How unequal were the Latins? The “strange” case of
Portugal, 1550-1770
Jaime Reis*
Álvaro Santos Pereira**
Conceição Andrade Martins*
*Instituto de Ciências Sociais, Universidade de Lisboa
** Simon Fraser University
For help and advice, the authors wish to express their gratitude to Maria Antónia Pires de
Almeida, Inês Amorim, Raquel Berredo, Leonor Freire Costa, Rui Esperança, Carlos Faísca,
João Ferrão, João Fialho, António Castro Henriques, Bruno Lopes, Filomena Melo, Esteban
Niccolini, Susana Pereira, Amélia Polónia, Isabel dos Guimarães Sá, Carlos Santiago-Caballero
and Ana Margarida Silva.
Address for correspondence: jaime.reis@ics.ul.pt
1
1. Introduction
Latin countries, whether in Europe or the Americas, are perceived nowadays by much
of the literature as having always been more unequal economically than others with
similar levels of income. In part, this view is encouraged by contemporary economic
inequalities, which show that Latin countries exhibit comparatively higher degrees of
income or wealth disparity (Lopez and Perry, 2008). At the same time, it has been
supported by the work of Engerman and others on the different paths of long term
development in the New World (Engerman and Sokoloff, 1997, Engerman, Haber and
Sokoloff, 2000). Their claim is that differing patterns of inequality go back a very long
way and were evident already in the earliest colonial period. Over time they became
more firmly entrenched and ultimately a powerful determinant of macro divergence
between, respectively, Latin America, and Canada and the USA. This colonial legacy
has thus become a central part of the conventional wisdom which explains the economic
differences displayed by post colonial societies (Frankema, 2009).
What is remarkable about this account is how little it has been subjected to empirical
verification. As Williamson (2009) has pointed out, there is very little hard fact to
substantiate it and the scarce available data actually point in the opposite direction to
that of the consensus view1. A second problem with the thesis of the long run
persistence of inequality in the Americas is that it is often specified poorly. A clear
1
In the case of the late 18th century USA, only two observations, to my knowledge, are known, neither of
which is entirely satisfactory. One, from Solow (1971), employs data on housing, which is problematic as
a means of assessing income or wealth inequality levels. The other is from Lindert (2000), citing Alice
Hanson, and is based on a fairly small number of probate inventories. For the entire territory of Latin
America, Milanovic et al. (2007) come up with a single measurement of aggregate inequality for the
fullness of the colonial era. Recently, Dobado and Garcia (2011) have come up with evidence which
suggests that at least Bourbon America was not “an exceptional unequal society” in international terms.
2
definition of the appropriate yardstick for comparison is lacking, i.e. wealth, income,
land ownership or rents? The same applies to the geographic terms of the comparison. Is
this supposed to pit the 13 Colonies against Latin America, or the New World against
the Latins in Europe? Within Europe should we seek to contrast England and the
Netherlands with Iberia? Finally, little attention has been given to the dating of these
assessments despite the likelihood that economic inequality will have displayed
differences across time, as well as across space.
Clearly there is scope for improvement in all these aspects of the international history of
economic inequality and the present paper tries to contribute to this in several ways. In
the first place, it adds to the presently rather small information pool by mobilizing a rich
and little studied seam of data from Portugal, a Latin country and a major colonial
power between the sixteenth and the eighteenth centuries. Secondly, it provides a new
set of reliable and consistent measures of inequality which can be used for meaningful
comparisons with other nations, as well as to show that this Latin country at least was
probably less unequal than England or the Netherlands. Thirdly, it sheds new light on
the long run determinants of inequality, in particular economic growth, and on the
applicability to pre-industrial societies of the Kuznets curve model. Finally, it shows
that the most plausible scenario for the less-dynamic economies of the Early Modern era
such as Portugal was a long term decline in inequality and a concurrent lessening of its
social polarization.
The paper is structured as follows. The next section sums up the current state of
knowledge regarding economic inequality in Early Modern countries. It also highlights
the problems of making long term comparative appraisals when heterogeneous data or
methods are employed. The third section presents the data used in the paper and
3
evaluates their reliability, consistency and representativeness.2 The fourth part outlines
the method whereby national indicators of economic inequality at long intervals can be
generated and displays the results obtained for Portugal. Its principal claim is that this
was a case of a long term inequality decline from the sixteenth to the 18th centuries and
was probably growing less unequal than the more advanced economies of Europe of the
time. The fifth section shows that van Zanden’s (1995) neo-kuznetsian hypothesis
linking macroeconomic performance and economic inequality fits awkwardly with this
account of Portugal’s inequality. It suggests an eclectic approach to a revision which
encompasses both the experiences of “declining” economies (Portugal) and “advancing”
ones (the Netherlands). The final section concludes.
2. Early Modern economic inequality: Data and estimation problems
2.1 Sources: social tables or fiscal data?
Historians of pre-industrial economic inequality have relied mainly on two types of
source: social tables and fiscal data. Both have advantages and weaknesses but, as table
1 below shows, this does not prevent the emergence of a consensual picture of the
principal features of Early Modern inequality.
Social tables are “highly educated guesses”, made by well-informed contemporaries,
who report average income and the number of its recipients by social class for a given
country, at a given moment (Milanovic, Lindert and Williamson, 2007 and 2011;
Williamson, 2009). One of their strengths is that usually they were organized to provide
This responds to Williamson’s (2009: 4) appeal to collect “new pre-industrial inequality evidence and
thus perhaps overthrow once and for all the historical persistence view that pervades modern debate about
Latin American inequality”.
2
4
full coverage of the country’s population or of a large region of a country. They are not
marred therefore by missing observation biases as occurs when there are significant
gaps in the distribution. Given that historically their aim was to assist in formulating tax
policies, they tend to be focused on income rather than wealth or other indicators. This
makes them an excellent source for calculating national indices of earnings inequality,
instead of having to employ proxies. Finally, they are particularly suitable for studying
societies such as those of the Early Modern period, in which mean income differences
between classes are large, intra-class divergence is small and class demarcation lines are
easy to detect. Estimating inequality on this basis is a comparatively simple undertaking
essentially because it only has to concern itself with between-class income differences
(Williamson, 2009).
Social tables are not free from defects, however. A doubt inevitably hangs over their
reliability. In part, it is a function of the extent of the guesswork involved and of the
aptness and biases of its author. In addition, they tended to be carried out in the service
of a ruler, to help maximize fiscal revenue, and this raises questions concerning possible
systematic distortions of the results.
Another important concern is the assumption that there was no variance within income
classes. Modalsli (2011) has argued, however, that Gini coefficients obtained from
social tables are prone to significant biases for two reasons which are related with this.
One is that in each class the entire group is concentrated at its mean income point,
which is improbable. The other is that there is no substantial overlap between the
income ranges of adjacent classes. If neither assumption is met, then this may imply an
upward correction of the estimated Gini coefficient of an order of between 30 and 100
per cent. Thus, there is a persistent downward bias in these estimates, and also a degree
of uncertainty as to its size.
5
A final worry regards the prospects of historical inequality studies based on social
tables. Not only is the pool of available material limited, but its supply is highly
inelastic. Since it seems unlikely that future research will uncover a great many more as
yet unknown ones of reasonable quality, historical inequality studies will have to be
written on the basis of the same small pool of information. This will be an important
limitation on the development of this field. 3 It seems important therefore to consider the
alternative offered by fiscal data.
Fiscal sources have been employed to good effect in a number of studies on the period
prior to 1800. Apart from being widely disseminated and well kept in archives, tax rolls
exist in large quantities and contain a massive amount of individual information
gathered at regular intervals on the households or householders of a given territory.
They are presumed to reflect, with reasonable accuracy, the personal income or wealth
of those listed at the particular moment (Soltow, 1987; van Zanden, 1995; Morrison and
Snyder, 2000). Their application was normally universal and they followed a consistent
methodology based on a flat rate charge. This made them less liable to error and
obfuscation. Generally speaking, the quality of these data in fact appears acceptable.
The emerging modern states of Europe armed themselves with surprisingly effective
mechanisms for estimating worth and collecting revenue. Tax rolls were put together
using the knowledge of crown officials, priests, magistrates and local “wise men”, and
were implemented on a house-by-house basis, to prevent evasion. Burocratic control
was used to dissuade favouritism or corruption, and a good deal of publicity
accompanied assessments, to help deter prevaricators.
3
Symptomatically, Williamson (2009) was obliged to calculate a number of Latin American Gini
coefficients for the pre-colonial, colonial and 19th century periods by means of regression analysis, given
the absence of social tables from which to construct them for this region. The title of the paper is
suggestively “History without Evidence”.
6
To obtain a national measure of inequality for a single year from a source of this type is
a highly laborious endeavour, however, which requires processing hundreds of
thousands of dispersed observations. Moreover, in most of Europe the administration of
such taxes was carried out by local authorities and the results were kept by them, not
centralized. To arrive at a satisfactory result consequently involves a heavy burden in
locating and pulling together a tremendous quantity of source material.4
An obvious problem with this information is its reliability. Despite all the precautions to
ensure procedural homogeneity, this was never easy to achieve. Ancien Regime France
offers a good example of attempts to implement universal direct taxes which were
repeatedly defeated in this aim. Given the pressure of local interests and the irresistible
tendency over time to reinterpret rules of fiscal assessment at lower levels of
administration, they tended to become idiosyncratic and the country was thus gradually
converted into a mosaic of fiscal localism (Morrison and Snyder, 2000).
