The Education Gender Gap: Evidence Following the Italian...

1

The Education Gender Gap: Evidence Following the Italian Unification#

Graziella Bertocchi* Monica Bozzano**

Preliminary Draft, September 2012

Abstract We investigate the historical determinants of the education gender gap in Italy in the late nineteenth century, immediately following the country’s Unification. We use a comprehensive newly-assembled database including 69 provinces over twenty-years sub-periods covering the 1861-1901 period. We explore the effect of economic factors such as contemporaneous measures of income and industrialization, institutional factors captured by the past domination prior to Unification, and socio-cultural factors associated with medieval family types. Our focal variable is a proxy for the effect of medieval commercial routes, which is found to be positively associated with female primary-school attainment, relative to that of men, at the non-compulsory upper-primary level.

JEL CODES: E02, H75, I25, J16, N33, O15.

KEYWORDS: Education gender gap, Italian Unification, medieval commerce, family types, institutions.

# Generous financial support from Fondazione Cassa Risparmio di Modena and the Italian University Ministry is gratefully acknowledged. *University of Modena and Reggio Emilia, RECent, CEPR, CHILD and IZA. Address: Viale Berengario 51, 41121 Modena, Italy, e-mail [email protected]. **University of Modena and Reggio Emilia and University of Pavia. Address: Viale Berengario 51, 41121 Modena, Italy, e-mail [email protected].

2

1. Introduction

The reversal of the education gender gap, to the advantage of women, has been part of a quiet revolution which has gradually transformed women’s lives in the vast majority of OECD countries (Goldin, 1998, 2006; Goldin et al., 2006). To understand the drivers of this change (as attempted for instance by Fernandez and Wong, 2011) is tantamount to understand the reasons for the previous, long-standing inferior condition of women. Gender inequality is a multifaceted phenomenon that, beside education, requires consideration of its economic and political dimensions.1 Historically, the gender gap in wages and labor force participation has been even deeper than that in education,2 not to speak about the fact that less than a century ago in most OECD countries women were not even granted the most elementary political rights, i.e., the right to vote.3 Even though getting a grasp of the evolution of the education gap only allows to capture the tip of the iceberg, to understand the empirical determinants of the education gender gap and its evolution would still represent a crucial advancement which we try to achieve in this paper. We start by assembling a new dataset that covers a sample of Italian provinces in the 1861-1901 period, i.e., the initial forty years after the country’s Unification. To our knowledge, this is the first time that these early provincial data are employed in econometric analyses. The historiography of the period reports that over this crucial time span for Italian history. For instance, on the basis of regional data A’Hearn et al., 2011 report that the gender gap in primary school enrolment declines from 12.5 to about 5 percent, reflecting a general process of gender equalization and also, to some extent, of convergence across pre-Unitary states.4 In our search for the determinants of the education gender gap, after controlling for economic, geopolitical and social factors, we find that the cross-sectional variation of the gap within Italian provinces is influenced by the medieval pattern of commerce, along the routes that connected Italian cities among themselves and with the rest of Europe. Between the end of the thirteenth century and the fourteenth century, Italy was at the center of a renewed process of expansion of trade, city growth, and economic and social development. This period, which is referred to as the “heroic era of the merchant” (Sapori, 1972), was in fact characterized by an intensification of international exchange throughout Europe which was in turn associated with the progress of mercantile science an practices, such as banking, payment tools, 1 Hausmann et al. (2007) presents the components of the Global Gender Gap index and Bozzano (2011) adapts them to contemporaneous Italian data. 2 For instance, Goldin (2006) traces an economic history of the transformation of American women’s status in employment and education for birth cohorts starting from the end of the 1800 and shows that the gender gap in college education has been narrower than the gender gap in labor force participation. 3 Bertocchi (2011) studies the process of women’s enfranchisement in the 1870-1930 period. 4 Bertocchi et al. (2012) report that the reversal of the education gender gap is confirmed in Italian data over the post-World War II period.

3

marine insurance, as well as mercantile laws and bookkeeping techniques (Gibbins, 1891). It is exactly at this stage that, within the merchants’ communities, women quickly acquired a special role: during the men’s years-lasting travels, their wives, mothers, sisters and daughters often needed to become literate in order to run the family business (Ricci, 2011). This peculiar position of women is amply documented: in the middle of the fourteenth century Boccaccio’s Decameron narrates of a merchant from Genoa who praises his wife for her ability to write, read and count “as well as a merchant”. In a distinct but related context, the commercial necessity of literacy is also stressed by Spufford (1995) and invoked by Hoftijzer (2001) to explain the relatively high female literacy rates in Amsterdam during the seventeenth century. The effect of commerce on the promotion of women’s education ran through a number of parallel channels: on the one hand, the physical absence of men forced women to take charge; on the other, this transfer of duties from men to women was facilitated by the fact that, contrary to other occupations, commerce did not require physical strength; moreover, women found themselves in charge not only of trade, but also of their children’s education, which resulted into an intergenerational transmission of the acquired skills, role models, and beliefs. Female literacy in fact is the first quantitative signal of a complex cultural phenomenon of transformation of a historically contextualized mentality (Briggs, 2000). The relevance of each of the above mentioned channels for the position of women in society has been recognized, in isolation, in the relevant literature. For instance, the impact on women’s labor force participation of men’s absence - as the result of a war - is by Goldin (1991) and Duby and Perrot (1998). The influence of women’s inferior physical strength on human capital accumulation is modeled by Galor and Weil (1996) and a similar argument to explain the shaping of gender roles is advanced by Alesina et al. (2011) on the basis of Boserup’s (1970) hypothesis that plough (rather than shifting) cultivation determined the historical gender division of labor.5 The role of mothers’ education on children’s outcomes is stressed by Schultz (2002) and Doepke and Tertilt (2009), while Fernandez (2012) and Fogli and Veldkamp (2011) explores the implications of the associated cultural change. For the case of medieval commerce, all three channels are present and reinforce themselves.

