 2001,

pag.205). The income tax adopted by Austria in 1849 also exempted land and buildings,

but imposed a rather progressive rate of taxation on profits and other professional incomes.

The latter tax rate rose to 20% during the first years of existence of the new tax (Kennan,

1910; Seligman, 1911). The Italian income tax adopted in 1864 did not tax incomes derived

from the ownership of land, but taxed profits and capital investments.

Secondly, owners of fixed assets also hoped that the introduction of the new income tax

would relief taxation of other assets, such as land. By contrast, owners of mobile assets were

more likely to express worries and concerns about the desirability of the introduction of a

new tax on income. The income tax would not only target profits (previously untaxed) but

also make manufacturing products more expensive in a context of increasing international

competition. Accordingly, owners of mobile capital were likely to oppose the new income

tax.5

Importantly, considerations about capital flight might account the lower tax rates that

we find in the initial income taxes. Concerns about capital flight were raised in numerous

occasions in the debates that preceded the income tax act in Westminster. “Let them harass

the manufacturer by taking and taking by an inquisitorial and offensive process by taking a

portion of those profits [...], he would probably transfer to other lands those pursuits which

they apparently contemplated with such strong disapprobation. [...] The wealthy capitalist

the skillful operative might quit their shores” (Sir W. Clay, merchant and Whig MP for

Tower Hamlets, HC Debate March 21, 1842). We interpret the risk of capital flight as an

5The economic hypothesis is consistent with Bates and Donald Lien (1985) seminal work, which alsodiscuses economic losses incurred by owners of fixed capital as a result of the increase in the importance ofmobile capital owners. Precisely, we claim that the income tax helped fixed-capital owners to rebalance theireconomic and political power vis-a-vis the new industrialists.

12

upper bound to the tax rates that the landed elites could impose the new sectors. External

constraints, however, should not prevent landed elites from levying tax rates on capital above

those on land whenever they are powerful enough, as the experience in Imperial Germany

suggests (Hallerberg, 1996).

This first hypothesis postulates the existence of sectoral conflict between landed and in-

dustrial elites over the adoption of the new income tax, much in line with the logic of sectoral

competition proposed by Ansell and Samuels (2015). Economic conditions are likely to af-

fect the relative bargaining power of owners of different assets and increase the probability

of income tax adoption. One such economic condition is the level of land inequality. Higher

levels of landholding inequality were more conducive to the selection of wealthier, conserva-

tive politicians (usually the landowner of the respective district) who were supportive the

adoption of the new tax with visible economic and political advantages. By contrast, we

expect to find delays in the adoption of the income tax in countries with lower levels of

income inequality, due to a lower bargaining power of landed elites in these contexts.

Note that our theoretical hypothesis about the effect of rural inequality about the ability

of elites to reach compromises about the new income tax is the opposite of the prediction in

the existing literature (Sokoloff and Zolt, 2007). In our account, rural inequality, everything

else constant, is expected to have a positive effect on the probability of income tax adoption.

This hypothesis is consistent with Hallerberg’s analysis of tax competition in the Wilhelmine

Germany, 1871-1914 (Hallerberg, 1996). Hallerberg documents how German states with

limited suffrage and strong landed elites in Parliament —with Prussia as the paradigmatic

case—were successful in shifting the incidence of direct taxation onto mobile capital owners.6

By contrast, states with more open franchises and a weaker representation of landed elites

in Parliament (such as Baden) increased the tax rate on mobile capital and labor together

with the land tax.7

6See also Hallerberg (2002) for a case-specific analysis.7It could be argued that industrial elites could pass on the burden of taxation to consumers by raising

intermediate or final good prices. For our argument to hold, we only need that the incidence to fall even ifonly partially on the modern sector, whose rents had gone virtually untapped under the old fiscal system.

13

2.2 Political considerations for income tax adoption: the effects

of electoral institutions

In many Western countries, political negotiations about the adoption of the income tax

took place in an environment characterized by restrictive electoral rules. Many of the elec-

toral systems that were in place in nineteenth century Western World used criteria such

as the wealth of different individuals or the payment of taxes to allocate electoral rights.

An important issue facing politicians that adopted the income tax was whether the new

law could enhance the existing electoral rights of incumbent elites, given existing electoral

rules. In this section, we consider these additional calculations about the political effects of

the income tax. Specifically, we assess the political calculations of incumbent elites in the

presence of a restrictive electoral rules that were in place at the time: the vote-tax link.

We define the vote tax link as an electoral provision that used payment of direct taxes

to reduce political participation of lower-income individuals. Specifically, the vote-tax link

is an electoral institution that conditions the right to vote on the prompt payment of direct

taxes. The vote-tax link was widely used in Europe, but also in the United States and Japan,

as listed in Table 1. All these countries, at one point, conditioned franchise for the national

elections on direct tax payments.8

Table 1 here

The existence of the vote-tax link establishes a convenient political tool for incumbent

political elites to restrict political participation. This electoral institution offered a tech-

nology that could be easily reconfigured as a tool of political exclusion once the new tax

was in place. The vote-tax link created two distinct opportunities of political exclusion of

That is, partial incidence suffices to make the adoption of the income tax in the interest of landed elites. Thestrong opposition to the income tax of the British MPs representing the industrial interest (further detailsbelow) suggests the passing the incidence of taxation entirely onto consumers was virtually impossible. Weare grateful to an anonymous reviewer for pointing out this necessary condition for our argument to hold.

8Table A-2 in the Online Appendix supplements this information by indicating in which countries thevote-tax link was actually effective when the income tax was adopted.

14

lower-income individuals. On the one hand, as the vote-tax link made the cost of political

participation visible, it was also likely to discourage demands of low-income voters for the a

further expansion of suffrage. On the other hand, the pecuniary cost of political participa-

tion was likely to depress the participation rates of the poorer strata among those qualified

to vote (i.e. the urban middle classes). With respect to fiscal outcomes, the vote-tax link

was seen as a guarantee of fiscal conservatism, which was widely embraced by elites during

the second half of the nineteenth century in the Western World. At the time, it was believed

that voters would refrain from demanding high expenditure programs (and thus, high taxes)

only if they were to fund these programs out of their own pockets (Daunton, 2001; Aidt,

Daunton and Dutta, 2010). Under that belief, a linkage between voting rights and income

taxation was seen as a natural mechanism to prevent overspending.9 In the words of William

Gladstone: “it is desirable in a high degree, when it can be effected, to connect the posses-

sion of the franchise with the payment of taxes”.10 For all these reasons, we expect that the

pre-existence of electoral rules that link tax payments to electoral participation should cre-

ate political conditions that facilitate the adoption of the income tax in a proto-democratic

setting.

Table 2

Table 2 summarizes our hypotheses about the desirability of an income tax in countries

with restrictive suffrage. These hypotheses present several plausible micro-logics that may

explain the early adoption of income taxes by non-democracies. We have identified the

economic and political considerations of owners of fixed and mobile assets, respectively.

With respect to economic gains, we have hypothesized that owners of fixed assets are more

likely to gain from the adoption of an income tax than owners of mobile assets. We have also

hypothesized that pre-existing electoral rules may offer political advantages to insiders once

the new income tax is adopted. By amplifying inequalities in representation, a new income

9 This same goal still inspired electoral provisions in many democracies in the late twentieth century(Ardanaz and Scartascini, 2013).

10Quoted in Matthew (1988, pag.127).

15

tax is likely to reinforce and enhance political advantages of incumbent elites. In this regard,

we expect that both owners of fixed and mobile assets should benefit from the adoption of

income tax in presence of this very particular electoral provision, the vote-tax link.11 These

common political gains might forge an informal alliance among elites from different sectors

who otherwise may disagree about the economic consequences of the new tax.12

3 A test of the economic and political gains of income

tax adoption

Next, we test for the economic and political gains hypotheses in modeling the adoption

of permanent income taxes in the Western World.13 We follow (Aidt and Jensen, 2009) and

establish the defeat of Napoleon in 1815 as the starting point of the analysis. The Napoleonic

Wars propelled the adoption of income taxes in various countries of the European continent:

the United Kingdom and France, but also the Austrian Empire, Belgium, Denmark, the

Netherlands, and Norway. After Napoleon’s defeat, the income tax was eliminated every-

where, but the precedent (that is, the technology) and the risk of its permanent adoption

11 Notice that industrial elites had to experience some gain from the new tax to secure quasi-voluntarycompliance, as it had been procured in 1799 when the income tax was temporary adopted in the UK towage the Napoleonic wars (Levi, 1988). An income tax would have been just too hard to administer withthe frontal opposition of the industrial elite.

12The vote-tax link is conceivable not the only electoral provision that might give a political advantage tothe incumbent elites. Electoral provisions that assign voters to different tiers in Parliament based on theirtax payments might do as well. This is the case of Prussia and other German principalities between 1871 and1913, where tax payments led to huge parliamentary malapportionment and facilitated the early adoptionof the income tax. Nevertheless, tax-based parliamentary malapportionment is an electoral provision thatis relatively rare in national parliaments in the Western World: Austria-Hungary had an estate-parliamentthat reserved one of the four (later five) curiae to big tax payers. Yet, the curia system was adopted in1861, only after the income tax had been adopted in 1849. Finland had a curia system similar to that inAustria-Hungary, but it was in place only during the period when Finland was under Russian control. Basedon the income tax adoption years, milder versions of tax-based malapportionment, such as the “plural vote”in Belgium and France, by which high tax payers receive a few additional votes, might not provide the sameclear incentives to the ruling elites.

13Australia, Austria-Hungary, Belgium, Canada, Denmark, Finland, France, Germany, Italy, Japan,Netherlands, New Zealand, Norway, Sweden, Switzerland, United States, United Kingdom.

16

remained.14

The structure of the data is Binary Time Series Cross Sectional (BTSCS). Beck, Katz

and Tucker (1998) prove the virtual equivalence between BTSCS and survival models. As

such, the adoption of income tax can be fitted with a standard logistic regression. We only

need to add a time trend of the years at risk of adopting the new tax to account for the

conditional hazard rate. For that purpose, we follow Carter and Signorino and fit a cubic

polynomial approximation to account for duration dependence (Carter and Signorino, 2010).

The Online Appendix includes a replication of the analysis using frailty models that allow for

unit-specific hazard rates, splines instead of flexible polynomials (Beck, Katz and Tucker,

1998), and complementary log-log models that do not assume ex ante any shape of the

hazard rate. The exact model specification is as follows.15

P (yit = 1|xit, yit−1 = 0) =1

1 + e−(xitβ+H(t−ti))(1)

Expression (1) models the adoption of permanent income tax y by country i at time t

(provided it has not been adopted yet) as a function of time-variant covariates xit and a

smooth function of the number of years a country has been at risk of adopting the income

tax: H(t− ti) = t+ t2 + t3, where t denotes the time elapsed since the observation enters the

sample. By expression (1), once a given unit adopts the income tax, it drops the sample.

Table 3 here

We begin the cross-national analysis by replicating Aidt and Jensen’s (2009) “suprising”

finding: namely, countries with limited franchise and proto-democratic institutions, proxied

by the Polity IV (Marshall and Jaggers, 2000), pioneered the adoption of the income tax. The

effects are robust to the inclusion of standard socio-economic covariates, a Europe fixed effect

14For those countries that were not independent in 1815, we assume that they only become at risk assoon as they gained independence. Again, we follow Aidt and Jensen’s (2009) approach.