The exceptional treatment under the tax law meted to certain individuals, groups of
individuals or occupational and other social classes presents even more formidable
difficulties. This was all the harder to eradicate as it involved very large numbers of
taxpayers and was an integral part of the social and economic fabric of pre-industrial
nations. In cases such as the clergy, the nobility, religious corporations, military officers
and servants of the Crown, fiscal exemptions were conceded having in view the
importance of their roles in society and in the power structure, or as a recompense for
services which, once rendered, then became fossilized in time (Santiago-Caballero,
2011). Given the substantial incomes and assets associated with these groups, a
Morrison and Snyder (2000) took a “small” sample of about 300,000 householders to study the income
distribution of 28 million Frenchmen in 1788. In all, they perused 189 rolls covering 56 departments.
4
7
downward bias in inequality estimates seems probable. At times, this could be
reinforced by placing a ceiling on high income valuations, though this effect could be
countered by striking the poor and the humble off the tax rolls, to save on the cost of
pursuing such a numerous and low income revenue target. Some “universal” taxes, on
wealth or on income, are known to have failed to reach as much as 70 per cent of the
population under consideration. (McCant, 2007; Soltow, 1980, 1985).5
2.2 Results and problems
The informational possibilities which arise when social tables and fiscal data are pooled
ought to ensure a rich empirical harvest on which to base the study of economic
inequality before 1800. Table 1 displays the majority of available estimates from the
literature on Early Modern Europe and the Americas. It reveals a set of thirty three
observations covering a wide range of situations and points in time. Altogether, they
originate in thirteen countries, cover the years from 1534 to 1808 and reflect the
distribution of economic achievement in terms of income, wealth accumulation or
expenditure on housing, usually on a household basis. It reveals a remarkable degree of
interest in the history of economic inequality and a considerable research effort. On the
other hand, it fails to go far enough towards satisfying the requirements of a
comparative analysis of the long run evolution of distributional disparities between
nations.
[Table 1 about here]
5
Morrison and Snyder (2000) correct for another bias which was widespread in Early Modern economies,
when assessments were made only with regard to monetized income. In fact, a large part of the population
living off agriculture was to some extent self-sufficient, with the result that a significant share of earnings
was ignored.
8
Apart from the deficiencies, vagaries and imperfections already noted above in the
sources, there are three major shortcomings to this collection. The first is that only part
of these country-year pairs can be used for our purpose. Roughly a third (13) only has
been collected from research materials which are truly “national” in scope (see col. 5).
None of the information on Spain, Germany or Italy satisfies this condition and parts of
that respecting the Netherlands, France and the United States suffer from the same.
Some of the estimates dubbed as “national” are, in fact, extrapolations from regional or
local data, sometimes in a manner which casts doubts on their representativeness. The
Dutch indicators of inequality are a case in point.6 One aspect is the poor coverage by
the data in question. In two instances, they are drawn from income tax findings (1742
and 1808), while in two others, from tax rolls relating to a levy on the value of
dwellings (1561 and 1732). All four of them, however, are heavily biased towards the
richer segments of the population and therefore have invited scepticism (McCants,
2007). A second problem arises because so-called “national” indices of inequality are
estimated by averaging population-weighted “local” Gini coefficients in order to obtain
the desired global figure. This procedure is incorrect, however, since the Gini metric is
not “decomposable” and consequently does not allow us “to write down a formula
giving total inequality as a function of inequality within the constituent subgroups, and
inequality between the subgroups” (Cowell, 2011: 64).7 This drawback is present in
many national inequality coefficients presented for the Netherlands, as well as in the
comparisons drawn by Williamson (2009) between inequality in Western Europe and
Latin America, all of them based on weighted arithmetic means of Gini coefficients.
6
For the Netherlands indicators, see Van Zanden (1995); and Soltow and Van Zanden (1998).
7
The measures of inequality which do permit this treatment are: variance, coefficient of variation,
Atkinson’s index, Dalton’s index, Theil’s index, MLD index, Herfindhal’s index and generalized entropy.
See Cowell ( 2011: 74). There is some doubt on this score, however. For a claim that Ginis can be
decomposed, see Soltow and Van Zanden (1998:18).
9
A final reason for shunning some of the estimates in table 1 is that the variables
summarized do not constitute a homogeneous group. Comparison between some pairs
can be therefore problematic. Proxying personal income with house rents, for example,
is bound to lead to difficulties unless it is preceded by a careful enquiry into the nature
of the relationship between these two variables.8 Likewise, comparing the evolution of
inequality in two countries, one of which - England and Wales – refers to income and
the other - Holland – to house rental values (Van Zanden, 1995) has to be questioned
too.
Finally, incomparability arises when the standard of inequality is applied to
different underlying data, e.g. income versus wealth (Van Zanden, 1995; McCants,
2007; Nicolini and Ramos-Palencia, 2011). This is only possible if the relationship
between the two variables is modeled with a reasonable degree of robustness so that one
can be converted into equivalent units of the other. At present, few exercises of this kind
are available, however, Nicolini and Ramos-Palencia (2011) and Williamson (2009)
being rare exceptions.9
Nevertheless, table 1 represents a considerable increment in our knowledge of historical
economic inequality. One of its achievements is to lend support to two notions which
have become consensual in the literature and will be important in the development of
the present paper. One is that Early Modern inequality in both Europe and the Americas
Expressions like “[rents] came quite close to the distribution of incomes” (Van Zanden, 1995: 648) need
to be substantiated. For a similar view on this relationship in late 18 th century North America, see Soltow
(1987).
8
9
Nicolini and Ramos-Palencia (2011) have estimated linear regressions of income on wealth plus control
variables. This should make it possible to infer the inequality of distribution of the former, indirectly,
from the latter. This second step has not yet been undertaken by them. Williamson (2009: table 2) has run
a regression to estimate the impact on the Gini coefficients of various historical cases in which one of the
explanatory variables is a dummy for the type of source used, i.e. “tax census” or “social table”. The same
procedure could be used with a dummy for “wealth” versus “income” data. This would provide a solution
for converting the values of one of these classes into the other and thus achieve comparability between
them.
10
was very pronounced by present day standards. In fact, it is rare to encounter for this
period Gini coefficients below 0.5, whilst the majority lie between 0.5 and 0.8. In the
case of wealth, this coefficient tends to be even greater, in some instances coming close
to the theoretical limit of inequality. Around 1800, the Scandinavian countries had the
highest scores – 0.9 or more – revealing societies where more than half the male
population over 20 years of age had no taxable property whatsoever, and a minute
proportion (0.1%) held more than half of the country’s entire wealth (Soltow, 1979,
1980 and 1985).10 Thus, on both sides of the Atlantic, “gross inequality” was
undoubtedly one of the hallmarks of pre-modern economies (McCants, 2007).11 The
practical significance of this finding for the present study is that it provides a
benchmark for the range within which any estimates produced by new research must
occur, if they are to be considered plausible.
The second generalisation comes from the possibility of ordering the sub-national
indices of inequality presented according to the socio-economic characteristics of the
human aggregates they represent. As has been noted, rural inequality in Europe during
these centuries tended always to be lower than urban (Van Zanden, 1995). In turn, the
scale of urban communities related to the extent of the disparities which they displayed.
In 18th century Spain, for example, the larger villages of rural Guadalajara were more
unequal than the smaller ones (Santiago-Caballero, 2011). In Madrid the Gini for
income was 0.77, while in Jerez de la Frontera, a small peripheral town, it was 0.500
10
The UN’s list of country income distribution reveals that outside Latin America and Africa, Ginis
above 0.5 do not exist. The majority of those in the group of the world’s most unequal societies have
Ginis between 0.5 and 0.6. In 2005, the average for the EU was 0.31 and the coefficient for the USA was
0.47.
Morrison and Snyder’s assertion that pre-revolutionary France’s Gini of 0.59 showed that this country’s
“income inequality was probably as high as it has ever been anywhere in modern times” thus strikes an
over-optimistic note (2000:77).
11
11
(Alvarez-Nogal and Prados de la Escosura, 2007). A finer distinction is exemplified by
income levels in the Dutch province of Overijssel in 1750. Among those of similar
scale, towns which were prosperous showed a higher degree of inequality and stagnant
ones, conversely.
These inequality rankings in association with social and economic characteristics can be
helpful in constructing national measures of income disparity in two ways. One is that
they provide an additional rule of thumb for testing the quality of freshly mined
evidence, when sources are problematic or procedures less reliable. If a small village
had an inequality measure above that of the country’s capital city, this would mean
either that it was an outlier or, more probably, that there was an error of estimation. The
other is that they enable us to select the stereotypical localities on which we can build a
grid to serve as the basis for calculating national indicators. The latter would serve to
calculate a weighted mean of “local” indicators that could be both statistically
significant and historically representative. If an appropriate yardstick (i.e. a
decomposable index) were used to aggregate them, this would yield a national measure
of inequality built on a properly calibrated set of local coefficients, and which would be
comparable internationally, as well as over time. In the following section we show how
we attempt such a procedure in the case of Portugal.
3. Data on economic inequality in Early Modern Portugal
The issue of economic inequality in 20th century Iberia has recently attracted some
attention.12 As regards Spain, this has spilled back in time, and a clear picture based on
solid empirical foundations is thus emerging even for the Early Modern period
(Santiago-Caballero, 2011; Nicolini and Ramos-Palencia, 2011; Alvarez-Nogal and Prados de la
12
See for example Alvaredo and Saez (2009), Prados de la Escosura (2008) and Atkinson et al. (2009).
12
Escosura, 2007, 2011). In the case of Portugal, the 20th century has also been the object of
some scholarly interest (Guillera (2011, 2010), Lains, Silva and Guilera (2008), Alvaredo
(2009) but its extension into earlier centuries, in the form of quantitative studies, has hardly
occurred.