In more detail, we perform our empirical investigation as follows. Our dependent variable is the ratio of female to male enrolment rates in primary school, which can be further disaggregated into a lower and an upper level, only the former being compulsory. Among the explanatory variables, we consider measures of the level of development which we capture with average male height, as a proxy of wealth, and industrialization. We also control for an initial condition for the education gender gap, which we measure with the ratio of the female to male literacy rate in 1861. Next, we control for the political geography of Italy prior to Unification: namely, we distinguish between territories ruled by the Savoy dynasty, the Austrian Empire, the Pope, and the Bourbons. This distinction should capture the effect of deep-rooted differences in political and educational institutions that may have shaped the evolution and persistence of gender disparities. Furthermore, since societal norms are often formed and transmitted through the family, we recognize the potential influence on gendered human capital accumulation of different family structures,

5 Since the plough system applied to the entire territory of Italy, we cannot impute to this channel regional differences in women’s outcomes.

4

according to a classification taken from Todd (1990) that we adapt to the Italian specific regional differentiation. Namely, we distinguish four family-types: incomplete stem, communitarian, egalitarian nuclear with delayed marriage, and egalitarian nuclear family with early marriage.6 Even after controlling for all the above correlates, we find that medieval commerce - after several centuries - is still affecting the education gender gap in the 1861-1901 period. In particular, we find that its effect on primary enrolment runs through enrolment in the upper primary level, while it is not significantly associated with the lower primary level which was compulsory and therefore less likely to be influenced by individual choices. Our findings are robust to a number of alternative specifications involving variants of dependent variables, regressors, samples, and estimation techniques. In particular, in order to control for omitted variables and measurement errors that are likely to bias our results, we perform 2SLS regressions where medieval commerce is instrumented with a variable capturing the presence of a university in early medieval times, i.e., between the eleventh and the thirteenth centuries. The rationale for the choice of this instrument comes from Cantoni and Yuchtman (2012) who, for a sample of German cities, show that medieval universities played a causal role in promoting the expansion of trade. The channels involved were the provision of juridical training, which in turn allowed the establishment of legal and administrative institutions, and the creation of networks of highly mobile individuals whose communication was facilitated by their ability to speak Latin as a common language. Finally, we show that medieval commerce exerts no significant influence on the level of schooling, which underscores its peculiar influence only though gendered human capital accumulation. Among the few papers that have focused on the education gender gap in historical perspective Becker and Wößmann (2008) stand out because they assess how Protestantism favored girls’ education by looking at sub-regional nineteenth-century data within Prussia. While a common explanation of such outcome is the stress on the study of the Bible within the Protestant tradition, Spufford (1995) has also noticed the coincidence between the high literacy rates in Protestant regions and its effect on commercial success. The rest of the paper is organized as follows. In Section 2 we provide a brief overview of the process that led to Italy’s Unification in 1861 and outline the process of human capital accumulation and the educational policies before and after this historical event. In Section 3 we present our newly-collected dataset on gendered educational attainment at the primary school level in 1861-1901. In Section 4 we show our empirical analysis and in Section 5 we conclude.

6 Following Del Panta et al. (1996), we add to Todd (1990) the latter distinction between egalitarian nuclear families with delayed vs. early marriage for women.

5

2. Education policies before and after Unification

The period immediately preceding Italian Unification is often referred to as Risorgimento, which runs from the end of the Napoleonic era in 1815 until 1861. The Unification of Italy represents a key event in the geopolitics of Europe (Dincecco et al., 2011) Ass illustrated in Figure 1, before this historical turning point Italy is divided among four main dynasties: the Savoys, ruling in the North West, i.e., in Piedmont, Liguria and Sardinia; the Habsburgs, holding direct control over the Lombardy-Venetia in the North East, as part of the Austrian Empire, as well as indirect control over the Granduchy of Tuscany and other minor Duchies nearby; the Pope, reigning over Rome and the Center South; and the Bourbons, in charge of the South and the Duchy of Lucca.

Figure 1. The geopolitics of Italy prior to Unification

Source: Shepherd (1926).

6

The pervasive regional differentiation in the institutional and cultural background inherited from Risorgimento represents the initial condition over which the Kingdom of Italy is established. Despite the fact that Unification favored convergence, the effect of initial conditions persists to the present days. In this brief narrative we are especially interested in the description of the processes that shaped the evolution of human capital accumulation through the education system. Pre-Unitary states presented deep differences along several dimensions, including literacy and schooling attainment, partially as a consequence of educational policies (A’Hearn et al., 2011). For instance, if in Lombardy-Venetia schooling was compulsory until age 12 and the Kingdom of Sardinia implemented a long term plan to regulate the educational system, in the Papal and Bourbons’ territories no such policies were present. In some areas of the peninsula there was not even a minimum of what primary school ought to transmit to pupils: in general in fact it was concerned with reading, writing and the first notions of maths, but in many cases it reduced to religion and, for girls, knitting and sewing.7 In other cases teachers were illiterate (De Mauro, 1963). Female education was largely neglected (Serristori, 1842) in most areas, with the exception of Lombardy, Rome, and the Duchy of Parma and Piacenza (Vigo, 1971). In the eve of Unification only 27 percent of the adult population, that means that only 5 million people over 22 millions of Italians were able to read. Only Piedmont and Lombardy showed literacy rates over 40 percent in line with neighbouring countries (A’Hearn et al., 2011). After Unification, the 1859 Casati Law disciplined the school system by introducing an initial compulsory two year level (extended to three year in 1877 by the Coppino Law) and a subsequent two year level (Bertocchi and Spagat, 1997; Bertola and Sestito, 2011). The Casati Law was explicitly inspired by the Prussian system of nationally directed education. The presence of separate schools for boys and girls was made mandatory for the local authorities even though Italy encountered several problems in putting into practice the initial objective of a free and mandatory primary education system, because of the regional differences inherited from the pre-Unitary polities and the differential supply and quality of primary schools. As of 1861, the gender gap in literacy was not too far from other European countries,8 and declined sharply to reach a very low level by 1901. A similar pattern was followed by the gap in primary enrolment: Table 1 offers a comparative perspective on schooling attainment for boys and girls in late nineteenth century Europe. Although distancing from Spain which shows the most dramatic disadvantage of girls with respect to boys in primary schools, Italy is clearly a laggard together with the Netherlands with 77 girls enrolled every 100 boys. The best performing countries are Great Britain, Belgium, Prussia, and France (all above 90 girls every 100 boys) followed by the Augsburg Kingdom (81 over 100).