15We borrow notation from Aidt and Jensen (2009).

17

that accounts for unobserved idiosyncrasies of the Old Continent, and, more importantly,

the usual suspect of any major innovation: i.e. war (Dincecco, 2011; Tilly, 1990).16

The remaining models in Table 3 and 4 seek to shed light on this counter-intuitive result

by examining the political and economic determinants of the adoption of income taxes. The

models reported in columns 2 to 5 in Table 3 assess the economic gain hypotheses, which

conjectures that politicians with ties to owners of fixed factors should be more supportive of

the adoption of an income tax as compared to politicians representing mobile assets. As the

income tax taxed income that was previously undetected, it imposed a higher burden on the

owners of mobile assets as compared to owners of fixed assets. Secondly we have argued that

expectations of future changes in the balance of power among owners of fixed and owners of

mobile assets are likely to affect their relative preferences about the desirability of the new

income tax. Particularly, we conjecture that in a time of major economic change landed elites

in limited democracies favored the adoption of income taxes because the incidence of this

tax fell mainly on the new urban elites. The income tax was, in other words, a mechanism

of redistribution of the tax incidence across sectors.

The observable implication of the economic gain hypothesis is that economic and political

factors that strengthen the relative bargaining power of owners of fixed assets are likely to

increase the probability of income tax adoption. We follow the empirical strategy in the

literature on democratization and proxy the economic power of landowning elites using

a measure of land concentration (Ansell and Samuels, 2015; Boix, 2003; Ziblatt, 2008):

specifically, we use the share of farms owned by families, as constructed by Vanhanen (2003)

from the agricultural censuses.17 We expect a positive relationship between this variable

16Refer to the Online Appendix for further specifications and measures of war-related activity.17Vanhanen (2003) offers decennial values of land ownership concentration. To impute in-between years,

we pursue flat-steep-flat coding. That is, we assign the last available value in the series to all subsequentobservations for which data is missing. Suppose that we have family farm information for 1900, 1910, and1920. From 1901 to 1909, we set family farm equal to its value in 1900, and from 1911 to 1919, we setfamily farm equal to the 1910’s record. Additionally, if the first observation for family farm follows thefirst observation for franchise, the first available value of family farm is extrapolated toward the past. Forinstance, if the first records of family farm and franchise for a given country are 1850 and 1820, respectively,family farm between 1820 and 1850 is set to its value in 1850.

18

and adoption of the income tax. The model reported in column 2 tests and confirms this

proposition. We find that countries with high levels of Landholding Inequality are more

likely to adopt the income tax.

One might argue that a broad range of factors proxying economic development, such as

the GDP per capita, population, or the level of urbanization, “the forces of development”

(Aidt and Jensen, 2009, pag.166), may affect the ability of a given country to adopt the

income tax as a permanent tax. Similarly, the prior experience with income taxation (either

as a temporary national level tax or as a local level tax), or the availability of other sources

of revenue may affect the probability of the adoption of a permanent income tax. Thus, one

conjecture is that the previous experience with income taxation at the national or local level

may facilitate the permanent adoption of this tax,18 whereas the presence of alternatives

sources of revenue may slow down the adoption of the income tax. All models control for

the former battery of factors, while column 3 specifically evaluates alternative sources of

revenue by assessing the relationship between tariffs and the adoption of income taxes. Here

we use Ad-Valorem Equivalent (AVE) tariffs (Lampe and Sharp, 2013), which are calculated

as the ratio of customs duty revenue to total imports for domestic consumption.19 We might

expect that when AVE tariffs are high enough, there might be no fiscal justification to adopt

an income tax. This variable holds a positive, yet not significant coefficient, contradicting

the hypothesis that higher levels of tariffs substituted for the need of countries to adopt

income taxes. This result is, however, not surprising to someone familiar with fiscal policy

development in the early nineteenth century. At the time, the income tax was not understood

as being a substitute, but as a complement of trade taxes, as the incidence of both taxes

fell on the trading sectors (Daunton, 2001, pag.83). Importantly for the economic gain

18Population and GDP per Capita (Aidt and Jensen, 2009) are linearly interpolated as to maximize thesimple size. Urbanization (Vanhanen, 2003) is interpolated following the flat-steep-flat strategy described inft. 17, as we use it to compute Rural Inequality, explained below. Data for local and temporary nationalincome taxes are drawn from Aidt and Jensen (2009).

19We lack tariff data for Finland, and Austria-Hungary prior to 1849. This explains the smaller N incolumn 3. The logarithmic transformation of this variable reported in column 3 and 5 seek to minimize theinfluence of extreme values.

19

hypothesis, Landholding Inequality keeps a positive sign and remains statistically significant

when we control for this alternative source of revenue.

Columns 4 and 5 repeat the same analyses using the measure of land inequality pro-

posed by Ansell and Samuels (2015): Rural Inequality. These scholars suggest adjusting

Vanhanen’s “family farms” for the actual density of rural population. “There is a possi-

bility that, even in countries where families own most of the cultivable land, most of the

rural population might not live on a family farm” (Ansell and Samuels, 2015, pag.101). To

properly measure overall inequality of agricultural landholding, they recommend to adjust

the measure of Family Farms by the size of the rural population, thus calculating “Rural

Inequality” as (1-Family Farms)(1-Urbanization)/100, where Urbanization, also taken from

Vanhanen (2003), is the percentage of urban inhabitants (ibid.).20 The relationship between

this variable and the probability of income tax adoption is similar to Landholding Inequal-

ity. The coefficient for Rural Inequality is positive and highly statistically significant in both

models reported in columns 4 and 5. The smaller magnitude of this coefficient is just a

byproduct of the wider range of this variable, which increases by two orders of magnitude.

Altogether, the results in Table 3, advance evidence consistent with the economic gains

hypothesis while confirming the surprising pattern identified by Aidt and Jensen. We in-

terpret these findings as suggesting that in conditions of restrictive political suffrage higher

rural inequality increased the ability of landholding elites to adopt the income tax to shift

the burden of taxation to the rising trading sector.

Next, we evaluate the political gains hypothesis, which conjectured that the calculations

of owners of different types of assets about the desirability of the income tax were affected

by an electoral institution linking individual voting rights to prompt tax payments, or the

vote-tax link. Specifically, we expect to find a positive relationship between this electoral

provision and the adoption of the income tax. Politicians elected under this type of pre-

20Notice that Rural Inequality exhibits higher over time variation as compared to Landholding Inequality,because rural population changes faster than land ownership. We normalize Rural Inequality to 100 so thatit has the same range than Landholding Inequality.

20

existing electoral rule are likely to regard the new income tax as a policy instrument that

maintains and reinforces their political advantage. Table 1 in the previous section reports

the countries and years in which the vote-tax link was in place in the countries included in

our analysis.

Table 4

The models in Table 4 build on the specifications presented in Table 3 and include controls

for socio-economic conditions, political institutions, warfare, and, importantly, landholding

inequality. The results reported in column 1 indicate that the existence of institutions

linking tax-payments to political participation facilitated the adoption of the income tax.

The coefficient is positive and statistically significant at 95%. In words, countries with pre-

existing electoral provisions conditioning the right to vote on the prompt payment of direct

taxes were more likely to adopt the income tax. In column 2, we assess how robust the

vote-tax link is to the replacement of Landholding Inequality for Rural Inequality. Results

are virtually identical to those in column 1. That is, the vote-tax link positively predicts the

adoption of the income tax.21

Finally, we seek to examine whether the vote-tax link is robust to the consideration of

an electoral provisions that is associated with the old regime (Mares, 2015) and is argued to

predict the adoption of the income tax: the secret ballot (Aidt and Jensen, 2009).22 Column

3 of Table 4 confirms that the positive effect of the vote-tax link is robust to the presence of

secret ballot, whereas the latter’s is not statistically different from zero. This result confirms

the unique political advantage that the Vote-Tax Link offers to the ruling class: namely, it

can exacerbate inequalities in political representation by conditioning political rights on the

prompt payment of direct taxes. We claim that this political gain entered the calculations

of both landed and industrial elites in pushing for the adoption of the income tax.

21These and Table 3 results are robust to country-random effects as well as shared frailty (refer to OnlineAppendix). Importantly, the likelihood-ratio test for significant frailty is negative, suggesting that coefficientsare not biased by country-level unobservables.

22Secret Ballot is a time-varying indicator, and it is drawn from Aidt and Jensen (2009).

21

4 The British Income Tax Act of 1842

To further test our hypotheses, we examine the adoption of the permanent income tax

in Britain in 1842. In Britain, the proposal for the adoption of the new tax originated with

and was pushed through the parliamentary deliberations by the conservative government of

Robert Peel. The examination of this crucial case provides us with the opportunity to test

the micro-foundations of our argument. We will begin by examining the most important

issues that were contested during the political deliberations over the new tax through a

qualitative analysis of the parliamentary deliberations. Next, we will test our argument

about the determinants of political support of an income tax through a quantitative analysis

of the decisive vote in Westminster that led to the adoption of this tax. We will further

probe our argument, by examining the incidence of the tax burden on different sectors after

the adoption of the new tax.

During the parliamentary debates over the adoption of the income tax, both opponents

and supporters of the new income tax noted that the new technology of taxation, which

assessed income had the potential to raise higher revenue when compared to existing taxes.

Yet both critics and supporters of the new income tax argued that the close monitoring of

income —which was required by the new tax—was also inquisitorial in nature. The tax

was attacked as a “threat to personal liberty, for it would require inquisitions scattered

throughout the country to enquire into private affairs”(Daunton, 2001, pag.82). Lord Dal-

meny, a liberal politician considered that the new income tax was “an imposition of the most

grievous impost accompanied by the most odious processes of inquisitorial investigation that

can be inflicted on a free nation” (HC Debate March 18, 1842). Others noted that the new

methods of income assessment violated individual privacy, while several members of parlia-

ment doubted the good will and intentions of the tax assessors. As one MP argued, “spies

and informers must be set to work, and a pecuniary interest must be given to them in the

additional amount they procured to the Exchequer” (HC Debate April 4, 1842).

We find strong and systematic evidence of a political conflict between politicians rep-

22

resenting rural and industrial interests during the political deliberations over the adoption

of the new tax. Robert Peel himself admitted that the new tax affected owners of capital

severely. As Peel stated, “I admit that the tax may press with additional severity on account

of the uncertain future of profits on that property which is derived from trade” (HC Debate

March 18, 1842). Other liberal politicians opposing the new tax expressed similar reserva-

tions to the new income tax. The political intentions of the new income tax, these critics

argued, were to “seize the profits of trade, to submit it to the legalized persecution of the

tax-gatherer. It is to uphold the monopolies of the agriculturist and planter, that he calls on

us to wring the pittance from impoverished industry, and scatters a horde of officials over the

land to violate the sanctuaries of private life” (HC Debate March 21, 1842). Similarly, “It

was felt in commercial towns that it was unjust to tax the hard-earned produce of industry

on the same footing as the income of realized property, annually received without toil and

without risk. It is felt too that the real burden of the income-tax, a Government inquiry into

men’s private affairs, falls exclusively upon the trading classes” (italics added, HC Debate

April 13 1842).