This is not to say that the topic has not any had its place in the historiography of the
Portuguese Early Modern period. On the contrary, in the earlier and non-quantitative
conventional wisdom, income and wealth disparities enjoyed a significant presence,
which focused on three aspects. Ancien Regime Portugal was seen as profoundly
unequal, the vast majority living in abject poverty, while a tiny minority of nobles and
high clergy controlled enormous shares of wealth and income. Such inequality was a
major barrier to the country’s long term economic development and growth, mainly
because intermediate groups lacked the economic clout to drive this process (Godinho
([1971] 1980). These asymmetries were far from static, however, and there is a
periodization to be considered in this respect. Income distribution was relatively
“democratic” prior to 1500 but became increasingly unequal during the first overseas
empire (1500-1600) and again, after 1700, as a result of renewed colonial expansion and
the widespread concession of state monopolies. It declined in the seventeenth century,
when the urban bourgeoisie was able to break the nobility’s grip on trade and shipping
(Cortesão, 1950).
None of these views has been supported by much if any evidence or developed with the
correct analytic tools. They add relatively little to the understanding of historical
distributional issues, which has stagnated and remained out of touch with international
developments in this field. The exception is Johnson’s (2001?) quantitative study based
on data from several Portuguese localities between the fourteenth and the eighteenth
centuries. This relied on Gini coefficients to trace the trend of economic inequality over
13
480 years, an evolution to which it gave a Malthusian interpretation - population density
drives inequality. It foundered, however, on a weak empirical verification, owing to the
piecemeal nature of the data, its haphazard geographic dispersion and the lack of
consideration for the size and nature of the localities in the sample.
Portugal suffers from no dearth of primary material for the study of economic
inequality, as Johnson’s study reveals. The present paper has followed therefore the
quantitative approach which he pioneered. It introduces, however, four significant
improvements. The first is to enlarge and refresh his pool of eighteen “soundings”,
taking advantage of the large body of sources whose potential has now come to light.
The second is to pick the localities where observations of income or wealth distribution
can be carried out in a systematic fashion dictated by the aim of the exercise, instead of
leaving the choice simply to the chance of discovery. The third is to concentrate data
gathering on three benchmark years separated by extended intervals. This renders the
overall effort less onerous, without prejudicing the possibility of an accurate perception
of long run change. We have opted for 1550, in the middle of the first overseas
expansion; 1770, towards the end of Portugal’s second great colonial cycle; and 1700,
as an intermediate position between the economic lethargy of the seventeenth century
and the relative dynamism of the next seventy years.13 The last is to select locations
accordance with the grid of settlement types outlined above. For every benchmark year
each one of the following classes are suitably represented in the sample: major cities
(Lisbon and Porto); major towns (over 1,000 households); small towns (significantly
13
Benchmark selection has also been influenced by data considerations. The availability of detailed
information on the extraordinary tax of 1565 makes this year a good choice. The decline in the
comprehensiveness of the Décima tax after 1789 – all manual workers were excluded - makes it difficult
to proceed beyond this moment.
14
less than 1,000 households); and rural settlements. All of them are presumed to be
characterised by distinctive time-invariant patterns of economic inequality.
Having created the sample, the next step is to compose the respective indices of local
inequality into a single measurement, which should be representative of the country as a
whole. Instead of the more popular Gini coefficient, we use the Theil index, which is
not inferior in other respects, but, as mentioned already, has the virtue of being
“decomposable” (Santiago-Caballero, 2011). It can be correctly used therefore for
calculating the weighted average of local inequality estimates. If the locations have been
correctly selected, this will be close to the true measurement of national inequality and
will permit an enormous saving in research effort.
14
It will also make it feasible to
gather the necessary information for constructing the minimum number of benchmarks
necessary for a perception of inequality trends in the long run.
Of the three standard sources for studying the history of Early Modern economic
inequality, two are of little use in Portugal. Social tables prior to 1800 are inexistent and
probate inventories are too scarce and patchy to be of great use. 15 This leaves us with
sole recourse to the personal rolls which arose from direct taxation. Fortunately, not
only are these surprisingly abundant and well kept, but also of very good quality for the
task at hand.
Regular direct taxes became an integral part of the fiscal system only during the 1640s.
Prior to this, however, the crown occasionally used direct impositions, on a part of or on
the entire country, as an extraordinary measure to meet exceptional revenue needs. This
14
In order to permit comparisons with the results of other researcher, our presentation of the data sample
displays local observations in the form both of Theil and Gini coefficients.
15
The first social table in Portugal - Franzini, (1843) - refers to the early 1840s. For work on probate
inventories, see Rocha (1994) and Madureira (1989).
15
practice went back well into the middle ages (Gonçalves, 1966, 1964) and seems to
have had a broad scope in terms of its incidence, though it usually fell on wealth alone.
Typically, the motives were war or preparations for defence; festivities on the occasion
of special events in the royal family, or dowries for princes; and the consolidation of
crown finances. They had to be agreed to by the representatives of the estates in the
cortes, with whom they were contractualized. They were meant to be unrepeatable, one-off
events, and were subjected to close public scrutiny.
Given their variety and heterogeneity, it is difficult to employ the rolls from these
extraordinary taxes for inter-temporal comparisons on a national level, as is our present
aim. An exception, however, seems to have been the levy, in 1565, of 40 million reais
“for the service of His Majesty”, to consolidate the royal debt. It was applied uniformly
throughout the country according to the same procedures as were used later when
regular direct taxes were introduced. It allowed some exceptions but in principle was
universal and encompassed a very high proportion of the entire population.16
We base our first benchmark on this source. For the second and third ones, we resort to
the décima, an income tax adopted in 1641 by Portugal to finance its war of
independence from the Spanish crown. This brought an end to all extraordinary direct
taxes and ushered in a new fiscal era in which this regular annual imposition fell on all
forms of personal income, however obtained.
The new impost was applied to the whole population and allowed no exceptions, and
consequently ignored the traditional privileges of the nobility and the clergy. It was due
on all types of earnings from real estate, in whatever form; profits and income from
16
This source is well known, having been used for Lisbon by Rodrigues (1970), for Loulé by Magalhães
(1970) and for Coimbra by Oliveira (1971). The full text of the Lisbon tax roll is in Livro (1947-8).
16
employment in business, trade, transport, manufacturing and agriculture; and interest on
loans. Each head of household was obliged to pay jointly on as many of these activities
as those living in his hearth were engaged in (female heads of household were also
liable), multiple-sourcing then being a common occurrence.17 Initially the charge was
10 per cent on net income and remained so until 1668. It was lowered to 4.5 per cent at
the end of the War of Independence, and then raised again to 10 per cent, in the early
18th century (in 1705), due to a renewed threat of war. In 1715, it fell back to 4.5 per
cent with the stabilization of the international relations, and only brought back to the 10
% rate in 1762, when international tensions and war broke out again. It remained at this
level until 1800.18
How helpful are both types of tax as proxies for studying the distribution of income in
Portugal? Simply in terms of aims and design, a good deal seems to commend them.
The stipulation that they must be universal in their application suggests that a very
substantial degree of coverage was intended. The constant assertion by the authorities
that, in carrying out their assessments, officials must always “keep justice and equality”
and “be especially careful in judging the commercial, business and manufacturing
activity of each subject” (Oliveira, 1971: 308) reveals that the wish that the fiscal
burden should be truly proportional to each household’s true economic capacity. 19 It is
important also to stress that in practise the process was surrounded by precautions to
discourage any evasion of the system, either by omission or by under-valuation.
17
This contradicts the belief that in pre-industrial economies most households had only one source of
income. See Morrison (2000).
18
This chronology was pieced together from the information in the legal data base Ius Lusitaniae at
http://www.iuslusitaniae.fcsh.unl.pt/.
19
This is demonstrated by the successive regulations which prescribed in ever increasing detail how the
assessors were to calculate the liability of each household, and multiplied the number of different
situations which might hypothetically arise.
17
The key moment in the valuation of the tax base was the compilation of the complete
list of householders in every locality of the country and their respective responsibilities.
This contained the personal identification of each subject and data for estimating
individual fiscal capacity. In each municipality the task was coordinated by four experts
appointed by the local authority for their knowledge of the district and its economic
activity. They, in turn, selected two assessors for each street who lived in it and would
have draw up its list of taxpayers. They were supposed to be familiar with its
inhabitants and their respective incomes and wealth. The entire operation was overseen
by a royal magistrate without local ties, who was appointed by the crown and on whom
his career depended. A control was that the municipal list was assembled following
rigorous geographic principles. In urban centres, it was done street-by-street and houseby-house. In the countryside, it was done hamlet-by-hamlet or farm-by-farm. Any gap
in the tally had to be justified, making it difficult to conceal the “absence” of a taxpayer
or to justify an erroneous appraisal of somebody whom all locals knew.20
Although obviously they cast a remarkably fine net, in practice it is not obvious to what
extent direct taxation rolls were inclusive. In terms of gross coverage, they clearly
reached a very high proportion of potential taxpayers. In 1565, in Lisbon and Coimbra,
those listed in the rolls of the Serviço tax were more than 90 per cent of all households
at the time. Later, in the era of the Décima, in all cases where population counts are still
available, we have found similar proportions. This does not exclude the possibility,
however, of their being enormous distortions in the manner in which taxable wealth or
20
In the instructions drawn up in 1654 on how to compile a tax roll for the décima the following steps
were prescribed: 1) start with the list of parishioners, who will be called one by one to describe their
employment and assets, as well as the income from each one 2) as a check, gather information on these
cases from other people 3) walk up and down the street to check the names of the residents and see if any
are missing or there are new residents 4) prepare the assessments and write them in the two copies of the
roll, one to be kept locally and the other to be sent to Lisbon.