7 The so-called “lavori donneschi”, i.e., “women’s work”, continued to be one of the main subjects taught in upper primary schools to girls compared to geometry and technical drawing for boys (Incatasciato, 1978). This distinction was abrogated only after World War II. 8 According to Cipolla (1969) the literacy differential between men and women was 14 percentage points, lower than in Spain (21) but twice that of France (7) and Prussia (5).

7

Table 1. International comparison

Country Year Number of girls over 100 boys in primary school

Italy 1862-63 77 France 1861 90 Great Britain 1858 95 Augsburg Kingdom 1862 81 Prussia 1858 92 Spain 1860 60 Belgium 1860 94 Netherlands 1857 78 Source: 1861 Census.

4. Data and descriptive statistics

In order to test our hypotheses on the determinants of the education gender gap, we compile a database on education in late nineteenth century Italy drawing from various sources. The main source is represented, for literacy, by census data and, for schooling, by data collected and published by the Italian Ministry of Agriculture, Industry and Commerce (MAIC). Our data cover 69 Italian provinces (at 1871 boundaries) in three points in time (1861, 1881, and 1901), spanning the initial forty years of the unified Kingdom of Italy. This allows us to organise the data as a panel in order to account for unobserved heterogeneity. It should be stressed that schooling data at the provincial level are only available until 1901, and that data collection at this level of disaggregation is resumed only after World War II. For our dependent variables, we collect schooling data by gender as enrolment rates for total population at the primary school age, which consists in the population between 6 and 12 (excluded) years of age. On this basis we calculate a measure of the education gender gap as the female to male ratio in enrolment rates in primary school.9 We also disaggregate our gap measure into a lower and an upper primary school level, only the former being compulsory. Beside data by gender, we also include general indicators of schooling defined as the number of pupils enrolled in primary school over the number of children of the corresponding 6-12 school age. Again we disaggregate this measure into lower and upper primary school level. To measure initial gendered human capital accumulation we rely on data on male and female literacy which for most provinces are available for 1861, so that we can construct initial cross province measures of the gender gap in literacy. However, for the provinces belonging to today’s Veneto and for Mantua initial data refer to 1867, for Rome to 1872: this is due to the fact that the former territories were annexed to the new born kingdom only in 1866, the latter in 1871. Literacy rates are calculated as the share of individuals aged 5 or older who are able to read and write and 9 This measure is equivalent to the UN definition of the Gender Parity Index, where 0 corresponds to extreme inequality and 1 to full equality.

8

the gender gap is measured as the female to male ratio (on a 0-1 range) of literacy rates.10 Since literacy data refer to stock information about population in 1861, they effectively embed the outcome of several previous decades of human capital accumulation and educational policies. We complete our dataset with a number of socio-economic variables. We include correlates of economic development which are likely to be related to human capital accumulation and its gendered dimension. We use two measures. The first is the average height of conscripts aged 20, as a proxy of wealth. Height is often employed in the economic history literature because it provides researchers with a measure of the stock of nutritional investment and therefore important indirect information on changes in the well-being of populations (Fogel, Engerman, and Trussel, 1982; A’Hearn and Vecchi, 2011). Data on average height for the three waves are taken from A’Hearn et al. (2009). The second measure is an index of industrialisation which should capture the economic structure of each province. The industrialization index is computed from census data collected by Ciccarelli and Fenoaltea (2012) as the share of value added (excluding construction) over the share of the male population over age 15. In order to control for the legacy of deep rooted differences in political and educational institutions that may have shaped the evolution and persistence of gender disparities, through subsequent policies and beliefs, we construct a set of dummies for the pre-Unification main ruling dynasties, i.e., Savoys, Augsburgs, Pope, and Bourbons. While these dummies capture the potential effect of the geopolitical order prevailing for a relatively short time span, i.e., from 1815 to 1861, we also explore the more distant past by employing a set of dummies for the prevailing political regimes during the fourteenth century: namely, the Signorie, the Communal Republics, the Papal State, the Kingdom of Sicily, and peripheral areas (as in De Blasio and Nuzzo, 2010). To detect the potential role played by the family in the transmission of gender norms, we include a set of dummy variables reflecting the prevailing family type in the Middle Age in each province building on the contribution of Todd (1990), 11 who distinguishes for the Italian peninsula three prevailing family types: incomplete stem, communitarian, and egalitarian nuclear. Todd’s classification is articulated over two leading axes, i.e., liberty vs. authority and equality vs. inequality. The former nexus is related to the rules of residence after marriage: more liberal family structures follow the neo-local rule, i.e., children leave their original family to live on their own after marriage, whereas the patriarchal structure consists in the practice of more than one generations living together and forming vertical and hierarchical relationships. In Todd’s classification the incomplete stem and the communitarian family types belong to the authoritarian model (also referred to as patriarchal) whereas the egalitarian nuclear belongs to the liberal one. According to the latter organising nexus families are distinguished between equal and unequal according to the prevailing inheritance rules among children. Following Del Panta et al. (1996), who specifically focus on Italy, we augment Todd’s classification to distinguish between two 10 Data are taken from censuses, where the ability to read and write is reported through the answers to the following question asked to the household head: “How many people in your family are able to read or read and write?”. 11 In a European perspective, Todd’s classification is employed by Duranton et al. (2009) to study the effect of family structure on various economic outcomes (labor force participation, employment in manufacturing and services, GDP per capita, Gini coefficients) and by Galasso and Profeta (2011) to study pension systems and redistributive policies.