A pervasive criticism of the income tax was that the tax imposed an unfair burden on

production and enterprise and that the incidence of the burden on different types of income

was unfair (Daunton, 2001, pag.83). Critics argued that “no tax could be devised which

could operate more unequally, more unjustly and more oppressively”. Opponents of the

new tax considered that it was unjust to impose the same tax burden on different forms of

income, as the latter originated through different processes. As liberal politicians argued,

imposing the same rate on “ ‘spontaneous’ income, on permanent property which could be

passed on from father to son and on precarious incomes based on personal exertion where

money had to be set aside to support dependents” was unjust. It was equally unjust to tax

“fluctuating income in the same proportion as permanent income” (HC Debate March 23,

1842). Yet liberals believed that the reasons for this inequality were political in nature and

that they could be attributed to the disproportionate influence of landed elites. “The reason

23

why the man of fluctuating income was severely taxed was, that the man of landed property

might escape taxation. The right hon. Baronet Peel did not dare to tax the landowner and

he therefore revenged himself upon the unhappy wretch who owned nothing but the faculty

of honestly earning high livelihood.”(HC Debate March 23, 1842).

In opposing the new income tax, members of parliament linked to Britain’s rising manu-

facturing sectors proposed several alternatives to the new tax. One set of proposals involved

scrapping the income tax altogether and replacing it with alternative taxes, while others

proposed changes to the income tax which lightened the burden of the new tax on the man-

ufacturing sector. Some of the proposals to augment the revenue-raising capacity of existing

taxes recommended the remission of duties on corn, the imposition of a legacy duty on

landed property or the increase in value of assessed taxes (HC Debate April 8, 1842; HC

Debate April 13, 1842). All alternatives to the income tax met however with defeat during

the deliberations. The remission of duties on corn, was rejected because it was believed that

such a tax could not generate the same level of income as compared with the income tax (HC

Debate April 13, 1842). Proposals to increase the level of assessed taxes, were also opposed

and eventually defeated on the grounds that these were even more burdensome as compared

to the income tax. “A great deal has been said about the inequality of an income-tax, but

the assessed taxes pressed with much greater inequality upon commerce and manufactures

than the income tax” (HC Debate April 13, 1842).

Alternative proposals made by Whig politicians recommended to lighten the tax burden

on the manufacturing sector within the income tax. This could be achieved either through

the imposition of different tax rates on funded or unfunded wealth or the exemption of various

incomes from the incidence of the new tax. Some of these proposals recommended to exempt

manufacturing interests altogether from the payment of the new tax (HC Debate April 29,

1842). Others recommended to exempt income derived from “alum mines, waterworks,

streams of water, canals, railways, bridges and ferries” from the payment of the new tax.

Finally, other proposals that were widely deliberated at the time recommended to impose

24

a lighter tax burden on incomes that were volatile, as compared fixed assets such as land.

Such proposals were, however defeated (HC Debate April 29, 1842). The argument that

eventually prevailed during the deliberations was that returns of land were as volatile as the

returns from commercial resources, making the differentiation of taxation for income from

volatile versus fixed assets unjust. As a result of the defeat of all these alternatives, Peel’s

income tax, a policy that was more favorable to the interests of landed elites was ultimately

adopted.

The marked division between politicians representing the interests of different sectors

was reflected in the final vote of the income tax, in May 31, 1842. Next we model the votes

of the different representatives as a function of partisanship, the background of the MP, and

the socio-economic characteristics of the districts that they represented. The results suggest

that, controlling for landholding inequality, district-level income and poverty, and the MPs

professions and social class, representatives of industrial districts opposed the income tax

while representatives of agricultural districts favored it.

4.1 A Roll Call Analysis of The 1842 Income Tax Bill

The parliamentary debates about the adoption of an income tax took place during a brief

period between March and May 1842. The third and last reading of the Income Tax was

held on May 31, 1842. 388 members of the parliament participated in this decisive roll call

votes, representing English, Welsh and Scottish constituencies. We collected biographical,

district and partisan information for these politicians (and were able to find this information

for 384 MPs). Our empirical analysis examines the relationship between these variables and

the vote for the income tax bill. We proceed stepwise. First, we evaluate the effect of district

characteristics, then biographical, and, lastly, partisanship.23

Table 5 here

23Among the early adopters, Great Britain is the only country that lacked a vote-tax link for its nationalelections. Such vote-tax link were, however, in place for local elections. Thus, in this section we only examinethe economic gain hypothesis explaining support for the new tax.

25

Table 5 reports the results of logistic models accounting for support or opposition to the

new income tax. The models reported in column 1 to 3 examine the relationship between

district levels characteristics and support for income tax, while including fixed effects for dif-

ferent regions (England, Wales and Scotland). Column 1 presents the simplest specification.

Here we regress the vote of MP on landholding inequality, controlling for the type of con-

stituency: urban vs. rural. Landholding inequality is computed as a Gini index of land size

at the county-level using the information of the first country-wide agricultural census in the

history of Great Britain, dated as of 1896, that we could locate (Agricultural Returns, 1896).

Using Vanhanen (2003) data, we can establish however that these values were closely corre-

lated to measures of landholding inequality in 1842, the year the income tax was adopted.

The national percentage of family farms in 1842 was 5%, and 8% only in 1890, suggesting

little change in the distribution of land between 1840s and 1890s (this share would reach

52% in 1975). Consistent with our hypothesis, the sign of this coefficient is positive. We

know that Conservatives were likely to be elected in rural districts while liberal MPs were

elected primarily in urban constituencies (Aydelotte, 1977, Table 7.1). Consistent with this

pattern, representatives of urban districts opposed the adoption of the income tax.

In column 2, we investigate the relationship between different economic characteristics of

districts and support for the income tax. Our variables include measures of population shares

of labor employed in agriculture, the old sector, and in retail, liberal professions (including

bankers), and manufacturing industry and industry, the new sector.24 The source for this

information is the 1831 UK Census, as coded by Southall, H.R. et al. (2004). Given that

this source covers England and Wales only, the number of observations in this and the next

column declines. We find that the higher the share of employment in the old-sector, the

more likely it is that a given representative will vote in favor of the income tax. In contrast,

a higher share of workers employed in new-sector activities correlates with a lower likelihood

that the representatives will support the income tax bill.

24Self-employed farmers are excluded, although results do not change when we include them

26

In column 3 we explore an alternative hypothesis, advanced by the literature on taxation

and democratization, which suggests that the income tax is adopted as a policy appeasing

the poor (Acemoglu and Robinson, 2000; Boix, 2003). To assess some of the implications

of these theories, we compute the district-level unemployment rate as well as the share

of individuals in houses rated under 10 pounds (Poor Homes) using information from the

1831 census. We also include a measure of population density, which we are using as a

district-level proxy of income25 One expectation of theories of democratization is that elites

in richer districts are more prone to support the adoption the income tax in an effort to offset

dramatic threats for expropriation. None of the three variables reach, however, statistical

significance at conventional levels. Fears of current or future redistribution do not seem

to explain the motives of support for the adoption of the income tax. Our alternative

explanation, premised on intra-sectoral competition among elites provides a better account

of the variation in political support for the new tax26

The models reported in columns 4 and 5 of Table 5 assess the relationship between

biographical characteristics of a legislator and their vote on the proposal to adopt the income

tax. Biographical information on legislators comes from the Aydelotte (1984) dataset, which

includes a wealth of political characteristics of all members of the British Parliament during

period 1841 to 1847. In column 4 we investigate whether MPs with a vested interest in

the business and financial sectors tend to oppose the income tax.27 The first coefficient is

negative and statistically significant, whereas the second is not statistically different from

zero. However, in subsequent models these effects reverse. We also tests for the relationship

between the aristocratic background of the MP and support for the income tax.28 This

25For a similar use of population density as proxy of income, see Acemoglu, Johnson and Robinson (2002)and Dincecco and Prado (2012).

26Likewise, proxies to district income such as population or population density, always hold positivecoefficients but never reach statistical significance.

27Aydelotte (1984) data includes a battery of items indicating whether MPs had invested in differenttypes of business.

28This variable has seven ordered categories: it takes value 0 for MPs that are unrelated to aristocracy.It takes value 1 for MPs whose family received a nobility title after 1837; value 2 for those who received atitle between 1820 and 1837; and successively until “before 1689”, the seventh category.

27

variable is positive but its statistical significance is not robust across different specifications.

The specifications reported in column 5 include a measure of the MP’s personal wealth.

This is a nine-category variable, ranging from $£2,000$ to $£1,000,000$. This variable

positively predicts the adoption of the income tax. This is consistent with the sectoral

redistribution of tax incidence, as landed elites were richest income group in 1842 Britain.

Finally, the results in columns 6 and 7 investigate the role of partisanship in predicting

the roll call. We use information reported in Aydelotte to construct this variable. Liberal

MPs opposed the income tax adoption (93% of them did). Not surprisingly, this variable

is strongly statistically significant in the expected direction. Importantly, inequality loses

statistical significance once we control for the partisanship of the MP. Due to the strong

alignment between district characteristics and partisanship, we interpret the change in the

coefficient of landholding inequality as an attenuation bias caused by controlling for a post-

treatment effect.

Political parties in the Peelite era were clearly defined. But the skeptical reader might

argue that partisan affiliations suffer from some rationalization of the coder (in this case,

Aydelotte 1984). To address this concern, the model reported in column 7 proxies policy

preferences with the voting record of the MP. To perform this test, we use Aydelotte’s

Guttman scale of the pro-land voting record. This scale is based on the behavior of each

MP in 9 different roll call votes affecting landed interests in the 1841-1847 session.29 This

scale tells us how strongly aligned each MP is with the interests of the landed sector. The

coefficient for this scale behaves as we would expect. The higher the alignment between the

politician and the interests of landed elites, the higher their support for the adoption of the

income tax.

To conclude our analysis of the British case, we investigate the incidence of the new tax

on different sectors during the years following its adoption. The British scheduled income

tax records the sources of various revenues collected by tax authorities, and it allows us

29For details, refer to the original Aydelotte (1984) codebook.

28

to calculate the fiscal pressure faced by each sector. Specifically, the 1842 income tax was

levied under five different schedules. Two of these schedules (A and B respectively) collected

income from the traditional sectors: income from land, real estate and agriculture. Schedule

D, by contrast, collected income from profit and liberal professions. During the decade of

the 1840s, 51.8% of total revenue came from schedules A and B, as compared to 29.4% from

schedule D (Parliamentary Papers, 1857). At first glance, one would see no harm done by the

new law to the industrial elite. However, over time the incidence of the tax was unfavorable

for industrial elites. Figure 3 presents the average tax rate for each of these two sectors for

the first 10 years of the law.

Figure 3 Here

In the first years following the adoption of the new tax, the value of assets in the tradi-

tional sectors doubled compared to the value of assets in the modern sector (Parliamentary

Papers, 1861). This explains why revenue stemming from land surpassed revenue from the

modern sector. However, keeping the value of the assets constant, the effective tax rate for

the modern sector was 30% larger than the rate on the traditional sector.30 Thus, from its

very inception, the incidence of the income tax fell more heavily on the modern sector. Even-

tually, total tax revenue from schedule D surpassed the total tax revenues from schedules

A and B. By 1907, the weight of income raised from schedules A and B accounted for 25.8

percent of the total tax revenues raised by the income tax. By contrast, total tax revenue

from schedule D accounted for 59 percent of tax revenue (Daunton, 2001, Table 7.2).