18
income were attributed to individuals in different income groups. In other words, were
the rich (and powerful) systematically favoured and the poor heavily burdened at the
moment of assessment? Figures 1A and B, which show the income distribution profiles
of the towns of Coimbra and Caminha, respectively in 1567 and 1767, suggest that this
did not occur on a scale which might seriously compromise our results. In both these
panels, as in all the other local observations of our sample, we note the expected very
long tail of extremely poor taxpayers and, at the other end, a very high concentration of
reais
Fig 1B Income distribution Coimbra 1567
16000
14000
12000
10000
8000
6000
4000
2000
0
1
141 281 421 561 701 841 981 1121 1261 1401 1541 1681 1821
taxpayers
income in the hands of a tiny share of wealthy heads of households.
Fig 1A Income distribution in Caminha1767
7000
6000
Reais
5000
4000
Series1
3000
2000
1000
0
1
26 51 76 101 126 151 176 201 226 251 276 301
taxpayers
19
More problematic for this research than biases caused by prevarication is their
likelihood as a consequence of exemptions given to certain social classes. The rules of
the Serviço tax, for example, declared that it applied to the “entire population” without
distinction, but then gave immunity to members of the religious orders and the royal
household, high judicial magistrates, doctors, university graduates, doctors of Theology,
knights and their widows and a few others. Moreover, it established a limit of one
million reais on taxable income, and a minimum tax of 16 reais on manual labourers
irrespective of their earnings or assets (Rodrigues, 1970). The former would bias the
measurement of inequality downwards, the latter would have the opposite effect, and it
may be that the two balanced each other out, though there is no way of controlling for
this. Outside Lisbon and therefore among the majority of the population, these
exceptions would be less numerous and would have had a lesser impact when gauging
distributional disparities.21
In the case of the Décima, the only significant exclusion was the clergy, since neither
the nobility nor all the other former privileged occupations and statuses were given any
relief from this obligation. Indeed, an examination of the tax rolls reveals the presence
throughout them of holders of titles of nobility and of the title of Dom. In the case of the
secular clergy - mainly parish priests – their names, properties and credit activities
appear in the rolls also. Although it is unclear whether they were obliged to pay the
corresponding tax, this enables us to include in our estimates a substantial part of their
incomes. On the other hand, the considerable revenues of the monastic orders and of the
bishoprics and archbishoprics and their distribution escape our scrutiny since they paid
21
Regarding the higher titular nobility, which appears only scarcely in the 1565 Lisbon tax roll, we
compensate for this by inputing available data on their estimated incomes from a 1529 list in Magalhâes
(2004), p.491.
20
a flat sum to the crown on account of the Décima. The burden of institutions was
allocated among them in a manner unknown to us. In any case, even if the number of
beneficiaries of these incomes and the amounts involved were known, there would be
no way of knowing how its was shared out within them.22
On the whole, the stock of rolls arising from direct taxes both before and after 1641
appears to offer a reasonable empirical basis for estimating a measure of economic
inequality at the national level across social classes, and allows us to make comparisons
over time. To obtain such indices, we need an adequate number of local estimates of
distributional inequality distributed among the four different types of population
settlement referred to above. Table 2 displays the shares in the national population
attributable to each of these categories, which will serve, in the following section, when
we need them to weight these “local” inequality measures in order to obtain a national
figure.
Table 2: Shares of national population by types of settlement
(thousands)
mid 16 c.
%
end 17 c.
%
late 18 c.*
Principal cities
(Lisbon and Porto)
144
8,34
205
9,32
238
8,21
Major towns
119
6,89
117
5,32
264
9,10
Minor towns
76
4,40
73
3,32
56
1,93
It is known that some orders were far “richer” than others but the distribution of income within each
one is impossible to know. For data on Church wealth, see Oliveira Marques and Fortunato de Almeida.
22
21
%
Countryside
1388
80,37
1805
82,05
2342
80,76
Total population
1727
100,00
2200
100,00
2900
100,00
Urbanization
rate
19,60
18,00
Source: Bairoch (1988) with adaptations.
Notes:major towns are >1,000 households; minor towns <1,000 households
Countriside= total - all towns
*= 1800
4. Results and discussion
Table 3 displays a set of summary statistics relating to direct taxation rolls over two
centuries which has been put together mainly from archives around the country. It is
organized into three panels – A, B and C - corresponding to each one of our three
benchmark years: 1550, 1700 and the 1770s. Aside from the standard Gini and Theil
measures of inequality, it shows the location and the date of production of each
document, whether it was from an urban or a rural setting, the number of households
described, and whether wealth or income were assessed. To give an idea of the levels of
prosperity across time and space, it includes the mean of the tax paid per household, a
very tentative indicator. Twenty one cases have been gathered to date, a number that is
growing still. Within each panel, they are ordered according to category: principal
cities, major towns, small towns and rural, dispersed settlement. They represent an
effort at constituting a geographically balanced sample, although some improvement is
still possible (see map).
[Table 3 about here]
All cases are complete, that is, they are supposed to include all households in the
locality which were liable to tax. The entire set therefore cuts across all income
22
19,24
categories and affords a global vision of income or wealth distribution at each specific
time and place. Three quarters of the present compilation are from Décima tax rolls,
making it therefore a reasonably homogeneous whole. Although we have discarded a
number of less satisfactory rolls used in Johnson (2001) because they contained
excessive fiscal exemptions and exclusions, some selection bias seems likely
nevertheless.23
Establishing comparisons on the basis of these inequality estimates raises two problems.
The first has to do with the “principal city” class comprising Lisbon and Porto. Owing
to the lack of consistent information for the two cities at each benchmark, we are
obliged to represent it first with Lisbon (1565), then with Porto (1700 and 1770), but
never able to use both at once. Between the 16th and the 18th centuries, Lisbon’s
population was always several times that of Porto, a differential which was bound to
translate, other things being equal, into a stronger inequality in the capital than its
northern counterpart.
23
Of his original eighteen rolls, we have retained here only five.
23
24
The second stems from the fact that, in classifying locations according to their
“structural inequality”, we have ignored whether regional and economic specificities
(e.g. administrative, manufacturing, port) might have affected their respective Theils
too. All we have taken into consideration is size and whether they were urban or rural.
Lack of data prevents us from dealing satisfactorily with the latter problem. A
satisfactory solution would require a substantially larger number of soundings than are
available. As regards the former, we try to overcome it by resorting to a linear
regression model to help us correct for possible biases arising from the urban size
factor. We estimate the expression
T =  +  * LHH +  * URB +  * T + 
UR?
(1)
in which T is the Theil coefficient (in logs), LHH is the log of thousands of households,
urb is a dummy for town or country (urban=1; country=0) and T is a dummy for the
century in which the inequality measure was taken. The error term is .
[table 4 about here]
Table 4 presents the results of this estimation, which are significant globally and for
each individual variable except for the 18th century dummy. This allows us to confirm
that size of locality (in logs of thousands of households) has an impact on the
measurement of inequality and that urban location has a strong inequality bias as well.
Moreover, it enables us to standardize the level of inequality for the “principal city”
category, whether we are basing it on smaller Porto or larger Lisbon. For example, we
can extrapolate what Lisbon’s inequality might have been in 1767, when it had three
25
times the households of Porto,
24
simply by increasing the latter’s Theil by a factor of
1,38.25
Before using the data in table 3 to analyse Portuguese income disparity, a preliminary
plausibility check is in order. One criterion is whether they conformed to the standards
of the Early Modern European context. We noted in section 2 that before 1800 measures
of inequality normally reached considerably higher levels than present day ones and this
is certainly the case with pre-industrial Portugal. Only two out of the eighteen Gini
coefficients displayed falls below 0.5, thirteen lie between 0.5 and 0.7, and three are
above 0.7. A second one is that at any given historical moment these values should fit a
descending pattern of inequality in accordance with size of locality and whether or not it
was urban. The principal cities should be highest, major and then lesser towns should
come next and rural communities should be lowest. Figure 2, presents the Gini
coefficient data from table 3 and orders them in terms of our four stereotypes of
inequality as well as by benchmark years. Where more than one observation is available
for a given time/location pair, we take their mean. In the other cases, we employ the
only value available. The conclusion is that the sample gives reasonable support to the
gradient hypothesis.26
24
Population figures for Portugal’s main towns et al. (1988).
25
The urban size effect is obtained by multiplying 300 percent (the proportion of Lisbon to Porto
households) by the elasticity of the “major cities”’ Theil (0.127 x 3.00 = 0.381). The inferred Theil for
Lisbon is equal therefore to 1.381 times that for Porto, i.e. 1.233.
The Theil indices for “major cities” have been corrected upwards in 1700 and in 1770 using Porto data
and the procedure described in the preceding footnote.
26
26
Theil
Fig 2 Local inequality in Portugal, 150-1770
1,8
1,6
1,4
1,2
1
0,8
0,6
0,4
0,2
0
1550
1700
1770s
Major
cities
Large
towns
Small
towns
rural
Apart from enabling us to undertake these two plausibility tests, figure 2 also provides
graphic evidence for a tentative answer to one of the main questions that motivate this
paper. The issue is the long run evolution of Portuguese inequality between the mid 16th
and the late 18th centuries. The hint provided by this figure comes from the general long
term decline visible throughout the data. The signs are not completely unambiguous,
however. Inequality in “large towns” falls and then rises, while in the major cities it
rises, then falls. On the other hand, those for small towns and the countryside evince an
unmistakable decline, which is important given that together they represent a large
majority of the population. Nevertheless, it would be wrong to conclude from this
sectoral distribution, without further examination, that country wide inequality was
necessarily declining. Even if all four sectors had experienced a common fall in their
respective Ginis, simultaneous inter-sectoral population movements could still have led
to an overall increase in inequality. This would happen, for example, if there had been
sufficiently large net shifts from the sectors with lower (rural) Ginis to those with higher
ones (i.e. urban centres with more than 5,000 inhabitants).