9

different models of egalitarian nuclear family: the first, prevalent in the North West, is associated with delayed marriage for women, the second, prevalent in the South, with early marriage. The marriage age is relevant since it represents a significant indicator of the degree of subordination of women with respect to men and therefore of inequality. As a result, we obtain a set of four dummies. Following the analysis of German cities in Cantoni and Yuchtman (2012), to detect the potential role played by history, we also create two further province-level indicators dating back to the Middle Age. Medieval commerce is a dummy variable that takes value one if the main town in the province was a commercial hub, a fair site, or the seat of a banking institution in late Middle Age, and zero otherwise. Figure 2 illustrates the map of medieval commercial routes in Europe. University is a dummy variable that takes value one if the province saw the onset of a university, or better a studium, between the eleventh and thirteenth century, and zero otherwise.

Figure 2. Medieval commercial routes in Europe

Source: Shepherd (1926).

More detail on the definitions and sources of variables, as well as on methodological issues, can be found in the Appendix.

10

Our dataset generates a set of stylized facts that can suggest preliminary hypotheses and considerations. The descriptive statistics reported in Table A1 reveal that in primary schools the average enrolment rate is 52.7 percent and the average gender gap is 0.83 (where 1 represents perfect equality). However there exists significant variation across provinces for both outcomes, ranging from 9.7 to 117.3 percent for the enrolment rate and from 0.05 to 1.51 for the gender gap. Moreover, the picture changes considerably it one focuses on the upper primary level, with an average enrolment rate of only 4 percent and a much larger gender gap at 0.53. Figure 3 shows the time evolution of enrolment rates for boys and girls across four macro-regions, at the primary level as an aggregate. In the North West – corresponding in 1861 to the Kingdom of Sardinia and Lombardy – schooling rates are uniformly larger for boys and girls, if compared to other areas, while the gender gap is smaller. At the opposite we find the South, ruled by the Bourbons. Despite evidence of convergence over time, we also find that at the end of the period the initial differences are far from eradicated.

Figure 3. Enrolment rates by gender at the primary level, by macro-region, 1861-1901

Figure 4 illustrate the same dynamics for the upper level of primary schools, where rates are computed by dividing the number of pupils enrolled by the number of children of the relevant schooling age. Enrolment rates at the upper level are uniformly much lower, presumably because this level was not mandatory. The gender gap in enrolment is also uniformly larger. North Western provinces lead again, with higher enrolments and lower gap, followed by the Centre and the North East, with the South far behind.

85.1

75.6

90.8

85.4

92 88.6

54.3

20.3

71.4

60.9

79.2

71.6

35.1

29.2

48.6

43

61.2

55.5

25.3

17.5

38.5 33.2

45.1 40.4

0

20

40

60

80

100

North West North East Center South1861 1881 1901 1861 1881 1901 1861 1881 1901 1861 1881 1901

Male Schooling Rate Female Schooling Rate

11

Figure 4. Enrolment rates by gender at the upper primary level, by macro-region, 1861-1901

In Figure 5 we look at regional differences in literacy rates at the beginning of the period. While the average literacy rate across the provinces is 23 percent, we find a striking divide between North Western territories, at 44 percent, and the rest of Italy, with the South at only 12 percent. Average male literacy is at about 30 percent against 16 percent for women, and again territorial differences are pervasive, with 52 percent of men and 36 percent of women being able to read and write in the North West against only 18.3 percent and 6 percent, respectively, in the South.

16.9

11.1

17.8

13

36.5

30.5

17.6

4.4

15.5

10.2

23.6

16

9

5.2

13.9

10

25.8

19.1

4.7

.7

10.7

5.7

15.8

10.5

0

10

20

30

40

North West North East Center South1861 1881 1901 1861 1881 1901 1861 1881 1901 1861 1881 1901

Male Upper Schooling Rate Female Upper Schooling Rate

12

Figure 5. Initial literacy rates, by macro-region, 1861

In Figure 6 we plot the education gender gap at the upper primary level, as defined in Table A1, against the initial literacy gender gap for each year in the sample, i.e., 1861, 1881, and 1901. For each year we find a positive association between the two variables even though the slope gets flatter and the intercept increases through time. In other words, the graphs confirm evidence of convergence between girls and boys but also of persistence of the initial gender differences.

44

52

36

23.7

31.5

15.9

22.9

28.9

17

12.1

18.3

6

0

10

20

30

40

50

North West North East Center South

Literacy 1861 Male Literacy in 1861 Female Literacy in 1861

13

Figure 6. The education gender gap at the upper primary level and the 1861 literacy gender gap, 1861-1901

0

.5

1

1.