Altogether, our roll call vote analysis of the distributional conflict over the adoption of

the income tax and Britain’s subsequent experience with revenue collection with the new tax

30 The absolute values of the tax rate and their difference might seem marginal for modern standards.However, these rates were considered high by the time and caused heated debates in Parliament preciselybecause it tapped previously unknown income and had major distributive consequences. Interestingly, bylaw all sectors were required to pay the same tax rate: 7d in the pound on income of 150 l. and upwards (onaverage, that represented a 2.9% tax rate). In practice, the fine print usually benefited those individuals withan economic interest in the traditional sectors. For instance, when income arose under different schedules, theabatement was allowed preferably from the amounts reported in schedules D and E, as claimed by Kennan(1910). Landed elites, in essence, were exempted from paying taxes for their activity in the newer sectors.

29

lends support to the economic hypothesis of income tax adoption proposed in this paper.

Consistent with our expectation, we find that politicians closely aligned with owners of

fixed assets supported the adoption of the new income tax, while politicians representing

the interests of owners of mobile taxes opposed its introduction. The historical information

about the incidence of the tax burden on different sectors suggests that the income tax

generated more revenue from industrial sectors. As such, the operation of the new tax

conformed with the intended preferences of the owners of fixed assets: the income tax led

to fiscal rebalancing, by involving the industrial sector more heavily in the financing of the

public sector.

5 Conclusion

The adoption of the income tax has paved the way for a dramatic increase in the size

of government expenditures and represents a decisive turning point in the evolution of the

modern state. The political decision to assess taxes on income gave governments the means

to tap into a variety of revenues that had been hitherto not assessed. The adoption of

the income tax also went hand in hand with the development of additional institutions

that sought to assess income that originated from various sources and ensure compliance

of citizens with their tax obligations. This paper seeks to provide an explanation of this

decisive turning point in the development of the modern fiscal state.

Existing explanations stress the importance of war as factor accounting for the evolution

of the modern state (Dincecco, 2011; Scheve and Stasavage, 2010; Tilly, 1990). Our paper

reconfirms this long-standing result of the literature. However, wars and preparations for

armed military conflicts are not sufficient in explaining political decisions to adopt the income

tax. The calculations of political elites about the desirability of this radically new instrument

of revenue extraction were affected by domestic economic conditions, on the one hand, and

by existing electoral rules, on the other hand. We provide a set of hypotheses about the most

30

decisive economic and political factors that affected the considerations of politicians about

the desirability of the income tax adoption. We test these hypotheses, by examining the

variation in the timing of income tax adoption across the Western World during the period

between 1815 and 1939.

Our paper identifies two empirical regularities in the timing of income tax adoption across

Western countries during this period. We find that countries with high levels of inequality

and low levels of suffrage pioneered the adoption of the new income tax. These findings

challenge a widely held proposition which suggests that democratization spurs fiscal redis-

tribution. The theoretical expectations of standard models of taxation and redistribution

(Acemoglu and Robinson, 2000; Boix, 2003) do not hold in this important policy domain.

To account for these empirical regularities, we examine the calculations of higher income

individuals in countries with restrictive electoral rules about the desirability of the new tax.

We suggest that owners of fixed and mobile assets will have different expectations about

their future tax liability under the new income tax. While owners of fixed assets supported

the income tax because of the ability of the latter to roll off a higher tax burden on owners

of mobile assets, mobile asset holders were likely to resist the latter. Our empirical analysis

confirms this hypothesis, which posits the existence of a sectoral conflict over the adoption

of an income tax. We find that both structural and short-term economic factors that in-

crease the economic bargaining power of landed elites increased the probability of income

tax adoption.

In many of the early adopters of income taxation, political negotiations about the intro-

duction of a new income tax took place in environments characterized by restrictive electoral

rules. The electoral rules among these non-democratic countries varied significantly in their

institutional design. Our paper examines how a very specific type of electoral rule that was

in place in many European countries prior to the adoption of universal suffrage —the vote-

tax link—affected the calculations of incumbents about the desirability of the new income

tax. Consistent with our expectation, we find that the presence of a vote-tax link increased

31

the probability of income tax adoption in non-democracies.

As an additional test of some of the observable implications of our explanation, we have

collected micro historical evidence analyzing the adoption of the income tax in Great Britain

in 1842. Our quantitative analysis of the final roll call votes leading to the adoption of the new

income tax found that the adoption of the new tax was supported by conservative politicians

from districts with high level of landholding inequality, but opposed by politicians from more

urban interests. The analysis of the parliamentary debates in the British parliament that led

to the adoption of the new tax reveal the existence of a strong inter-sectoral conflict between

representatives of landed and manufacturing elites over the design of the new income tax.

Some of the most contested issue during these deliberations concerned the rates of taxation

of income from fixed or more mobile sources. We show that the fine-print of the final income

tax bill conformed to the intentions of conservative politicians to enlist resources from the

manufacturing sector for fiscal purposes and that alternative tax proposals advanced by

liberal politicians were defeated.

The findings in our paper open up interesting avenues of future research for the politics of

taxation and redistribution in non-democratic regimes, while emphasizing intra-elite compe-

tition as the ultimate cause of major fiscal innovations, much in line with Ansell and Samuels

(2015). Our findings suggest that existing models of redistribution operate with assumptions

about preferences for taxation that are too simplistic and that overlook important sources

of cleavages political incumbents in countries with limited electoral suffrage. We leave it to

future research to examine whether the surprising support of owners of high fixed assets for

policies with high redistributive potential such as the income tax can be also found in cases

of other policy areas.

32

References

Acemoglu, Daron and James A. Robinson. 2000. “Why Did the West Extend the Franchise?Democracy, Inequality, and Growth in Historical Perspective.” The Quarterly Journal ofEconomics 115(4):1167–1199.

Acemoglu, Daron, Simon Johnson and James A. Robinson. 2002. “Reversal Of Fortune:Geography And Institutions In The Making Of The Modern World Income Distribution.”The Quarterly Journal of Economics 117(4):1231–1294.

Agricultural Returns. 1896. Agricultural Returns for Great Britain, Board of Agriculture,London, 1896.

Aidt, Toke S., Martin Daunton and Jayasri Dutta. 2010. “The Retrenchment Hypothe-sis and the Extension of the Franchise in England and Wales.” The Economic Journal120(547):990–1020.

Aidt, Toke S. and Peter S. Jensen. 2009. “The taxman tools up: An event history study of theintroduction of the personal income tax.” Journal of Public Economics 93(1-2):160–175.

Ansell, Ben and David J. Samuels. 2015. Inequality and Democratization: An Elite Compe-tition Approach. New York: Cambridge University Press.

Ardanaz, Martin and Carlos Scartascini. 2013. “Inequality and Personal Income Taxation:The Origins and Effects of Legislative Malapportionment.” Comparative Political Studies47(9).

Aydelotte, William O. 1977. Constituency Influence on the British House of Commons, 1841-1847. In The History of Parliamentary Behavior, ed. William O. Aydelotte. PrincetonUniversity Press pp. 225–246.

Aydelotte, William O. 1984. British House of Commons Roll Call Data, 1841-1847, SN:7384. Inter-university Consortium for Political and Social Research.

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

Beck, Nathaniel and Jonathan N. Katz. 2001. “Throwing out the Baby with the Bath Water:A Comment on Green, Kim, and Yoon.” International Organization 55(2):487–495.

Beck, Nathaniel, Jonathan N. Katz and Richard Tucker. 1998. “Beyond Ordinary Logit:Taking Time Seriously in Binary Time-Series-Cross-Section Models.” American Journalof Political Science 42(4):1260–88.

Becker, Sascha, Francesco Cinnirella, Erik Hornung and Ludger Woßmann. 2014. “iPEHD- The ifo Prussian Economic History Database.” Historical Methods: A Journal of Quan-titative and Interdisciplinary History 47(2):57–66.

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

33

Boix, Carles. 2003. Democracy and Redistribution. New York: Cambridge University Press.

Caramani, Daniele. 2000. Elections in Western Europe since 1815. London: MacmillanPress.

Carter, David B. and Curtis S. Signorino. 2010. “Back to the Future: Modeling TimeDependence in Binary Data.” Political Analysis 18(3):271–292.

Daunton, Martin. 2001. Trusting Leviathan: the Politics of Taxation in Britain, 1799-1914.New York: Cambridge University Press.

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

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

Flora, Peter, Franz Kraus and Winfried Pfenning. 1983. State, Economy, and Society inWestern Europe 1815-1975. London: Macmillan Press.

Gutierrez, Roberto G. 2002. “Parametric frailty and shared frailty survival models.” StataJournal 2(1):22–44.

Hallerberg, Mark. 1996. “Tax Competition in Wilhelmine Germany and Its Implications forthe European Union.” World Politics 48(3):324–357.

Hallerberg, Mark. 2002. “The Political Economy of Taxation in Prussia, 1871/1914.”Jahrbuch fur Wirtschaftsgeschichte 2:11–34.

HC Debate April 13. 1842. Hansard, HC Deb 13 April 1842 vol 62 cc377-456.

HC Debate April 29. 1842. Hansard, HC Deb 29 April 1842 vol 62 cc1256-317.

HC Debate April 4. 1842. Hansard, HC Deb 04 April 1842 vol 61 cc1251-90.

HC Debate April 8. 1842. Hansard, HC Deb 08 April 1842 vol 62 cc84-174.

HC Debate March 18. 1842. Hansard, HC Deb 18 March 1842 vol 61 cc839-929.

HC Debate March 21. 1842. Hansard, HC Deb 21 March 1842 vol 61 cc944-1025.

HC Debate March 23. 1842. Hansard, HC Deb 23 March 1842 vol 61 cc1118-219.

HCPP. 1830-1. “Accounts and Papers.” House of Commons Parliamentary Papers X.9(202).

Kennan, Kossuth Kent. 1910. Income Taxation: Methods and Results in Various Countries.Milwaukee: Burdick & Allen.

Lampe, Markus and Paul Sharp. 2013. “Tariffs and income: A time series analysis for 24countries.” Cliometrica 7(3):207–235.

34

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

Lieberman, Evan S. 2002. “Taxation data as indicators of state-society relations: possibilitiesand pitfalls in cross-national research.” Studies in Comparative International Development36(4):89–115.

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

Lindner, Ralf and R.-O. Schultze. 2005. United States of America. In Elections in theAmericas: A Data Handbook, ed. Dieter Nohlen. Oxford University Press.

Mares, Isabela. 2015. From open secrets to secret ballots: The adoption of electoral reformsprotecting voters against intimidation. New York: Cambridge University Press.

Marshall, Monty G. and Keith Jaggers. 2000. Polity IV Project: Political Regime Charac-teristics and Transitions, 1800-2010. Center for International Development and ConflictManagement. University of Maryland.

Mason, R. H. P. 1969. Japan’s First General Election 1890. New York: Cambridge UniversityPress.