A second line of enquiry which can now be handled is the placement of Portugal in the
international ranking of economic inequality at this time. Was it unusually “unequal” by
27
European, American or Spanish standards? Figure 3 puts together the available relevant
evidence from tables 1 and 3. Unfortunately, this is less complete than one could have
wished, the reason being that consistent local data, which would also be comparable to
that gathered for Portugal is scarce in other countries European, and simply does not
exist for the Americas. For instance, Ginis estimated from rents and wealth have been
discarded because most Portuguese measurements are drawn from income rather than
wealth distributions. Our analysis is thus confined to the late 18th century and focuses on
a small handful of case studies which are nevertheless sufficient for our purpose.
Fig 3 Inequality in Portugal and other countries, 18c.
0,9
0,8
0,7
Gini
0,6
Spain
0,5
Netherlands*
0,4
Portugal
0,3
0,2
0,1
0
Pr
Ru
La
Sm
in c
rge
ral
all
ip a
t ow
tow
lc
n
n
it y
The main finding, on the basis of Ginis for Spain, the Netherlands and Portugal, is that a
strong pattern of international contrasts fail to emerge. The dispersion of local
inequality measures is not seriously pronounced in the categories of “principal cities”
and “countryside” and in the other two there is no clear leader. The fact that “large
towns” in Castile display less income disparity than Portugal and the Netherlands is
difficult to interpret given that they are smaller than their counterparts and that we lack
28
adequate information regarding this category for Spain. In any case, Portugal certainly
does not emerge as the most unequal. On the other hand, what is striking is the “gross
inequality” (McCants, 2007; Soltow, 1979, 1880, 1985) which appears to have
permeated uniformly the nations of Early Modern Europe, including Portugal. This
evidence, however, is too weak to allow any analysis of how inequality trends were
shaped by economic prosperity or any other factors, in particular in the case of the less
well-off Iberian countries. We therefore turn now to estimating, as accurately as
possible, the degree of economic inequality in Portugal for each of benchmarks we have
selected.
Given the already noted limitations of the Gini methodology, this exercise relies on
local estimates of Theil indices [GE(1)] from column 8 of table 3. It thus takes
advantage of the decomposability of this index (Sala-i-Martin, 2002) while using it,
however, inversely in relation to the usual procedure. Rather than starting from an
existing national estimate - which we do not possess - and decomposing it into subindices of inequality, we follow an “additive strategy”. This means calculating a
national Theil from the available “local” Theils weighted by their respective populations
and per capita outputs. The standard expression for this is
T =  (Nj/N)*(j/M)*Tj +  (Nj/N)*(j/M)* ln (Mj/M)
(2)
where T is the national Theil, Nj and N are the sizes of the “local” and national
populations respectively, Mj and M are their respective production means and Tj are the
local Theil indices of each of the four reference groups. The means of the local
production are proxied by nominal per capita tax ratings (col. 5, table 3), standardized
by the variations in the rate of taxation (col. 9), and the national mean is their
population-weighted average (see table 2 for the decomposition of population).
29
[Table 5 about here]
The result is shown in table 5. The path of economic inequality it reveals is a downward
sloping line, with a low gradient from the middle of the sixteenth century to 1700 and a
sharper one in the ensuing seven decades or so of the eighteenth century. This estimate
reveals several unusual features. In the first place, it represents possibly the longest time
span (1550-1770s) for which we have consistent and comparable national indices of
inequality for an Early Modern European country.27 Secondly, it is the first instance of a
non-core economy of the Early Modern period for which inequality measurement exists
on a large temporal scale. In the third place, the steady decline of Portuguese inequality
over more than two centuries sets it clearly at odds with the better-studied examples of
England and the Netherlands. Of these two, the former witnessed a decline during the
first half of the eighteenth century and then a rise over the next fifty years. The latter
experienced a prolonged increase in income disparity between 1590 and 1740, followed
by stabilization up to around 1800 (Soltow and van Zanden, 1998; Lindert, 2000). This
is especially important from the viewpoint of testing the relevance of the Kuznets curve
for the pre-industrial age, given that it is on these two case studies alone that this
discussion has rested empirically so far. The present new run of Portuguese Theil
indices thus provides a chance, for the first time, of verifying this approach in the
context of an economy which was quite different in more than one aspect from its
northern counterparts.
5. Portuguese inequality and the Early Modern Kuznets curve
27
Morrison’s (2000) comparative study encompasses several significant time spans for different countries
but they are shorter and cover mainly the history of modern industrialisation.
30
The original Kuznets curve model (1955) was devised for analysing economic
inequality in societies undergoing processes of Modern Economic Growth. More
recently, it has been adapted to account for the worsening levels of inequality witnessed
in the Early Modern period among its more dynamic economies. It incorporates three
mechanisms. One of them highlights the fact that during Early Modern economic
growth a certain share of the labour force moved from low to high productivity
occupations, thereby reinforcing income disparity. This process involved the growth of
cities, under the influence of international trade expansion, export industry growth,
rapid capital formation and the inordinate accumulation of urban fortunes (Van Zanden,
1995; Morrison, 2000; Lindert, 2000). A second occurred as a consequence of the
growth of per capita income in a context of fixed natural resources and led to a transfer
of “functional income” from labour to the other factors of production – land and capital
– mainly in the agricultural sector. A third arose from scarcities in the human capital
and skill markets, which were also generated by the rising prosperity of these
economies and led to an increased remuneration of these assets. The joint consequence
of all this was a movement towards greater inequality, which started before the onset of
industrialization and linked up to the classical inverted U shaped curve identified earlier
by Kuznets. This gave place to what has been termed a much longer “super Kuznets
curve” (van Zanden, 1995), linking pre-industrial and industrial economies in the
perspective of the study of their respective inequalities.
The case of Portugal contradicts this picture. Instead of a “super Kuznets curve” during
the Early Modern what emerges is a progressively more equal society. Instead of a
dynamic macro performance, it comes out as a stagnant economy in which structural
change is notable for its absence. Does this apparent clash with the current conventional
31
wisdom require a different explanation? Or is it in fact the counterfactual which
confirms the neo-Kuznetzian model?
This section discusses two issues. One of them is whether the descending Portuguese
inequality curve can be ascribed to some of its economic features which are most
relevant to this question. The other is whether a connection can be established between
Portugal’s relative economic decline and the long trend improvement of its
distributional profile. If it did, it would not only enable us to fit this case-study into a
more encompassing version of the super Kuznets curve concept. It would also allow us
to include in the model European countries with divergent macroeconomic
characteristics whilst accounting for their respective inequality profiles with a common
vector of causes.
Figure 4 (panel A) presents the information needed for a preliminary analysis. First we
note that, despite an unusual burst of growth in the first half of the 18th century,
Portugal’s overall macroeconomic performance was characterised by a stagnant
behaviour of GDP per capita between 1500 and 1800, which, by the latter date, had
declined by some 20 per cent relative to its early 16th century level. This was
accompanied by a marked stability of the urban share of the population, a favourite
indicator of structural change, which hovered persistently just below the 20 per cent
level (Figure 4 panel B) during the same period. The same constancy can be equally
seen at a more detailed level of scrutiny (table 2). The shares of population in large,
medium and small towns also remained essentially unchanged. If we put these facts
together with the country’s three inequality benchmarks, the picture which comes to
light suggests the opposite of the scenario depicted for the dynamic and less equal
economies of Northwest Europe. While in the latter, “economic development,
urbanization and capital accumulation […] went hand in hand with an increase in
32
inequality” (van Zanden, 1995: 649), in Portugal we find growing equality concurring
with the opposite of those features.
100
100
80
GDP pc
120
80
60
60
40
40
20
20
0
0
Theil
Figure 4 - Panel A: Explaining Theil
1500 1550 1600 1650 1700 1750 1770 1800
Theil
GDPpc
Linear (GDPpc)
0,50
80
0,40
60
0,30
40
0,20
20
0,10
0
0,00
shares
100
15
00
15
50
16
00
16
50
17
00
17
50
17
70
18
00
Theil
Fig 4 - Panel B: Explaining Theil
Theil
share non agric
share urban
To verify the plausibility of an “enlarged” neo-Kuznetzian model, we need to check
whether Portugal’s history of economic inequality can be interpreted with the same
tools as the Netherlands and England have been by van Zanden (1995). To begin with,
we clarify further the question of economic structure. An alternative to the urbanization
33
indicator, namely the share of the non-agricultural labour population, is displayed in
table 4-B. This combines urban and rural workers who were occupied in the tertiary and
secondary sectors, and, in a sense, is perhaps a better indicator of structure when
analysing a society in which a significant part of the “modern sector” may not have had
an urban location. The graph shows that structural change did occur from the fifteen
hundreds to the seventeen hundreds, although the economic stagnated and yet there was
no worsening of the distributional situation, on the contrary.
To resolve this, we have to consider that despite these macro-economic conditions, nonagricultural workers were able nevertheless to increase their share in the labour force.