0

.5

1

1.5

0

.5

1

1.5

.2 .4 .6 .8 1

1861

1881

1901

Female to male ratio upper schooling rate

Fitted values

Female to male ratio Literacy 1861

14

In Figure 7 we illustrate the bivariate relationship between our focal variable, medieval commerce, and the education gender gap at the upper primary level. The figure highlights the fact that provinces that hosted a commercial hub in the Middle Age are associated with a more advantageous condition for women throughout the period under examination and in each region (with the only exception of the North east in the initial year of the sample).

Figure 7. The gender gap at the upper primary level and medieval commerce, by macro-region, 1861-1901

.48

.68 .67

.75.79

.84

.27.25

.62

.69.66

.69

.24

.58

.68

.78

.71.75

.07

.23

.46

.59.62

.67

0.2

.4.6

.8ge

nder

gap

in u

pper

prim

ary

scho

ols

northwest northeast center south1861 1881 1901 1861 1881 1901 1861 1881 1901 1861 1881 1901

non commercial hubs medieval commercial hubs

Equivalently, Table 2 reports descriptive statistics for our measures of gender equality in human capital accumulation distinguishing between provinces that were a commercial centre and provinces that were not. Medieval commercial hubs exhibit higher mean scores at all levels of schooling, but their superiority is especially clear at the upper primary level, whit a score of 0.62 against 0.47 for the rest of the sample. These initial observations put forth strong evidence of a relevant effect of the past legacy of trade on subsequent cultural norms regarding gender roles.

15

5. Regression analysis In order to assess the determinants of education inequalities across gender at the primary school level in Italy, over the 1861-1901 period immediately following the Unification, we regress the education gender gap on a number of covariates. Table 3 presents OLS regressions for three alternative dependent variables. We first examine the female to male ratio in enrollment rates for the primary level as a whole (Column 1). Enrollment rates are measured as the number of pupils of each gender over the school age (i.e., 6-12 years of age) population of the corresponding gender. Then we turn to the lower and upper primary levels taken separately (Columns 2 and 3, respectively). Because of lack of data at the provincial level, the denominators of the enrollment rates at the upper and lower level are always the corresponding 6-12 population figures. Since the lower primary level was compulsory, it is at the upper level that differences in the treatment of boys vs. girls emerge more clearly. In Column 1 we find that the controls we insert for male mean height, as a proxy for pre-existing wealth, and industrialization, are both insignificant, while the coefficient of the initial gender gap in literacy is highly significant and positive. The latter variable has to be interpreted as a measure of the initial endowment in human capital as well of gender role norms and beliefs. Its significant effect signals the presence of persistence, despite the fact that the time dummies also indicate a simultaneous process of reduction of the gap throughout the period. Turning to the set of dummies that describe the political geography of the country prior to Unification, where the Savoy Kingdom is the reference category, we find that the gap is significantly larger in the territories ruled by the Pope and the Bourbons, even though the size and significance of the coefficients is larger for the latter. Next we find that the dummies capturing family structure, with egalitarian nuclear family with delayed marriage as the reference category, indicate a smaller gap for the egalitarian nuclear with early marriage and a larger one for the incomplete stem type, while the communitarian type is not significantly different. Finally, the variable capturing medieval commerce is associated with a positive but insignificant coefficient (the coefficient is significant at 10% in a specification omitting the initial condition for literacy). In Columns 3 and 4 we separate the lower and upper primary

Table 2. The gender gap and medieval commerce, summary statistics, 1861-1901 Female to male ratio

schooling rate Female to male ratio

lower schooling rate Female to male ratio

upper schooling rate

Whole sample

Medieval commercial

hubs Others

Whole sample

Medieval commercial

hubs Others

Whole sample

Medieval commercial

hubs Others

Obs 207 102 105 207 102 105 207 102 105

Mean 0.833455 0.877647 0.790526 0.854258 0.89787 0.811891 0.550165 0.626824 0.475695

Std. Dev. 0.206254 0.209335 0.194758 0.216332 0.212345 0.212677 0.269699 0.253488 0.265162

Min 0.058806 0.133151 0.058806 0.045682 0.081783 0.045682 0 0 0

Max 1.515913 1.515913 1.135 1.339314 1.339314 1.162496 1.326245 1.326245 0.898016

T-test 0.087*** 0.0859 *** 0.15*** ( 0.028) ( 0.0295) ( 0.036)

T-test is the two-sample mean comparison test (Ha=diff~=0). P-values are reported in parentheses. *, **, and *** denote significance at 10%, 5%, and1% levels.

16

school levels. Previous results still hold for most regressors, even though at the lower level Austrian territories are also displaying a larger gap than the Savoy Kingdom, while at the upper level the Papal territories are no longer associated with a larger gap, possibly because of the influence of the relatively well developed system of religious schools in Rome. At the upper level, the egalitarian nuclear family type with early marriage loses significance if compared with Column 1. Medieval commerce gains significance but only at the upper level, indicating that provinces that were more closely involved with trade in the period running from the thirteenth to the fourteenth centuries exhibit a more favorable position for women in terms of education.