Matthew, H.C.G. 1988. Gladstone, 1809-1874. New York: Oxford University Press.

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

Nohlen, Dieter. 2005. Elections in the Americas A Data Handbook Volume 1. North America,Central America, and the Caribbean. Oxford: Oxford University Press.

Nohlen, Dieter and Philip Stover. 2010. Elections in Europe. Baden-Baden: Nomos.

Parliamentary Papers. 1857. First Report of The Commissioners of Inland Revenue on ThenInland Revenue, Session 1, vol. IV.65 no. 2199.

Parliamentary Papers. 1861. Report from the Select Committee on Assessment and Collectionof Income and Property Tax, vol. VII.1, 339 no. 503.

Popitz, Johannes. 1926. Einkommensteuer. In Handworterbuch der Staatswissenschaften, ed.Ludwig Elster, Adolf Weber and Friedrich Wiese. Jena: Gustav Fischer pp. 400–84.

Ranis, G. 1959. “The Financing of Japanese Economic Development.” The Economic HistoryReview 11(3):440–454.

Salmon, Philip. 2002. Electoral Reform at Work: Local Politics and National Parties, 1832-1841. Woodbridge: Boydell Press.

Sarkees, Meredith Reid and Frank Wayman. 2010. Resort To War. CQ Press.

Scheve, Kenneth and David Stasavage. 2010. “The Conscription of Wealth: Mass Warfareand the Demand for Progressive Taxation.” International Organization 64(4):529–561.

35

Schumpeter, Joseph A. 1918. The Crisis of the Tax States. In Joseph A. Schumpeter: TheEconomics and Sociology of Capitalism, ed. Swedberg Richard. 1991 ed. Princeton, NJ:Princeton University Press.

Seligman, Edwin R.A. 1911. The Income Tax: a Study of the History, Theory and Practiceof Income Taxation at Home and Abroad. 2004 ed. New York: Macmillan Press.

Singer, J. David, Stuart Bremer and John Stuckey. 1972. Capability Distribution, Uncer-tainty, and Major Power War, 1820-1965 v4.0. In Peace, War, and Numbers, ed. BruceRussett. Beverly Hills: Sage pp. 19–48.

Sokoloff, Kenneth L. and Eric M. Zolt. 2007. “Inequality and the Evolution of Institutions ofTaxation: Evidence from the Economic History of the Americas.” NBER Working Paperspp. 83–138.

Southall, H.R. et al. 2004. Great Britain Historical Database: Census Data, Parish-levelPopulation Statistics, 1801-1951. SN 4560. Colchester, Essex: UK Data Archive.

Statistics New Zealand. 1893. The New Zealand Official Yearbook. Statistics New Zealand.

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

Vanhanen, Tatu. 2003. Democratization: A Comparative Analysis of 170 Countries. NewYork: Routledge.

Ziblatt, Daniel. 2008. “Does Landholding Inequality Block Democratization.” World Politics60(4):610–41.

36

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Uni

ted

King

dom

184

2

020406080100Franchise

1810

1830

1850

1870

1890

1910

1930

year

Aust

ria 1

849

020406080100Franchise

1860

1880

1900

1920

1940

year

Italy

186

4

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Japa

n 18

87

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

New

Zea

land

189

1

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Nor

way

189

2

020406080100Franchise

1830

1850

1870

1890

1910

1930

year

Net

herla

nds

1893

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Swed

en 1

902

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Den

mar

k 19

03

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Fran

ce 1

914

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

USA

191

3

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Aust

ralia

191

5

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Can

ada

1917

020406080100Franchise

1870

1890

1910

1930

year

Ger

man

y 19

20

020406080100Franchise

1810

1835

1860

1885

1910

1935

year

Finl

and

1920

020406080100Franchise

1830

1850

1870

1890

1910

1930

year

Belg

ium

192

2

020406080100Franchise

1840

1860

1880

1900

1920

1940

year

Switz

erla

nd 1

939

Figure 1: Income Tax Adoption and Franchise in Western Europe. The vertical line indicatesthe year of the permanent adoption of the tax.

37

1520

2530

3540

-15 -12 -9 -6 -3 0 3 6 9 12 15Adoption

Figure 2: Direct Taxation as Share of Central Government Revenue for Early Adopters anda 15-year window frame before and after the permanent adoption of income tax. Localpolynomial smoothed line with 90 CI. Source: Flora, Kraus and Pfenning (1983).

38

22.

53

3.5

4Ta

x R

ate

1842 1844 1846 1848 1850 1852 1854year

schedule D schedules A and B

Figure 3: United Kingdom 1843-1853: The tax Rate of the Traditional sector (dashed curve)and Modern sector (solid curve). The rates represent the ratio of actual tax collection tototal assessed value of assets cataloged on schedules A and B (Traditional sector) and D(Modern sector). Source: Parliamentary Papers (1861) and Parliamentary Papers (1857).

39

Table 1: Vote-Tax Link for the Election to the National Lower Chamber

Years Effective space Years Effective

Australia - Italyb 1848-1912Austria-Hungary 1849-1907 Japanc 1890-1925Belgium 1831-1893 Netherlands 1849-1917Canada - New Zealand -Denmark - Norwayd 1885-1897Finland - Swedene 1866-1907France 1817-1848, 1850-1852 Switzerland -Germany - United Statesf 1788-1859Prussiaa 1848-1913 United Kingdomh -a Replaces Germany in the Online Appendix. b Male Adults under 21 had to paya minimum tax from 1912 to 1919. c Specific source: Mason (1969). d Payingincome taxes was a sufficient condition, not necessary e Paying income taxes wasa sufficient condition, not necessary. From 1909 to 1945, those who had tax debtwere excluded from the right to vote. f Article 1, section 2 of the US AmericanConstitution entitles the individual states to design their own election laws for theselection of their representatives in Congress. From 1788 to 1850, most states passedtax-based requirements for enfranchisement Lindner and Schultze (2005). h Prompttax payment of local taxes, not direct, for national suffrage (Salmon, 2002). Generalsources: Flora, Kraus and Pfenning (1983), Caramani (2000), Nohlen (2005), andNohlen and Stover (2010).

40

Table 2: Summary of Hypotheses under the Assumption of Future Changes in EconomicPositions.

Economic gains fromincome taxes

Political gains from in-come taxes

Owners of fixed assets(old elites)

Yes Yes

Owners of mobile assets(new elites)

No Yes

41

Table 3: The Adoption of the Income Tax as a function of Representative Insti-tutions, War, Landed Inequality, and Socio-Economic Controls.

(1) (2) (3) (4) (5)

Franchise -0.045** -0.043** -0.052 -0.045** -0.058(0.021) (0.021) (0.036) (0.022) (0.036)

Polity -0.414*** -0.489*** -0.516** -0.489*** -0.546***(0.140) (0.163) (0.204) (0.161) (0.212)

Landholding Inequality 0.047* 0.083*(0.025) (0.043)

Rural Inequality 0.061** 0.120**(0.027) (0.056)

War 1.611 1.683 1.000 1.760* 1.080(1.096) (1.065) (1.072) (1.060) (1.048)

WWI participant 1.386 1.304 1.421 1.286 1.423(1.053) (0.905) (1.162) (0.950) (1.242)

ln(Population) 0.728 0.776 0.896* 0.770 0.941*(0.498) (0.482) (0.490) (0.499) (0.534)

ln(GDP/cap) 3.657* 5.116** 5.557 5.162** 6.101*(2.103) (2.565) (3.407) (2.520) (3.456)

Urbanization 0.163*** 0.156** 0.240*** 0.198*** 0.347***(0.060) (0.061) (0.062) (0.072) (0.106)

Temporary Income Tax -5.310*** -6.729*** -8.464*** -6.860*** -9.147***(1.584) (2.456) (3.028) (2.410) (3.248)

Local Income Tax 0.775 1.668** 2.313 1.719** 2.227(0.580) (0.793) (1.537) (0.821) (1.550)

ln(Tariff) 0.626 1.034(0.905) (1.079)

Constant -1.646 -2.881 -10.877 -4.079 -15.905(6.003) (4.803) (7.879) (4.747) (9.919)

Observations 849 849 675 849 675Countries 17 17 15 17 15Duration Dependence YES YES YES YES YESEurope FE YES YES YES YES YES

To model duration dependence, all models include a third-order flexible polynomialof the number of years elapsed without the income tax. Country-clustered robuststandard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

42

Table 4: Tax Barriers and the Income Tax.

(1) (2) (3)

Polity -0.494*** -0.482*** -0.494***(0.176) (0.171) (0.175)

Franchise -0.026 -0.027 -0.027(0.023) (0.025) (0.025)

Landholding Inequality 0.059* 0.060*(0.034) (0.033)

Rural Inequality 0.073**(0.034)

Vote-Tax Link 1.748** 1.707** 1.871*(0.683) (0.701) (1.106)

Secret Ballot -0.278(2.606)

War 1.903* 1.957* 1.920(1.090) (1.094) (1.170)

WWI Participant 1.136 1.079 1.148(0.873) (0.937) (0.955)

ln(GDP/cap) 4.578* 4.529* 4.813(2.494) (2.497) (4.011)

ln(Population) 0.812* 0.795 0.799*(0.477) (0.488) (0.484)

Urbanization 0.175*** 0.221*** 0.173**(0.064) (0.079) (0.074)

Temporary Income Tax -6.978*** -6.996*** -7.082**(2.564) (2.442) (2.814)

Local Income Tax 1.843* 1.843** 1.835*(0.967) (0.929) (0.991)

Constant -6.464 -7.668 -5.906(5.743) (5.723) (9.954)

Observations 849 849 849Countries 17 17 17Temporal Duration Yes Yes YesEurope FE Yes Yes Yes

To model duration dependence, all models include a third-orderflexible polynomial of the number of years elapsed without theincome tax. Country-clustered robust standard errors in paren-theses. *** p<0.01, ** p<0.05, * p<0.1

43

Tab

le5:

Roll

call

vote

of

the

Bri

tish

Inco

me

Tax

Bil

lof

May

31,

1842:

0,

again

st,

1in

favor.