This could only happen without any impact on inequality presumably because the
corresponding changes in pay and in productivity were quite small. They would earn
therefore only slightly more than unskilled workers, and therefore failed to reach levels
of compensation akin to those of skilled urban workers, as would have happened in a
more dynamic economic transition. Such a shift from the primary sector would in fact
have lowered rather than raised disparities since it would have enlarged the earnings of
a part of the formerly unskilled and poorest elements of society, without going so far as
to raise them to the level of the better-off strata of the labour force. Had it been
otherwise, such a shift would have increased the national Theil or Gini indices, as
posited in the original Kuznetzian formulation.
At present, we are not in a position to verify this hypothesis. This would require an
analysis of the micro data regarding individual earnings which is to be found in tax
rolls. These are available but have not yet been processed. The aim would be to
establish what sort of distributional readjustment between the deciles had occurred
during times of long run economic stagnation, with a particular focus on the lower 40 or
50 percent of incomes, i.e. the very modest ones earned only skilled and barely skilled
34
workers. It might reveal the effect of switching from agricultural to non-agricultural
activity without leaving the countryside and show how and to what extent, in a stagnant
Early Modern economy, alterations in employment could lead to the attenuation of its
inequality, rather than its elevation.
The next step in this enquiry takes us to the topic of shifts in “functional income” and
their influence on inequality. In Early Modern Portugal, an increasing scarcity of land
relative to a prevalently agrarian population drove down productivity in the primary
sector. Consequently the real wage of unskilled labour, which seems to have enjoyed a
fairly integrated market, fell sharply in the long run too, as Table 4-C shows. One might
40
40
20
20
0
0
real wage; agric product.
60
theil
Agric productivity
real wage
18 00
60
17 70
80
17 50
80
17 00
100
16 50
100
16 00
120
15 50
120
15 00
Theil
Figure 4 - Panel C: Explaining Theil
have thought that this reduction in the reward to the mass of the labour force would
have been reflected also in how total income was shared among production factors, and
have led to a shift in favour of the landed classes. A rise in the general distribution of
income would seem a logical outcome to this.
In fact, the ratio of rent to wages, the proxy usually employed to evaluate inequality in
pre-industrial societies, yields an initially unexpected result (Bertola et al, 2010). Figure
5 shows that during the 17th and the first half of the 18th century (data for the 16th
35
century is unavailable) the “functional income” division clearly favoured labour, i.e. the
poorest deciles of the population. During the latter part of the eighteenth century, this
movement was reversed in favour of those controlling the land, although by 1770, it had
only regained the level reached in 1650. This is hardly what one would expect in the
light of what occurred in the Anglo-Dutch economies in this respect, but there can be
little doubt as to its role it played in shaping income disparity, which was to reduce it up
to the mid 18th century, and to have lessened it afterwards.
1771
1760
1749
1738
1727
1716
1705
1694
1683
1672
1661
1650
1639
1628
1617
1606
18,0
16,0
14,0
12,0
10,0
8,0
6,0
4,0
2,0
0,0
1595
ratio
Fig 5 Rent-Wage ratio, 1595-1780
This is not the place to discuss in depth the long term movement in Portugal of the price
of land relative to that of labour but three clues for resolving this puzzle can
nevertheless be suggested. One is that the trend of this ratio in Portugal is very similar
to that in Spain (Alvarez-Nogal and Prados de la Escosura, 2011), which lends
credibility to our data and implies that a broader, Iberian explanation is necessary
though currently not available. Another is the corroboration by recent research on the
great landed aristocratic families of the post 1750, whose losses of income and wealth
have been described for that period as “the twilight of the Great”. Finally, we may
hypothesize that this may have been an instance of Milanovic, Wiliamson and Lindert’s
(2010: 268) rule that “in a declining economy, a given measured inequality translates
36
into a greater extraction rate”, something that ruling groups might be fearful of having
to live with. With real wages falling over the long run, the landed may have been ready
to accept an income shift in favour of labour and lower inequality, rather than having to
increase coercion in order to augment the extraction rate.
Fig 6 Ratio of skilled-unskilled wages
2,50
ratio
2,00
1,50
1,00
0,50
ratio
1772
1759
1746
1733
1720
1707
1694
1681
1668
1655
1642
1629
1616
1603
1590
1577
1564
0,00
Linear (ratio)
We now turn to the impact on income disparities of differences in pay arising out of
differences in productivity and which relate in turn to the distribution of acquired
manual skills and intellectual competences. The evidence in figure 6 represents the first
of these, measured by the standard metric of the ratio of skilled-to-unskilled wages. It
indicates that during the two centuries under observation the skills market failed to
generate any pressure for raising distributional inequality and, if anything, it contributed
towards lowering it. This is not a startling discovery, given the picture we have already
of the Portuguese economy. In it, pressures for structural changes and for redistributive
movements through the market for factors of production were on the whole weak and
apparently were rapidly met by supply forces, even in urban settings. It contrasts clearly
with how these markets and ratios evolved in England and the Netherlands in times of
significant economic growth and the different effect this had on their inequality levels.
37
In contrast to the skill premium, comparatively little scholarly attention has been paid to
the more subtle income redistributions which also took place in Early Modern societies,
between medium and higher deciles, as opposed to among the lower ones, or between
the lowest and the highest. Of these two types, the former might may be especially
relevant not only because they can generate greater equality, but also encourage social
situations in which the middle strata gain ground on the very wealthy and powerful,
thereby creating dynamics of social mobility which are conducive to growth.
There are two ways of exploring these aspects. One is an analysis of deciles, which
compares the evolution over long time spans of the shares in total income of the low,
middle and high deciles respectively. This can be done with the help of the fiscal
records at our disposal, a feasible task which is currently being undertaken. The other is
to look at the ebb and flow of the incomes of the intermediate segments of this society,
of individuals who lived neither from large land ownership, political office or major
business, nor from the sweat of their brows. Typically, they were engaged in “white
collar” activities, namely the liberal professions or the administrative classes which
served the state. In an earlier paper (Pereira et al, 2010), the earnings of these two
groups were compared to those of unskilled workers in Coimbra, a major Portuguese
town with a large and diverse non-manual salaried labour force. Between 1550 and
1770, the nominal wages of unskilled labour increased by a factor of three. In the same
interval, the salaries of professors of the university, choristers and organists, burocrats,
librarians and surgeons, rose by between four and six times, a trend which resembles
what happened with salaried employees in cities like Amsterdam and Vienna (van
Zanden, 1995).
Shifts such as these could obviously be responsible for some of the downward
movement in Theil estimates which we observed in table 5, and in reality were probably
38
comprehended more extensive groups than those in the example we have just given.
They were also occurring throughout the country amongst larger farmers who were
successfully transforming agricultural practices, small merchants who were rising
thanks to the protection of the state, returned colonial civil servants and the like. This is
a vast and complex reality and much of the currently available evidence on it is
anecdotal. The signs of a society moving towards greater equality and driven by a
middle stratum in Portugal is, however, a hypothesis which deserves further
exploration.
This section, on balance, has served two purposes. Firstly, it has helped to corroborate
the reliability of our Theil estimates by showing a good fit between them and some of
the principal features of their economic context. In the second place, it has shown that
the Portuguese inequality trend between the 16th and the 18th centuries can be
satisfactorily explained in terms of the same variables as underpinned the earlier
examination of the very different Dutch and English experiences in this domain. This
suggests the possibility of a common European model to fit these disparate situations
and one which has at the base the macro performance of the respective economies. At
the same time and in the light of the evidence furnished by the Portuguese case-study,
two new items may be proposed in order to enrich the analysis. One is that, for the sake
of generality, the notion of a “super Kuznets curve” is heuristically less important, if not
altogether dispensable, given that it is absent in some countries, possibly in a majority
of them at this time. The other is that for the completeness of the inequality account
social, political and institutional factors are important too. This means that not only we
need to study more countries but we also have to know much more about the conditions
under which the Milanovic et al’s extraction rate influenced distribution in them.
5A. Latin Inequality in comparative perspective
39
A final point takes us back to the question of whether Latin inequality was more
pronounced than in non-Latin Europe. Our reasoning starts with the notion that
medieval inequality was a good deal less than what it became during the Early Modern
period. As regards Portugal, this appears to be correct, the prevailing view of the nonquantitative historiography being that its society was relatively egalitarian until the
creation of its empire started in earnest from 1500. Johnson (2001) has adduced
evidence in support of this and of the subsequent alteration of Portuguese society in the
16th century by rapid urbanization, overseas trade and the development of a centralized
state. Together, all of these led to the emergence of very wealthy strata of Portuguese
businessmen, courtiers, financiers and so on, living mostly in the capital. We can thus
presume that income distribution worsened from, say, the end of the fifteenth century
and that this proceeded until the mid or late 1500s.
By this time, the factors that had provoked this change were weakening. The empire
was in retreat, and it seems reasonable to infer that the long swing of falling inequality
portrayed in table 5 had started. In the Netherlands, a similar process of economic
development and concurrently rising disparity of incomes started later, circa 1590, and
was to go on until the mid-eighteenth century. A plausible scenario might then be that
when Portugal’s inequality peaked it did so earlier than elsewhere in Europe. At this
stage it may well not have been higher than in the contemporary Netherlands.
Meanwhile, from 1600 or perhaps a bit sooner, Dutch inequality rose continuously for a
century and a half, a time in which in Portugal it fell steadily. By the 18 th century
therefore, there is reason to believe that Portuguese inequality had been overtaken and
that thereafter, as it continued to fall might therefore be falling behind the Netherlands.
In other words, Portugal may well have been less unequal than the Dutch.
40
6. Conclusions
This paper set out to make three new contributions in the study of Early Modern
economic inequality. The first has been to enlarge and improve significantly the pool of
information on Portugal and thus contribute to the burgeoning European data base in
this field. This has shown that fiscal information is not only a rich source, but one which
promises to break a looming data bottleneck in the field. Using a Theil approach, the
result has been a set of national, consistent benchmark estimates over a considerable
time span, which shows that income distribution in Portugal declined steadily over two
hundred years, starting from the middle of the sixteenth century.