Table 3. The determinants of the education gender gap, Italy, 1861-1901: OLS (1) (2) (3)

Estimation Method: Pooled OLS

Female to male ratio schooling

rate

Female to male ratio lower

schooling rate

Female to male ratio upper

schooling rate Height -0.0100 -0.0122 0.00604

(0.00907) (0.00934) (0.0103) Industrialization -0.0153 -0.00191 -0.0450

(0.0364) (0.0381) (0.0454) Female to male ratio literacy 1861 0.498*** 0.518*** 0.438*** (0.0997) (0.107) (0.124) Year 1881 0.189*** 0.187*** 0.309***

(0.0312) (0.0336) (0.0333) Year 1901 0.223*** 0.236*** 0.376***

(0.0340) (0.0354) (0.0355) Augsburgs -0.0412 -0.0557* 0.0199

(0.0289) (0.0306) (0.0484) Pope -0.0757* -0.105** -0.0637

(0.0429) (0.0494) (0.0659) Bourbons -0.163** -0.114* -0.199**

(0.0820) (0.0590) (0.0840) Egal. nucl. family early marriage 0.196** 0.168*** 0.102

(0.0815) (0.0508) (0.0717) Incomplete stem family -0.119** -0.112* -0.133**

(0.0564) (0.0597) (0.0546) Communitarian family 0.0216 0.0358 -0.0503

(0.0378) (0.0393) (0.0548) Medieval commerce 0.0200 0.0230 0.0566**

(0.0241) (0.0260) (0.0257) Constant 2.119 2.462 -0.797

(1.447) (1.490) (1.636)

Observations 207 207 207 R-squared 0.48 0.44 0.65 Adj. R-squared 0.44 0.41 0.63 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

17

In Table 4 we repeat the same exercise with OLS regressions with regional fixed effects reflecting the macro-regions, where the North West is the reference category. Results from this specification, where height and industrialization are dropped, confirm the results from Table 3. In particular, the effect of medieval commerce on female educational attainment relative to that of males is again significantly positive at the upper primary level.

Table 4. The determinants of the education gender gap, Italy, 1861-1901: OLS with fixed effects (1) (2) (3)

Estimation Method: LSDV

Female to male ratio

schooling rate

Female to male ratio lower

schooling rate

Female to male ratio upper

schooling rate Year 1881 0.179*** 0.174*** 0.315***

(0.0285) (0.0309) (0.0307) Year 1901 0.208*** 0.216*** 0.387***

(0.0290) (0.0299) (0.0301) Female to male ratio literacy 1861 0.514*** 0.564*** 0.383***

(0.118) (0.133) (0.146) Augsburgs -0.0309 -0.0428 0.00896

(0.0281) (0.0296) (0.0501) Pope -0.0549 -0.0798* -0.0667

(0.0424) (0.0462) (0.0668) Bourbons -0.258*** -0.210*** -0.213**

(0.0777) (0.0626) (0.0930) Egal. nucl. fam. early marriage 0.265*** 0.222*** 0.157*

(0.0836) (0.0639) (0.0818) Incomplete stem family -0.0170 -0.00961 -0.0445

(0.0513) (0.0574) (0.0586) Communitarian family 0.100* 0.102** 0.0196

(0.0510) (0.0510) (0.0693) Medieval commerce 0.0189 0.0223 0.0577**

(0.0234) (0.0254) (0.0252) North East -0.131*** -0.129*** -0.0934

(0.0405) (0.0425) (0.0579) Center -0.0661 -0.0492 -0.0490

(0.0471) (0.0489) (0.0619) South 0.0699 0.101* -0.0749

(0.0461) (0.0519) (0.0731) Constant 0.452*** 0.435*** 0.185**

(0.0718) (0.0821) (0.0925)

Observations 207 207 207 R-squared 0.50 0.47 0.66 Adj. R-squared 0.46 0.43 0.63 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

To check the robustness of our results, we run several other variants of the regressions in Table 3. We modify the dependent variables, using simply the ratio of female to male enrollment, rather than the corresponding rates, and adding an index of masculinity of the school age population (i.e.,

18

female over male population of age 6-12) as a further control. Our results are unchanged. We also estimate the same regressions over a sub-sample excluding the provinces located in Veneto and Mantua, which were part of the Austrian territories, as well as Rome. We exclude these provinces because they show missing values for a few variables in the initial year a few years.12 Results are again similar. Moreover, our results are robust to the use of a different source for initial literacy. Namely, rather than census data we use data from wedding registers. We also employ dummies for alternative geographical entities, i.e., the regions defined by the 1884 Inchiesta Jacini. However, these dummies largely overlap with those previously employed for the political geography of the country prior to Unification. Finally, we obtain similar results, concerning the role of medieval commerce, when we replace our geopolitical dummies with dummies that reflect more distant, fourteenth century political regimes, i.e., communal republics, Papal territories, the Kingdom of Sicily and other peripheral areas. For brevity, we do not report these additional regressions here. A general conclusion we can drive from Tables 3 and 4 is that, over the first decades after Unification, after controlling for a number of covariates, the evolution of gender inequalities in education is strongly influenced by initial literacy. Moreover, geopolitical and sociological differences also matter. Finally, the influence of the medieval patterns of trade is still perceived after controlling for the above factors at the upper primary, non-compulsory level of schooling. The results we obtained so far may be biased by omitted variables and measurement errors. In particular, it can be argued that both the dependent variables and medieval commerce may be driven by a common third factor. Thus, in order to obtain consistent estimates, we now proceed with 2SLS estimation where our focal variable, medieval commerce, is instrumented with a dummy that captures the presence of a university in even earlier times. For a sample of medieval German cities Cantoni and Yuchtman (2012) show that medieval universities played a causal role in promoting the expansion of trade that led to Europe's Commercial Revolution. Universities provided juridical training which facilitated the establishment of the legal and administrative institutions which were required for growth. They also favored the creation of networks of highly mobile individuals whose communication was facilitated by their ability to speak Latin as a common language. The identifying assumption here is that the creation of a university, or studium, in the early Middle Ages is likely to be the result of independent initiative of groups of people (students and professors) who decided to organize themselves for cultural purposes (Ascheri, 1994)13 and therefore unlikely to have a direct impact on outcomes in the late nineteenth century period. Results are reported in Table 5. Column 1 reports the first stage, common to the second stages presented in Columns 2-4, one for each dependent variable. When medieval commerce is regressed on university, as well as the other regressors, the coefficient of the latter is positive and highly significant. Together with the partial F statistics, reported for each second stage in Columns 2-4, this confirms the relevance of the instrument we select. In Columns 2-4 we present the second stages for the primary level as a whole, the lower primary and the upper primary levels, respectively. The effect of medieval commerce is now triple and significant at 5% at the upper primary level, for which the Hausman test does reject the null hypothesis of exogeneity. Medieval commerce remains insignificant at the lower, 12 See the Appendix for further details. 13 See also Percoco (2010) on the influence of universities on development in Italy.