(1)

(2)

(3)

(4)

(5)

(6)

(7)

Urb

and

istr

ict

MP

-2.3

16***

-1.8

16***

-2.0

12***

-2.1

38***

-2.2

58***

-2.6

45**

-0.1

89

(0.3

49)

(0.4

47)

(0.5

26)

(0.3

48)

(0.3

95)

(1.2

44)

(0.5

80)

Lan

dh

old

ing

Gin

i5.7

45***

6.2

45***

7.0

05***

4.9

36**

5.4

40**

-1.9

47

-4.9

94

(2.2

17)

(2.2

17)

(2.3

28)

(2.2

50)

(2.5

08)

(6.9

71)

(4.5

64)

Old

sect

orem

plo

ym

ent

7.7

79***

(2.9

19)

New

sect

orem

plo

ym

ent

-2.7

06*

(1.4

71)

Un

emp

loym

ent

Rat

e-1

.863

(1.6

29)

Poor

Hom

es0.0

10

(0.0

10)

Pop

ula

tion

Den

sity

0.0

04

(0.0

05)

Bu

sin

ess

MP

’sb

ackgr

oun

d-0

.886**

-1.1

12**

1.4

17

0.7

58

(0.4

22)

(0.5

12)

(1.1

85)

(0.7

54)

Ban

kin

gM

P’s

bac

kgr

oun

d0.3

12

0.6

02

-1.5

06*

-1.2

92*

(0.4

38)

(0.4

97)

(0.9

00)

(0.6

97)

Ari

stocr

at0.0

79

0.1

09

0.2

12

0.2

99**

(0.0

59)

(0.0

70)

(0.2

33)

(0.1

35)

Per

son

alW

ealt

h0.1

26*

(0.0

74)

Wh

ig-9

.429***

(1.4

77)

Pro

-lan

dvo

tin

gre

cord

0.6

86***

(0.1

00)

Con

stan

t-2

.499*

-3.0

50**

-2.8

32

-2.2

01

-3.0

72*

6.2

00

-7.9

97***

(1.3

17)

(1.3

28)

(1.7

24)

(1.3

75)

(1.6

70)

(5.5

15)

(3.0

22)

Ob

serv

atio

ns

384

310

268

384

321

384

384

Reg

ion

FE

YE

SY

ES

YE

SY

ES

YE

SY

ES

YE

SC

ounty

-clu

ster

ed,

rob

ust

stan

dard

erro

rsin

pare

nth

eses

***

p<

0.0

1,

**

p<

0.0

5,

*p<

0.1

44

Online Appendix

The Online Appendix includes

• Data Details

– Table A-1: Summary Statistics of the Crossnational Data.

– Table A-2: Value of the Vote-Tax Link on the Year of Adoption of the Income

Tax.

– Table A-3: Summary Statistics of the British Data.

– Figure A-1: Land Inequality Over Time vs. the Adoption of the Income Tax.

– Figure A-2: Rural Inequality Over Time vs. the Adoption of the Income Tax.

• Robustness Tests

– Table A-4: Dropping Cases with minimum variation in Rural Inequality.

– Table A-5: Germany replaced for Prussia.

– Table A-6: Additional measures of war.

– Table A-7: Sequencing of land ownership and income tax adoption.

– Accounting for unit-level unobserved heterogeneity.

∗ Table A-8: Weibull Models with Shared Frailty.

∗ Table A-9: Splines and Conditional log-log models.

– Table A-10: Secret Ballot vs. the Vote-Tax Link with Aidt and Jensen (2009)

specification.

i

Tab

leA

-1:

Sum

mar

ySta

tist

ics

ofth

eC

ross

nat

ional

Dat

a.

Variable

Mean

Std

.Dev.

Min.

Max.

NSource

Inco

me

Tax

Ad

opti

on

Date

0.02

0.14

01

849

Aid

tan

dJen

sen

(200

9)F

ran

chis

e49

.818

38.3

570

107.

35a

849

Aid

tan

dJen

sen

(200

9)P

oli

tyIV

1.67

86.

866

-10

1084

9M

arsh

all

and

Jag

gers

(200

0)V

ote-

Tax

Lin

k0.

244

0.43

01

849

Ref

erto

Tab

le1

Sec

ret

Ball

ot

0.46

90.

499

01

849

Aid

tan

dJen

sen

(200

9)L

an

dh

old

ing

Ineq

uali

ty60

.398

16.5

0929

100

849

Van

han

en(2

003)

Ru

ral

Ineq

uali

ty51

.089

14.2

6119

.895

849

Van

han

en(2

003)

War

0.16

10.

368

01

849

Sar

kees

and

Way

man

(201

0)W

WI

Part

icip

ant

(tem

por

al)

0.01

30.

113

01

849

Ow

nW

WI

Part

icip

ant

(per

man

ent)

0.01

50.

123

01

849

Ow

nM

ilit

ary

Siz

e(%

pop

)1.

343

1.83

60

12.2

6370

3S

inge

r,B

rem

eran

dS

tuck

ey(1

972)

GD

Pp

erC

apit

a(0

00)b

2421

1071

737

6390

849

Aid

tan

dJen

sen

(200

9)P

op

ula

tion

(000)b

1565

820

120

119

9760

684

9A

idt

and

Jen

sen

(200

9)U

rban

Pop

ula

tion

(%)

15.1

599.

106

340

849

Van

han

en(2

003)

Tem

pora

ryIn

com

eT

ax0.

687

0.46

40

184

9A

idt

and

Jen

sen

(200

9)L

oca

lIn

com

eT

ax

0.41

70.

493

01

849

Aid

tan

dJen

sen

(200

9)E

uro

pe

0.75

90.

428

01

849

Ow

nA

VE

Tari

ff0.

111

0.09

0.00

60.

477

675

Lam

pe

and

Sh

arp

(201

3)aB

etw

een

188

7an

d18

91,

inN

ewZ

eala

nd

,th

ees

tim

ated

elec

tora

teas

ap

erce

nta

geof

the

enfr

anch

ised

age

and

sex

grou

pis

gre

ater

than

100.

bG

DP

/cap

and

Pop

ula

tion

are

scal

ed1/

10,0

00in

the

emp

iric

alan

alysi

s.

ii

Table A-2: Pre-existence of the Vote-Tax Link on the Year of Adoption of the Income Tax.

Adoption year Link space Adoption year LinkUK 1842 noa Denmark 1903 noAustria 1849 yes France 1911 noItaly 1864 yes Australia 1915 noJapan 1887 nob Canada 1917 noNew Zealand 1891 no Finland 1920 noNorway 1892 yes Belgium 1922 noPrussia 1891 yes Germany 1920 noNetherlands 1893 yes USA 1913 noSweden 1902 yes Switzerland 1939 noa From 1832 to 1865, enfranchisement for the national elections was conditionalon paying local taxes, but not income taxes. bJapan adopted the income tax evenbefore Parliamentarism. The first Diet convened in 1890, and it had a vote-tax linkin place.

iii

Tab

leA

-3:

Sum

mar

yst

atis

tics

ofth

eB

riti

shD

ata

Vari

ab

leM

ean

Std

.D

ev.

Min

.M

ax.

NU

nit

Sou

rce

Inco

me

Tax

Rol

lC

all

0.65

70.

475

01

388

Dis

tric

t∗A

ydel

otte

(198

4)C

ity

0.61

60.

487

01

388

Dis

tric

tA

ydel

otte

(198

4)B

usi

nes

sM

P’s

bac

kgr

ound

0.17

80.

383

01

388

Dis

tric

tA

ydel

otte

(198

4)B

ankin

gM

P’s

bac

kgr

ound

0.08

20.

275

01

388

Dis

tric

tA

ydel

otte

(198

4)A

rist

ocr

at1.

907

2.37

80

638

8D

istr

ict

Aydel

otte

(198

4)P

erso

nal

Wea

lth

4.18

81.

756

19

325

Dis

tric

tA

ydel

otte

(198

4)W

hig

0.34

80.

477

01

388

Dis

tric

tA

ydel

otte

(198

4)P

ro-l

and

voti

ng

reco

rd18

.041

5.22

311

2238

8D

istr

ict

Aydel

otte

(198

4)O

ldSec

tor

0.04

70.

074

00.

504

313

Dis

tric

tSou

thal

l,H

.R.

etal

.(2

004)

New

Sec

tor

0.09

80.

107

01.

026

313

Dis

tric

tSou

thal

l,H

.R.

etal

.(2

004)

Unem

plo

ym

ent

Rat

e0.

077

0.07

70

0.80

827

5D

istr

ict

Sou

thal

l,H

.R.

etal

.(2

004)

Poor

Hou

ses

81.8

0216

.494

14.2

3799

.821

299

Dis

tric

tH

CP

P(1

830-

1)L

andhol

din

gIn

equal

ity

0.63

60.

072

0.43

90.

813

385

Cou

nty

Agr

icult

ura

lR

eturn

s(1

896)

∗ Dis

tric

tre

fers

toE

lect

oral

Dis

tric

t.T

he

latt

eris

nes

ted

wit

hin

county

.

iv

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Uni

ted

King

dom

184

2

0255075100Land Inequality

1810

1830

1850

1870

1890

1910

1930

year

Aust

ria 1

849

0255075100Land Inequality

1860

1880

1900

1920

1940

year

Italy

186

4

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Japa

n 18

87

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

New

Zea

land

189

1

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Nor

way

189

2

0255075100Land Inequality

1830

1850

1870

1890

1910

1930

year

Net

herla

nds

1893

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Swed

en 1

902

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Den

mar

k 19

03

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Fran

ce 1

911

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

USA

191

3

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Aust

ralia

191

5

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Can

ada

1917

0255075100Land Inequality

1870

1890

1910

1930

year

Ger

man

y 19

20

0255075100Land Inequality

1810

1835

1860

1885

1910

1935

year

Finl

and

1920

6065707580Land Inequality

1830

1850

1870

1890

1910

1930

year

Belg

ium

192

2

0255075100Land Inequality

1840

1860

1880

1900

1920

1940

year

Switz

erla

nd 1

939

Figure A-1: Landholding Inequality (100-Family Farm) vs. Adoption of the Income Tax.Starting year varies by data availability and country’s independence.

v

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Uni

ted

King

dom

184

2

500

2500

4500

6500

8500

Rural Inequality

1810

1830

1850

1870

1890

1910

1930

year

Aust

ria 1

849

500

2500

4500

6500

8500

Rural Inequality

1860

1880

1900

1920

1940

year

Italy

186

4

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Japa

n 18

87

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

New

Zea

land

189

1

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Nor

way

189

2

500

2500

4500

6500

8500

Rural Inequality

1830

1850

1870

1890

1910

1930

year

Net

herla

nds

1893

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Swed

en 1

902

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Den

mar

k 19

03

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Fran

ce 1

911

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

USA

191

3

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Aust

ralia

191

5

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Can

ada

1917

500

2500

4500

6500

8500

Rural Inequality

1870

1890

1910

1930

year

Ger

man

y 19

20

500

2500

4500

6500

8500

Rural Inequality

1810

1835

1860

1885

1910

1935

year

Finl

and

1920

500

2500

4500

6500

8500

Rural Inequality

1830

1850

1870

1890

1910

1930

year

Belg

ium

192

2

500

2500

4500

6500

8500

Rural Inequality

1840

1860

1880

1900

1920

1940

year

Switz

erla

nd 1

939

Figure A-2: Rural Inequality (100-Family Farm)(100-Urban Population) vs. Adoption ofthe Income Tax. Starting year varies by data availability and country’s independence.

vi

Dropping Cases with minimum in Rural Inequality

The first values of family farms and urbanization for Finland, New Zealand, and Aus-

tralia in Vanhanen (2003) date as of 1918, 1908 and 1908, respectively. These values are

extrapolated until the date of independence for Australia and New Zealand, and 1815 for

Finland. This explains why the land inequality series for these countries are flat over the

period, even for Rural Inequality, for which temporal variation is supposed to be high due

to the rapid change in urbanization rates in this period. To ensure that the results are not

driven by these de facto fixed effects, Table A-4 reruns the main specifications upon drop-

ping these three countries from the sample. Results hold: high levels of Land and Rural

Inequality, as well as the Vote-Tax Link, predict the adoption of the income tax also in the

absence of Australia, New Zealand and Finland.

vii

Table A-4: Main models rerun after dropping countries with low variation in Rural Inequal-ity: Australia, New Zealand, and Finland.