The second aim was to introduce a new player in the debate on inequality on both sides
of the Atlantic and thereby broaden the debate surrounding the roots of economic
backwardness in Latin America. A satisfactory international comparison using Portugal
is as yet not possible and it is unclear how this country would stand in it vis-à-vis Brazil
and other former Latin American colonies. The Portuguese data confirm, however, that
high inequality was indeed the norm during these centuries, in Europe as in Latin
America. Moreover, it shows that, at least by the eighteenth century, if not earlier,
Portugal was probably less unequal than the Netherlands and possibly France too.
Exceptional “Latin inequality” in the Early Modern period may after all simply be a
fiction.
Our third finding lies in the interpretation of these results. Given the aim of reaching an
encompassing explanation, we focused on the task of unifying our understanding of two
well-known and contrasting situations. One was the case of economically successful
countries like England and the Netherlands, where the rise of per capita GDP over a
long period led to an increase in distributional inequality. The other was Portugal, which
41
experienced slow structural change and economic stagnation and followed a downward
trend of income disparity. An explanatory model in the Kuznets tradition, using the set
of variables proposed in van Zanden’s (1995) study, was applied. It showed its capacity
to account for both rising and declining experiences of inequality. It also revealed that
the notion of a “super Kuznets curve” probably reflects only a small part of what
happened in pre-industrial Europe and is not essential to the broader analytical effort we
have pursued here. This is something, however, that only future research can elucidate.
It also brought to light the need to provide a role for non-economic factors in this
discussion, this time in the tradition of Milanovic et al’s concept of inequality extraction
ratios, again, a task for future research to worry about.
References
Alvarez-Nogal, Carlos and Prados de la Escosura, Leandro (2007), “The Decline of
Spain (1500-1850): Conjectural Estimates, European Review of Economic History, 11,
pp. 319-366.
Alvaredo, Facundo and Saez, Emmanuel (2009), “Income and Wealth Concentration in
Spain from a Historical and Fiscal Perspective”, Journal of the European Economic
Association, 7 (5), pp.1140-67.
Alvaredo, Facundo (2009), Top Incomes and Earnings in Portugal 1936-2005
Explorations in Economic History, 46, (4), pp. 404-417.
Atkinson, A., Piketty, T. and Saez, E. (2009): “Top Incomes in the Long Run of
History”.National Bureau of Economic Research (NBER) Woking Paper No. 15408.
Cambridge: Massachusetts, October.
Bairoch, Paul, Batou, Jean and Chèvre, Pierre (1988), La Population des Villes
Européennes, 800-1850 (Geneva: Librairie Droz).
Bertola, L., Castelnovo, C., Rodríguez, J., and Willebald, H. (2009): “Income
Distribution in the Southern Cone during the First Globalization and Beyond”.
International Journal of Comparative Sociology 50, 5-6, pp. 452-485.
Bertola, L., Prados de la Escosura, L. and Jeffrey G. Williamson (eds) (2010) “Specail
Issue on Latin American Inequality”, Revista de Historia Económica. Journal of Iberian
and Latin American Economic History, 28, pp.219-26.
42
Carvalho, Joaquim de and Paiva, José Pedro (1989), “A Diocese de Coimbra no Século
XVIII. População, Oragos, Padroado e Títulos dos Párocos”, Revista de História das
Ideias, pp. 175-268.
Conde, Antónia Ralho (2002/3), “O Mosteiro de S. Bento de Cástris e a Décima
Eclesiástica”, Revista Portuguesa de História, XXXVI, pp. 161-72.
Cortesão, Jaime (1950), “Alexandre de Gusmão e o Tratado de Madrid”…
Cowell, Frank A. (2011), Measuring Inequality (Oxford: Oxford University Press).
Dobado Gonzalez, Rafael and Garcia Montero, Héctor (2011), “Colonial Origins of
Inequalitty in Hispanic America? Some Reflections Based on New Empirical
Evidence”, Revista de Historia Económica. Journal of Iberian and Latin American
Economic History, 28, pp.253-77.
Engerman, Stanley L. and Sokoloff, Kenneth L. (1997), “Factor Endowments,
Institutions and Differential Paths of Growth among New World Economies: A View
from Economic Historians of the United States” in Stephen Haber (ed), How Latin
America Fell Behind. Esays on the Ecoinomic Histories of Brazil and Mexico, 18001914 (Stanford: Stanford University Press).
Frankema, Ewout (2009) Has Latin America Always Been Unequal? A Comparative
Study of Asset and Income Inequality in the Long Twentieth Century. Boston: Brill.
Franzini, Marino Miguel (1843), Considerações acerca da Renda Total da Nação
Portuguesa e sua Distribuição por Classes com algumas Reflexões sobre o Imposto da
Décima (Lisboa: Imprensa Nacional).
Godinho, Vitorino Magalhães (1965), Os Descobrimentos e a Economia Mundial
Godinho, Vitorino Magalhães (1968-71), Ensaios (4 vols), (Lisboa: Sá da Costa).
Godinho, Vitorino Magalhães ([1971] 1980), Estrutura da Antiga Sociedade
Portuguesa (Lisboa: Arcádia, 4th ed.).
Gonçalves, Iria (1964), O Empréstimo Concedido a D. Afonso V nos Anos de 1475 e
1476 pelo Almoxarifado de Évora (Lisboa: Ciência Técnica Fiscal).
Gonçalves, Iria (1964), Pedidos e Empréstimos em Portugal durante a Idade Média
(Lisboa: Ciência Técnica Fiscal).
Guilera, Jordi (2010): “The evolution of top income and wealth shares in Portugal since
1936”, Revista de Historia Económica – Journal of Iberian and Latin American
Economic History, Volume 28, Issue 1.
Guilera, Jordi (2011), “Estimating inequality, understanding history: Did inequality
shape the long path to democracy in Iberia?” Iberometrics V.
43
Hoffman, Philip, Jacks, David, Levin, Patricia A. and Lindert, Peter (2002), “Real
Inequality in Europe since 1500”, Journal of Economic History, 62, pp.322-355.
Johnson, Harold B. (2001), “Malthus Confirmed? Being some Reflections on the
Changing Distribution of Wealth and Income in Portugal (1309-1789)”, unpublished
manuscript.
Kuznets, Simon (1955), “Economic Growth and Income Inequality”, American
Economic Review, 45, pp. 1-28.
Lains, Pedro, Gomes, Esther and Guilera, Jordi (2008), Are Dictatorships more
unequal? Economic Growth and Wage Inequality During Portugal’s Estado Novo,
1944-1974, Working Paper Universidad Carlos III.
Lindert, Peter H. (2000), " Three Centuries of Inequality in Britain and America” in
Anthony B. Atkinson and François Bourguignon (eds), Handbook of Income
Distribution, Volume 1, pp. 167-216.
Livro do Lançamento e Serviço que a Cidade de Lisboa fez a el-Rei Nosso Senhor no
ano de 1565 (1947-48), ( (Lisboa, Câmara Municipal, 4 vols.)
Lopes, Maria Antónia (2000), Pobreza, Assistência e Controlo Social. Coimbra (17501850) (Viseu: Palimage).
Lopez, J. Humberto and Perry, Guillermo (2008), Inequality in Latin America:
Determinants and Consequences, The World Bank: Latin America and the Caribbean
Region, Policy Research Working Paper 4504, Washington D.C.
Madureira, Nuno Luís (1989), “Inventários: Aspectos do Consumo e da Vida Material
em Lisboa nos Finais do Antigo Regime”, Master’s Dssertation, Universidade Nova de
Lisboa.
Magalhães, Joaquim Romero de (1970), Para o Estudo do Algarve Económico durante
o Século XVI (Lisboa: Cosmos).
Magalhães, Joaquim Romero de (1993), O Algarve Económico, 1600-1773 (Lisboa:
Estampa).
Magalhães, Joaquim Romero (2004), “Dinheiro para a Guerra: as Décimas da
Restauração”, Hispania, LXIV, pp. 157-182.
Marques, A. H. Oliveira (1980), Ensaios de História Medieval Portuguesa (Lisboa,
Vega).
McCants, A. (2007): “Inequality among the poor of eighteenth century Amsterdam”.
Explorations in Economic History 44, pp. 1-21.
Milanovic, Branko, Lindert, Peter H. and Jeffrey G. Williamson (2007), Measuring
Ancient Inequality, World Bank Policy Research Working Paper No. 4412.
44
Milanovic, Branko, Lindert, Peter H., & Williamson, Jeffrey G. (2010) “Pre- Industrial
Inequality”, Economic Journal, 121, pp. 255-272.
Modalsli, Jorgen (2011), “Inequality and Growth in the very long Run: Inferring
Inequality from Data on Social Groups”, Department of Economics, Oslo University
Working paper # 11/2011.
Monteiro, Nuno Gonçalo Freitas (1998), O Crepúsculo dos Grandes. A Casa e o
Património da Aristocracia em Portugal (1750-1832) (Lisboa: Imprensa Nacional-Casa
da Moeda).
Morrisson, Christopher (2000), “Historical perspectives on income distribution: The
case of Europe” in Anthony B. Atkinson and François Bourguignon (eds), Handbook of
Income Distribution, Volume 1, pp. 217-260.
Morrisson, Christian and Snyder, Wayne (2000), “The Income Inequality of France in
Historical Perspective”, European Review of Economic History, 4, pp. 59-83.