19

compulsory level, and as a consequence also at the aggregate level, but here the Hausman test does not reject the null so that the OLS estimates are to be preferred.

Table 5. The determinants of the education gender gap, Italy, 1861-1901: 2SLS (1) (2) (3) (4)

Estimation Method: 2SLS

Medieval Commerce

Female to male ratio schooling

rate

Female to male ratio lower schooling

rate

Female to male ratio upper schooling

rate University 0.454***

(0.0776) Medieval commerce 0.0535 0.0233 0.152**

(0.0606) (0.0687) (0.0605) Height -0.0278 -0.0110 -0.0122 0.00332

(0.0311) (0.00903) (0.00902) (0.0104) Industrialization 0.115 -0.0213 -0.00196 -0.0620

(0.110) (0.0358) (0.0386) (0.0447) Year 1881 0.0295 0.190*** 0.187*** 0.311***

(0.0791) (0.0301) (0.0326) (0.0328) Year 1901 0.0499 0.225*** 0.236*** 0.380***

(0.0913) (0.0330) (0.0340) (0.0361) Female to male ratio literacy 1861 1.037*** 0.463*** 0.518*** 0.340***

(0.245) (0.0976) (0.114) (0.128) Augsburgs 0.0622 -0.0374 -0.0557* 0.0309

(0.119) (0.0268) (0.0295) (0.0478) Pope 0.0641 -0.0691* -0.105** -0.0450

(0.160) (0.0417) (0.0504) (0.0656) Bourbons 0.175 -0.156** -0.114* -0.179**

(0.171) (0.0791) (0.0603) (0.0818) Egal. nucl. fam. early marriage -0.106 0.185** 0.168*** 0.0704

(0.135) (0.0775) (0.0533) (0.0695) Incomplete stem family -0.133 -0.118** -0.112* -0.132**

(0.144) (0.0543) (0.0578) (0.0527) Communitarian family -0.309** 0.0219 0.0358 -0.0494

(0.124) (0.0360) (0.0381) (0.0529) Constant 4.272 2.279 2.463* -0.342

(4.990) (1.440) (1.439) (1.652)

Observations 207 207 207 207 R-squared 0.33 0.47 0.44 0.63 Adj. R-squared 0.29 0.44 0.41 0.60 F test (First stage) 34.25 34.25 34.25 Endogeneity (p-values) 0.5564 0.9968 0.0597 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

20

Despite the fact that our results so far point to a powerful influence of medieval commerce on relative female attainment, up to the beginning of the twentieth century, one may wonder whether medieval commerce might have had an influence not only on the gender gap, but more generally on overall schooling. In order to verify whether indeed medieval commerce affected male and female enrollment differently, we now turn to the determinants of schooling. Results are presented in Table 6 for three separate dependent variables: the primary schooling rate is measured as the number of pupils enrolled in primary schools over the population of age 6-12; the lower and upper primary schooling rates are the number of pupils enrolled in lower and upper primary schools, in both cases over the population of age 6-12 since disaggregated data by age sub-classes are not available at the provincial level. In Column 1 we find that height and industrialization are now highly significant determinants of schooling at the aggregate level: higher height, as a proxy for wealth, is positively associated with schooling, while at the margin the effect of industrialization is negative. The year dummies and the initial literacy level in 1861 (measured for the population over age 5) are also positively significant. The political geography of the country prior to Unification is hardly significant, while we find a large explanatory power for family structure, with the reference category, i.e., the nuclear family with delayed marriage, being associated with more schooling. Medieval commerce now displays a negative but insignificant coefficient. When we turn to Columns 2 and 3, we observe some differences in the effect of some of the regressors. For instance, the negative impact of industrialization is no longer present at the upper level. However, medieval commerce remains always insignificant, suggesting that the influence it exerts on female vs. male education up to 1901 has something to do with its peculiar gendered effect.14

14 We obtain similar results for OLS estimates with regional fixed effects and for 2SLS estimates where medieval commerce is instrumented with university, even though exogeneity is not rejected at either level of schooling.

21

Table 6. The determinants of schooling, Italy, 1861-1901: OLS

(1) (2) (3)

Estimation Method: Pooled OLS Schooling

rate Lower

Schooling rate Upper

Schooling rate Height 1.519** 1.108* 0.283*

(0.658) (0.656) (0.146) Industrialization -8.780*** -10.08*** 1.218

(3.140) (3.154) (0.798) Year 1881 14.46*** 13.42*** 0.923***

(1.716) (1.783) (0.328) Year 1901 20.89*** 17.29*** 3.858***

(1.928) (1.919) (0.419) Literacy 1861 1.092*** 1.019*** 0.0830*** (0.122) (0.126) (0.0297) Augsburgs 2.777 5.133 -2.265**

(3.573) (3.780) (1.128) Pope 6.755* 7.282* -1.039

(4.056) (4.174) (1.061) Bourbons 0.657 3.435 -3.769***

(4.315) (4.314) (1.410) Egal. nucl. fam. early marriage -17.48*** -19.26*** 2.781*

(4.067) (4.086) (1.501) Incomplete stem family -6.586* -8.085* 1.500

(3.884) (4.117) (1.118) Communitarian family -19.18*** -19.41*** 0.413

(3.548) (3.633) (0.894) Medieval commerce -1.107 -1.184 -0.101

(1.431) (1.504) (0.333) Constant -214.0** -148.1 -45.32*

(104.2) (103.6) (23.11)

Observations 207 207 207 R-squared 0.86 0.82 0.56 Adj. R-squared 0.85 0.81 0.54 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. To further explore the gendered effect of medieval commerce on education, we also run separate regressions where the dependent variables are female and male primary enrollment: in Table 7 commerce is not significant for either dependent variable, even though the absolute size of its negative coefficient is indeed twice as large for male schooling if compared to female. The same is true in unreported regressions where the dependent variable is disaggregated between the lower and upper primary level.