(1) (2) (3)

Franchise -0.083*** -0.084*** -0.074**(0.031) (0.030) (0.034)

Polity -0.418** -0.425** -0.313**(0.182) (0.194) (0.144)

Landholding Inequality 0.067* 0.116(0.035) (0.074)

Rural Inequality 0.079**(0.037)

Vote-Tax Link 2.982*(1.690)

War 1.840 1.859 2.092**(1.155) (1.146) (1.031)

WWI Participant 1.925 1.947 2.356(1.597) (1.741) (2.198)

ln(Population) 1.086** 1.027** 0.901*(0.527) (0.490) (0.483)

ln(GDP/cap) 5.854* 5.762* 4.627(3.054) (2.953) (2.856)

Urbanization 0.118 0.192 0.167(0.096) (0.119) (0.116)

Temporary Income Tax -6.309** -6.580** -6.068*(3.020) (3.238) (3.320)

Local Income Tax 4.016*** 3.911*** 6.359**(1.414) (1.244) (3.153)

Constant -5.345 -6.596 -16.444(3.760) (4.304) (10.228)

Observations 795 795 795Duration Dependence YES YES YESCountries 14 14 14Europe FE YES YES YES

To model duration dependence, all models include a third-orderflexible polynomial of the number of years elapsed without theincome tax. Country-clustered robust standard errors in paren-theses. *** p<0.01, ** p<0.05, * p<0.1

viii

Replacing Germany for Prussia

In Table A-5 we replace Germany for Prussia. The former adopted the federal income tax

in 1920, after WWI. Prussia, as many other German principalities, adopted province-level

income taxes earlier, as soon as 1891, when franchise was restricted and a vote-tax link was

in place. Table A-5 replicates the main specifications replacing Germany for Prussia. Given

the size of Prussia, we use the Germany values for Land Inequality, War, WWI Participant,

ln(GDP/cap), and Urbanization. Aidt and Jensen (2009) franchise variable is defined as the

electorate (for parliamentary elections) in percentage of the enfranchised age and sex group,

before women’s suffrage, male population only. In Prussia, only men 20 years old or above

were allowed to vote. We draw the electorate size for Prussia from Wahlen in Deutschland

(http://www.wahlen-in-deutschland.de), and we approximate the share of male 20 years old

or above based on census data from the Preussische Statistik and Becker et al. (2014). Total

population is drawn from Flora, Kraus and Pfenning (1983). We set Temporary Income

Tax to 1 (Prussia adopted one during the Napoleonic wars), and Local Income tax to 0, as

Prussian municipalities had never operated this tax prior to 1891. Results suggest that both

Landholding and Rural Inequality, as well as the Vote-Tax Link, predict the adoption of the

income tax even if we replace Germany for Prussia.

ix

Table A-5: Main Models Upon Replacing Germany for Prussia.

(1) (2) (3)

Franchise -0.042* -0.043* -0.028(0.025) (0.025) (0.028)

Polity -0.466*** -0.452*** -0.439***(0.155) (0.153) (0.161)

Land Inequality 0.046** 0.055**(0.018) (0.024)

Rural Inequality 0.054***(0.021)

Vote-Tax Link 1.313*(0.777)

War 0.915 0.964 1.013(1.133) (1.143) (1.197)

WW I Participant 1.923*** 1.994*** 1.951***(0.724) (0.740) (0.732)

ln(Population) 0.826** 0.798** 0.700**(0.330) (0.325) (0.354)

ln(GDP/Cap) 5.879*** 5.620*** 5.015**(2.123) (2.065) (2.303)

Urbanization 0.096* 0.134** 0.108*(0.052) (0.062) (0.055)

Temporary Income Tax -6.328*** -6.197*** -6.062**(2.282) (2.201) (2.386)

Local Income Tax 1.052 1.032 1.143(0.718) (0.763) (0.723)

Constant 0.951 -0.204 -2.058(3.003) (3.068) (4.436)

Observations 842 842 842Duration Dependence YES YES YESEurope FE YES YES YES

To model duration dependence, all models include a third-orderflexible polynomial of the number of years elapsed without theincome tax. Country-clustered robust standard errors in paren-theses. *** p<0.01, ** p<0.05, * p<0.1

x

Additional measures of War

Table 3 in the main text includes two controls for war: one is being at war at time

t, the other participating in WWI between 1914-1918. We included these two covariates to

investigate how robust the early adoption of the income tax among proto-democratic regimes

is once we control for the main cause of fiscal innovations (Dincecco, 2011; Tilly, 1990; Scheve

and Stasavage, 2010). Table A-6 considers additional measures of war. Columns 1 and 2

include the first and the second lag of war. The adoption the income tax might follow

war if the latter created debt that has to be honored once armed conflict is over (we do

not report models with additional lags as results do not vary). In column 4 we evaluate

whether participation in WWI has any lasting effect. WW1 Participant in the main text is

equal to one for WWI participants during war years and zero otherwise. WW1 Participant

(permanent) equals one for WW participants during war years and all that follow. That is,

it captures any lasting effect of participating in WWI. Finally, in column 5 we control for

the magnitude of commitments to and preparation for war efforts by including a measure

of military mobilization, measured by the share of military personnel to total population

(source: Singer, Bremer and Stuckey (1972)). Irrespective of how we adjust for the effect of

war on the adoption of income taxes, low franchise and low Polity IV are strong predictors

of the adoption of this tax, thus confirming the intuition in Figure 1.

xi

Table A-6: Different Measures of War

(1) (2) (3) (4) (5)

Franchise -0.041** -0.048** -0.054** -0.050** -0.084***(0.018) (0.021) (0.024) (0.023) (0.021)

Polity -0.442*** -0.484*** -0.522*** -0.391*** -0.474*(0.151) (0.164) (0.190) (0.151) (0.272)

War 1.656 1.447 1.720 1.571 1.672(1.081) (1.053) (1.097) (1.151) (1.201)

First lag of War 1.298 0.962(0.917) (0.686)

Second Lag of War 1.210(0.894)

WW I participant (permanent) 2.696***(0.850)

Military Personnel/Population 0.181(0.362)

ln(Population) 0.798 0.782 0.742 0.642 1.178**(0.518) (0.517) (0.526) (0.520) (0.540)

ln(GDP/cap) 3.894 4.546* 5.057* 3.405* 3.131(2.389) (2.615) (2.859) (2.023) (4.299)

Urbanization 0.172*** 0.163** 0.156** 0.161*** 0.239**(0.066) (0.070) (0.076) (0.058) (0.101)

Temporary Income Tax -5.702*** -6.047*** -6.367*** -4.976*** -8.014**(1.747) (1.784) (1.904) (1.642) (3.755)

Local Income Tax 0.906* 1.073** 1.227** 0.598 2.744(0.548) (0.526) (0.544) (0.574) (1.706)

Constant -2.018 -0.412 0.893 -1.280 -6.412(6.790) (7.318) (8.043) (5.109) (11.460)

Observations 849 844 837 849 703Temporal Duration YES YES YES YES YESEurope FE YES YES YES YES YES

To model duration dependence, all models include a third-order flexible polynomial of the numberof years elapsed without the income tax. Country-clustered robust standard errors in parentheses.*** p<0.01, ** p<0.05, * p<0.1

xii

Sequencing of land ownership and income tax adoption.

In two cases of the crossnational analysis, Japan and Belgium, the income tax is im-

mediately followed by a decrease in land ownership, proxied by Vanhanen’s (2003) family

farms. One might argue that this sequence is artificially induced by the decennial nature

of Vanhanen’s database, in which records are only offered on the eight year of the decade

(e.g. 1908, 1918, 1928). Based on the decennial structure, it could be the case that an

income tax, despite having been adopted after a land reform, appears as it preceded land

decentralization, thus reversing the true causal sequence. For instance, suppose that a land

reform in country x is passed in 1860. This would be reflected only in 1868 in Vahnanen’s

decennial data. Suppose now that the income tax in that country is adopted in 1865. If

we do not correct for the decennial coding schedule in Vanhanen, it would read as if the

income tax had been adopted before the land reform, not after. Based on Figures A-1 and

A-2, this issue could affect Belgium and Japan. How sensible are the results to Vanhanen’s

coding-induced sequencing? We examine this separately.

Japan. The income tax was adopted in Japan in 1887 as a part of the Meiji restoration,

which involved a bundle of political and economic reforms, among these, a land reform.

The latter started in 1873 with the adoption of Land Tax Reform, and it was completed

progressively during the 1870s and 1880s (Bird 1977).31 The effect of this reform is picked

in Vanhanen’s decennial data only in 1888. The previous data point, dated as of 1878, does

not account for any effect of the land reform. In particular, from 1873 to 1887 family farm

for Japan equals 0 (i.e. maximum landholding inequality). In light of Vanhanen’s coding de-

cision, one might suspect that the sequence of events in the dataset could bias the results in

our favor. To evaluate the possibility of bias, we can modify the rural inequality variable so

that it early recognizes the land liberalization under the Meiji’s reform. We do this twofold:

31Richard M. Bird. 1997. “Land Taxation and Economic Development: The Model of Meiji Japan”,Journal of Development Studies, Vol. 13(2): 162- 174.

xiii

first, we extrapolate the value of 1888 family farm entry (the first in the series that recognizes

the Meiji’s land reform) all the way back to 1873 (this is consistent with our flat-steep-flat

strategy, as explained in ft.17). Second, we linearly interpolate the value of land inequality

between 1873 and 1888. One way or another, after these changes are made, the land reform

is forced to precede the adoption of the income tax. We replicate the models in Table 3 in the

manuscript, using Rural Inequality as proxy for the bargaining power of the landed elites.

Columns 1-2 and 3-4 in Table A-7 report the results for the backwards extrapolation and

linear interpolation, respectively. Results are virtually identical to those reported in Table 3

even though, in column 3, Rural Inequality does not reach conventional levels of statistical

significance. All other variables behave as in Table 3. This is an important result, because

Japan constitutes a paradigm of the logic that we are advancing in this paper: that is, in

scenarios of limited representation, old-regime landed interests may benefit from the income

tax due to its economic and political gains.

Belgium. The income tax is adopted in 1922 in Belgium. Land Inequality, as measured

by the share of family farms, remains stable at 22% since 1858, except for the 1898-1917

period in which it decreases to 21% only to return to 22% in 1918. Then, this value jumps

to 30% in 1928.32 To confirm that the results are not driven by Vanhanen’s decennial cod-

ing (for instance, land could have been redistributed in 1919, 1920 or 1921 despite having

artificially remaining fixed to 21% until 1928), we replace all the post-1918 values of family

farm for the 1828 value: that is, 30%. This way, we assume that the reduction of land

inequality precedes the adoption of income taxes. Results with the revised coding of Rural

Inequality are reported in columns 5 and 6 in Table A-7. The coefficients for this variable

remain positive and statistically significant.

Overall, Table A-7 suggest that results in Table 3 are not driven by Vanhanen’s decennial

32We have not found evidence of any major land reform in Belgium right after WWI.

xiv

coding.

Table A-7: Sensibility Analysis: Pushing Land Decentralization Backwards in Japan andBelgium.