Nicolini, Esteban and Ramos-Palencia, Fernando (2011), “Comparing Income and
Wealth Inequality in Pre-Industrial Economies. Lessons from Spain in the 18th
Century”, unpublished paper presented at the Iberometrics V meeting, Barcelona.
Oliveira, António de (1971), A Vida Económica e Social de Coimbra de 1537 a 1640
(Coimbra: Faculdade de Letras da Universidade de Coimbra).
Prados de la Escosura, Leandro (2008), “Inequality, Poverty and the Kuznets Curve in
Spain, 1850–2000”, European Review of Economic History, 12, pp. 287–324.
Reis, Jaime, Martins Conceição A. and Costa, Leonor F. (2011), “New Estimates for
Portugal’s GDP per Capita, 1500-1850”, unpublished manuscript.
Rocha, Maria Manuela (1994), Propriedade e Níveis de Riqueza: Formas de
Estruturação Social em Monsaraz na Primeira Metade do Século
XIX (Lisboa: Cosmos).
Rodrigues, José Albertino (1970), “Ecologia Urbana de Lisboa na Segunda Metade de
Século XVI “, Análise Social, VIII, pp. 96-115.
Sala i Martin, Xavier (2002): The Disturbing ‘Rise’ of Global Income Inequality.
National Bureau of Economic Research (NBER) Working Paper No. 8904. Cambridge:
Massachusetts, April.
Santiago-Caballero, Carlos (2011), “Income Inequality in Central Spain, 1690-1800”.
Explorations in Economic History 48 (1), pp. 83-96.
Silva, Alvaro Ferreira da (1987), “Família e Trabalho Doméstico no Hinterland de
Lisboa: Oeiras, 1763-1810”, Análise Social, XXIII, pp. 531-562.
45
Sokoloff, Kenneth L. and Engerman, Stanley L. (2000), "History Lessons. Factor
Endowments, and Paths of Development in the New World", Journal of Economic
Perspectives, 14, pp. 217-232.
Soltow, Lee (1979). ‘Wealth Distribution in Denmark in 1789’, Scandinavian Economic
History Review 27(2): 1-18.
Soltow, Lee (1980). ‘Wealth Distribution in Norway and Denmark in 1789’,
HistoriskTidsskrift (Oslo), 59: 221-35.
Soltow, Lee (1985). ‘The Swedish Census of Wealth at the Beginning of the 19th
Century’, Scandinavian Economic History Review 33(1): 1-24.
Soltow, Lee (1987), “The Distribution of Income in the United States in 1798:
Estimates Based on the Federal Housing Inventory”, Review of Economics and
Statistics, 69, pp. 181-5.
Soltow, Lee and van Zanden, Jan Luiten (1998): Income and Wealth Inequality in the
Netherlands, 16th to 20th century. Amsterdam: Het Spinhuis.
Steckel, Richard H. and Moehling, Carolyn M. (2001), “Rising Inequality: Trends in the
Distribution of Wealth in Industrializing New England”, Journal of Economic History,
61, pp.160-83.
Williamson, Jeffrey G. (2009), History without Evidence: Latin American Inequality
since 1491, NBER Working Paper #14766.
Zanden, Jan Luiten van (1995), “Tracing the Beginning of the Kuznets Curve: Western
Europe during the Early Modern Period”, Economic History Review, XLVIII, pp. 64364.
Jan Luiten van Zanden, Joerg Baten, Peter Földvari and Bas van Leeuwen (2011) The
Changing Shape of Global Inequality 1820-2000: Exploring a new dataset GEH
Workiing Paper Seriies #1.
46
TABLE 1
The Measurement of Economic inequality:
Europe and Americas, 1500-1800
Place
SPAIN
Castille
Castille
Guadalajara
Madrid
Jerez
Castille/Palencia
Castille/Palencia
Castille/Palencia
Year
Population
Gini
coefficient
Location
Economic
variable
c.1750
1752
16901800
late 18 c.
c. 1750
c.1750
c.1750
c.1750
na
3,945 families
14,000
producers
160,000
na
106,000
97,000
9,000
0,39 - 0,56
0,525
towns/ country
towns/ country
Y
Y
0,500
0,770
0,500
0,527
0,523
0,557
rural
urban
rural?
rural/small town
rural
small town: Palencia
Y
na
na
Y
Y
Y
1688
5,700,000
0,556
whole country
Y
BRITAIN
England and
Wales
England and
Wales
England and
Wales
1759
na
0,552
whole country
Y
1801/1803
9,000,000
0,593
whole country
Y
GERMANY
Weimar
Eisenach
Augsburg
1542
1542
1604
426 hh
632 hh
na
0,640
0,650
0,890
town
town
town
rents
rents
W
town
whole country
whole country
city
city
southern towns
rural
whole country
W
rents
rents
rents
Y
Y
Y
NETHERLANDS
Alkmaar
Netherlands
Netherlands
Amsterdam
Amsterdam
southern towns
Netherlands
Netherlands
1534
1561
1732
1740-82
1742
1742
1742
1808
2,100,000
>0,750
0,560
0,61
0,584
0,690
0,700
0,600
0,630
FRANCE
Lyons
France
1545
late 18c
27,300,000
>0,750
0,590
city
whole country
W
Y
NORTH
AMERICA
USA
1774
919 HH
0,694
dispersed
Wealth
na
47
USA
1774
919 HH
0,642
dispersed
USA
1798
40,000HH
0,706
whole country
Wealth
house
values
1800
1800
1800
640,000
574,000
65,000
0,923
0,914
0,937
all>20 - whole
country
only rural
only urban
Wealth
Wealth
Wealth
1800
1800
1800
222,000
206000
16000
0.905
0.896
0.953
all>20 - whole
country
rural
urban
Wealth
Wealth
Wealth
1789
900,000
0,952
all >26 - whole
country
Wealth
Norway
1789
900,000
0,881
all >26 - whole
country
Wealth
LATIN AMERICA
New Spain
1790
4,500,000
0,635
whole region
Y
ITALY
Ivrea
Ivrea
Padova
Padova
1544
1632
1549
1642
na
4,500 inhabts
35,000 inhabts
32,700 inhabts
0.658
0.692
0.743
0.809
town
town
town
town
w
w
w
w
SWEDEN
Sweden
Sweden
Sweden
FINLAND
Finland
Finland
Finland
DENMARK
Denmark
NORWAY
Note: HH=household; Y= income; W= wealth; R= housing rents
48
Table 2: Shares of national population by types of
settlement
(thousands)
mid 16 c.
%
end 17 c.
%
late 18
c.*
Principal cities
(Lisbon and Porto)
144
8,34
205
9,32
238
8,21
Major towns
119
6,89
117
5,32
264
9,10
Minor towns
76
4,40
73
3,32
56
1,93
Countryside
1388
80,37
1805
82,05
2342
80,76
Total population
1727
100,00
2200
100,00
2900
100,00
Urbanization rate
19,60
18,00
Source: Bairoch (1988) with adaptations.
Notes:major towns are >1,000 households; minor towns <1,000 households
Countriside= total - all towns
*= 1800
49
%
19,24
TABLE 3
Gini and Theil coefficients: Local economic inequality
1
locality
2
date
3
4
5
6
Tx
paid
households
per
taxed
urban/rural cap Wealth/Income
7
8
9
Gini
Theil
Tax
rate
Panel A -circa 1550
Lisboa
1565
15020
U
430
Y
0,798 1,392
na
Coimbra
1567
1334
U
118
Y
0,697 1,178
na
Loulé
1564
694
U
(302)
Y
0,714 1,127
na
Loulé
1564
507
R
(156)
Y
0,553 0,538
na
4,50%
Panel B - circa 1700
Porto
16981701
5000
U
263,6
Y
0,660 0,998
Portalegre
siza
1725
1489
U
609
Y
0,648 0,887
1690
311
U
569
Y
0,656 0,829
4,50%
1698
706
U
293
Y
0,577 0,688
4,50%
1725
1690
1011
313
R
R
680
1260
Y
Y
0,541 0,559
0,635 0,791
4,50%
4,50%
Avis
Vila do
Conde
Portalegre
(siza)
Avis
Panel C- 1770s
Porto
1767
12000
U
1768
Y
0,663 0,893
10%
Guarda
1763-9
1042
U
735
Y
0,744 1,360
10%
Viseu
Caminha
Vila do
Conde
1763
1767
390
459
u
u
Y
0,696 1,024
0,585 0,753
10%
1763
706
u
709
Y
0,559 0,663
10%
1763
124
R
639
Y
0,461 0,279
10%
Vila do
Conde
50
Caminha
Galveias
1767
1753
2021
107
R
R
2720
0,483 0,401
0,61 0,750
Sources: Johnson (2001); various archives.
Note: Porto population is for whole city but inequality counts are for a number of parishes
only: see population file; "hh taxed" are actually simply hh
Table 4
Determinants of Theil Coefficients: a regression
(ordinary least squares)
Dependent variable: Theil coefficient
Explanatory
variables
coefficients
constant
-1,472***
0,007
16th c.
0,317**
0,049
18th c.
-0,123
0,533
Log households (000s)
0,127*
0,072
0,601***
0,000
Urban vs. Rural
p values
N = 22
R squared = 0,775
Adjusted R squared = 0,699
Sources: table 3
Notes: coefficients significant at 10, 5
and 1 percent levels denoted by respectively one, two or three asterisks
51
10%
4,50%
Table 5
Global Income Inequality in
Portugal:
Theil, 1550-1770s
Theil
within groups
across
groups
1550
0,918
0,817
0,0978
1700
0,872
0,768
0,102
Ca. 1770
0,654
0,624
0,034
Source: Data from table 3. For adjustments
and interpolations, see text.
52