JAPAN BELGIUMBackwards Linear Backwards

extrapolation Interpolation extrapolation(1) (2) 3 (3) (4) (6) (5) (6)

Polity -0.488*** -0.474** -0.478*** -0.484*** -0.483*** -0.522***(0.152) (0.192) (0.152) (0.181) (0.156) (0.195)

Franchise -0.044** -0.055 -0.045** -0.054* -0.044** -0.057(0.020) (0.036) (0.021) (0.033) (0.022) (0.035)

Rural Inequality 0.066* 0.112* 0.054 0.099* 0.059** 0.113**(0.040) (0.063) (0.034) (0.059) (0.027) (0.051)

ln(Tariff) 0.443 0.517 0.839(0.747) (0.808) (0.960)

War 1.729 1.043 1.722 1.042 1.765* 1.079(1.060) (1.134) (1.055) (1.091) (1.071) (1.075)

WW1 Participant 1.212 1.361 1.241 1.348 1.293 1.476(0.909) (1.174) (0.929) (1.163) (0.958) (1.305)

ln(GDP/cap) 4.537** 4.146 4.646** 4.599* 5.044** 5.696*(2.193) (2.551) (2.303) (2.753) (2.430) (3.231)

ln(Population) 0.816* 0.837 0.793 0.860* 0.754 0.884*(0.472) (0.533) (0.485) (0.512) (0.487) (0.512)

Urbanization 0.206*** 0.312*** 0.196*** 0.305*** 0.200*** 0.342***(0.080) (0.106) (0.072) (0.091) (0.074) (0.102)

Temporary Income Tax -7.021*** -8.435*** -6.741*** -8.287*** -6.815*** -8.896***(2.486) (3.190) (2.364) (2.836) (2.367) (2.988)

Local Income Tax 1.618** 2.580* 1.527** 2.307* 1.650** 2.200(0.724) (1.422) (0.757) (1.378) (0.794) (1.520)

Constant -5.162 -14.611 -4.206 -13.454 -4.181 -15.335(5.732) (9.772) (5.147) (9.867) (4.912) (10.188)

Observations 849 675 849 675 849 675Flex Polynomial Yes Yes Yes Yes Yes YesEurope FE Yes Yes Yes Yes Yes Yes

Country-cluster standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

xv

Accounting for unobserved heterogeneity

The models in the main text assume a common hazard rate for all countries, once we

control for all relevant covariates. However, we cannot guarantee that we have controlled for

every relevant variable. In an OLS framework, we would fit a country fixed effect to account

for unobserved heterogeneity that simultaneously influence the unit-specific baseline hazard

and the dependent variable. However, a full vector of country fixed effect cannot be fitted

to BTSCS data, as they would completely determine the duration of the subject. There is

still one option left to account for unit-specific heterogeneity: we can work with a Weibull

duration model with gamma heterogeneity (Beck and Katz, 2001, pag.491) an shared frailty

(Gutierrez, 2002), which accounts for unobserved unit heterogeneity in a similar fashion than

random effects models.

Specifically, frailty models add randomness to each units’ underlying propensity to adopt

the income tax as a function of explanatory variables combined with a stochastic term. This

kind of models are appropriate when we suspect that some units are more (less) likely to

adopt the income tax regardless of the values of the observed independent variables (Beck

and Katz, 2001). Shared frailty models are used when the unobserved frailty is shared

among groups of observations (i.e. the same country over time). Accordingly, the shared

frailty model is said to be equivalent to a random effect model for survival data (Gutierrez,

2002).

Shared frailty models are computationally demanding. Historical cross-national time-

series data, like the one we have here, is not best suited to run this complex models, as

variables tend to be slow moving, covariate correlations high, and left censoring to differ

between units. Still, in Table A-8 we report two set of stable models: one without the

covariates that we use to test our hypotheses, and one including all of them at once. Again,

they are estimated with a Weibull model with gamma heterogeneity and unit-specific (or

shared) frailty.

Results hold. That is, higher land or rural inequality, as well as the vote-tax link, predict

xvi

the adoption of the income tax when we adjust for unit-specific unobserved heterogeneity.

More importantly, the variance of the frailty, θ, is not statistically significant or is only at

10%. This implies that the unobserved unit effect that these models account for is small.

Accordingly, the common hazard rate that we assume in the main text seems is reasonable,

while making the estimation of the models less cumbersome.

For the sake of comprehensiveness, in Table A-9 we also fit splines models to account for

duration dependence (Beck, Katz and Tucker, 1998) and complement them with complimentary-

log log models (Carter and Signorino, 2010), which add full flexibility to the hazard rate by

not assuming ex ante any particular shape (i.e. increasing, decreasing, or non-monotonic).

Again, results hold.

xvii

Table A-8: Accounting for unobserved unit heterogeneity: Weibull models with gammaheterogeneity and shared frailty

(1) (2)

Franchise -0.040** -0.002(0.017) (0.022)

Polity -0.112 -0.185(0.124) (0.147)

Rural Inequality 0.282***(0.097)

Vote-Tax Link 9.386***(2.658)

War 2.466** 4.023**(1.100) (1.607)

WW1 Participant 1.395 0.437(1.231) (1.322)

ln(Population) -0.391 -0.266(0.355) (0.606)

ln(GDP/cap) -1.654 1.434(1.317) (2.416)

Urbanization 0.236*** 0.562***(0.085) (0.129)

Temporary Income Tax -2.291 -3.670**(1.765) (1.850)

Local Income Tax 1.318 6.527***(0.937) (2.300)

Constant -25.720*** -52.552***(6.326) (9.981)

θ 0.536 2.352*(0.799) (1.333)

Observations 844 844Number of groups 17 17Duration Dependence YES YESEurope FE YES YES

Standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

xviii

Tab

leA

-9:

Rob

ust

nes

ste

sts:

Splines

and

Clo

glog

Model

sSplines

Cloglog

(1)

(2)

(3)

(5)

(4)

(5)

(6)

Fra

nch

ise

-0.0

37*

-0.0

40*

-0.0

22

-0.0

44**

-0.0

46**

-0.0

26

(0.0

21)

(0.0

22)

(0.0

21)

(0.0

21)

(0.0

22)

(0.0

23)

Pol

ity

-0.2

01**

-0.2

04**

-0.1

99**

-0.4

89***

-0.4

91***

-0.5

00***

(0.0

87)

(0.0

86)

(0.0

91)

(0.1

45)

(0.1

45)

(0.1

61)

Lan

dh

old

ing

Ineq

ual

ity

0.0

49**

0.0

68***

0.0

47*

0.0

59*

(0.0

21)

(0.0

25)

(0.0

25)

(0.0

35)

Ru

ral

Ineq

ual

ity

0.0

66***

0.0

61**

(0.0

23)

(0.0

26)

Vot

e-T

axL

ink

1.8

26***

1.6

21**

(0.7

02)

(0.7

55)

War

2.3

93*

2.4

97*

2.8

48*

1.5

70

1.6

64

1.7

32

(1.2

91)

(1.2

85)

(1.5

48)

(1.0

67)

(1.0

55)

(1.0

74)

WW

1P

arti

cip

ant

2.6

69*

2.6

87*

2.5

51*

1.2

48

1.2

34

1.0

72

(1.5

63)

(1.6

32)

(1.3

58)

(0.7

84)

(0.8

29)

(0.7

66)

ln(P

opu

lati

on)

0.0

32

0.0

08

-0.0

13

0.7

98*

0.7

97*

0.8

39*

(0.2

22)

(0.2

28)

(0.2

48)

(0.4

35)

(0.4

55)

(0.4

33)

ln(G

DP

per

cap

)1.6

19

1.7

29

1.4

08

5.2

30**

5.3

01**

4.7

21**

(1.2

48)

(1.2

45)

(1.3

01)

(2.3

54)

(2.3

32)

(2.3

02)

Urb

aniz

atio

n0.1

03**

0.1

49***

0.1

04**

0.1

51***

0.1

93***

0.1

69***

(0.0

46)

(0.0

53)

(0.0

46)

(0.0

58)

(0.0

69)

(0.0

62)

Tem

por

ary

Tax

-3.1

37***

-3.2

59***

-3.1

22***

-6.7

14***

-6.8

57***

-6.9

49***

(1.0

43)

(1.0

24)

(1.0

15)

(2.2

70)

(2.2

26)

(2.3

99)

Loca

lT

ax1.7

78***

1.8

95***

2.2

31***

1.6

62**

1.7

23**

1.8

18*

(0.6

24)

(0.7

02)

(0.6

42)

(0.7

95)

(0.8

24)

(0.9

60)

Con

stan

t-6

0.7

62**

-62.1

34**

-65.3

70**

-2.5

08

-3.6

81

-5.9

35

(27.4

53)

(27.4

95)

(26.5

67)

(4.6

38)

(4.5

38)

(5.8

07)

Ob

serv

atio

ns

849

849

849

849

849

849

Sp

lin

esw

ith

lin

ear

tren

dY

ES

YE

SY

ES

NO

NO

NO

Fle

xib

leP

olyn

omia

lof

Tim

eE

lap

sed

NO

NO

NO

YE

SY

ES

YE

SE

uro

pe

FE

YE

SY

ES

YE

SY

ES

YE

SY

ES

Cou

ntr

y-c

lust

ered

rob

ust

stan

dard

erro

rsin

pare

nth

eses

.***

p<

0.0

1,

**

p<

0.0

5,

*p<

0.1

xix

Secret Ballot vs. the Vote-Tax Link

Aidt and Jensen (2009) find that the Secret Ballot is a strong predictor of the adoption

of the income tax. Table 4 indicates that the effect disappears when we control for the

vote-tax link. Here we run the same exercise but keeping Aidt and Jensen (2009) original

specification, which includes a squared term for Franchise, a control for Tax Technology (an

additive index of census, local tax, primary school enrollment, and a dummy variable for high

urbanization), and linguistic proximity between countries. Likewise, duration dependence

here is fitted with splines (Beck, Katz and Tucker, 1998). Results in column 1 confirm

Aidt and Jensen (2009) finding: the presence of the Secret Ballot increases the likelihood of

adoption of the income tax. Column 2 suggests that this effect vanishes once we control for

the Vote-Tax Link, which remains statistically significant. In other words, results in Table

4 are independent of specification.

xx

Table A-10: Secret Ballot vs. the Vote Tax Link assessed with Aidt and Jensen (2009)specification

(1) (2)

Secret Ballot 2.286* 0.925(1.338) (1.146)

Vote-Tax Link 2.228*(1.305)

Franchise -0.269*** -0.282***(0.098) (0.100)

Franchise Squared 0.002** 0.002***(0.001) (0.001)

War 2.404** 2.619*(1.168) (1.366)

ln(GDP/cap) -0.712 -0.473(1.253) (1.478)

ln(Population) 0.181 0.174(0.271) (0.270)

Tax Technology 1.296** 1.191**(0.536) (0.556)

Temporary Income Tax -2.514*** -2.765***(0.834) (1.044)

Linguistic Distance 2.133* 2.379*(1.150) (1.293)

Constant -50.098*** -50.408**(19.197) (19.848)

Observations 948 948Splines YES YESEurope FE YES YES

Country-clustered robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

xxi