22

Table 7. The determinants of schooling by gender, Italy, 1861-1901: OLS

(1) (2)

Estimation Method: Pooled OLS Male

schooling rate Female

schooling rate Height 1.710** 1.633**

(0.657) (0.737) Industrialization -7.765** -9.827***

(2.989) (3.579) Year 1881 10.82*** 18.15***

(1.642) (2.221) Year 1901 16.52*** 25.15***

(1.838) (2.475) Male literacy 1861 0.953*** (0.0980) Female literacy 1861 1.174*** (0.149) Augsburgs 4.686 0.477

(3.566) (3.813) Pope 9.828** 2.976

(4.074) (4.387) Bourbons 4.699 -3.327

(4.524) (4.977) Egal. nucl. fam. early marriage -22.95*** -13.52***

(4.162) (4.737) Incomplete stem family -3.659 -11.27**

(3.369) (5.188) Communitarian family -20.91*** -18.85***

(3.318) (4.017) Medieval commerce -1.565 -0.794

(1.408) (1.768) Constant -243.0** -230.6*

(104.1) (117.3)

Observations 207 207 R-squared 0.86 0.81 Adj. R-squared 0.86 0.79 Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

To sum up, we find that medieval commerce still affects the education gender gap throughout the entire 1861-1901 period, with an especially strong influence at the non-compulsory, upper primary level. Our findings are confirmed after a number of robustness checks, including 2SLS estimation where medieval commerce is instrumented with the presence of an ancient university and a comparison with its effect on the schooling rate, which turns out to be insignificant, pointing to its peculiar influence not so much on overall human capital accumulation but rather on its gendered differential speed.

23

6. Conclusion This work has attempted at exploring the empirical determinants of the education gender gap in Italy and its evolution in historical perspective and represents a crucial advancement in the literature about long term persistence of deeply rooted factors of economic and social development. This has been made possible through a newly assembled dataset that covers a sample of Italian provinces in the 1861-1901 period, i.e., the initial forty years after the country’s Unification. In our search for the determinants of the education gender gap, after controlling for economic, geopolitical and social factors, such as pre-existing ruling dynasties and prevailing family-types, we find that the cross-sectional variation of the gap within Italian provinces is influenced by the medieval pattern of commerce, along the routes that connected Italian cities among themselves and with the rest of Europe. We suggest that having a past as medieval commercial hub might have created the favourable preconditions for the transformation of mentality about the role of women in society which has ended up in a more egalitarian set of beliefs transmitted through generations.

24

References A’Hearn, B., Auria, C., Vecchi, G., 2011. Istruzione. In Vecchi, G. (ed.), In Ricchezza e in Povertà: Il Benessere degli Italiani dall'Unità a Oggi. Il Mulino, Bologna. 159-206. A'Hearn, B., Peracchi, F., Vecchi, G., 2009. Height and the Normal Distribution: Evidence from Italian Military Data. Demography 46, 1-25. Alesina, A.F., Giuliano, P., Nunn, N., 2011. On the Origins of Gender Roles: Women and the Plough. NBER Working Papers No. 17098. Ascheri, M., 1994. Istituzioni Medievali. Il Mulino, Bologna. Becker, S.O., Wößmann, L., 2008. Luther and the Girls: Religious Denomination and the Female Education Gap in Nineteenth-century Prussia. Scandinavian Journal of Economics 110, 777-805. Benavot, A., Riddle, P., 1988. The Expansion of Primary Education, 1870-1940: Trends and Issues. Sociology of Education 61, 191-210. Bertocchi, G., 2011. The Enfranchisement of Women and the Welfare State. European Economic Review 55, 535-553. Bertocchi, G., Brunetti, M., Torricelli, C., 2012. Is It Money or Brains? The Determinants of Intra-Family Decision Power. CEPR Discussion Paper No. 9017. Bertocchi, G., Spagat, M., 1997. Il Ruolo dei Licei e delle Scuole Tecnico-Professionali tra Progresso Tecnologico, Conflitto Sociale e Sviluppo Economico. In Rossi, N. [ed.], L'Istruzione in Italia: Solo un Pezzo di Carta? Il Mulino, Bologna. 421-459. Bertola, G., Sestito, P., 2011. A Comparative Perspective on Italy’s Human Capital Accumulation. Bank of Italy Economic History Working Papers No. 6. Boccaccio, G., 1972. The Decameron. McWilliam, G. H. (transl.). Penguin Books, London. Boserup, E., 1970. Woman’s Role in Economic Development. George Allen and Unwin Ltd., London. Bozzano, M., 2012. Assessing Gender Inequality among Italian Regions: The Italian Gender Gap Index. University of Pavia, Department of Economics and Quantitative Methods Quaderni di Dipartimento No. 17. Briggs, C.F., 2000. Literacy, Reading, and Writing in the Medieval West. Journal of Medieval History 26, 4397-4420. Cantoni, D., Yuchtman,N., 2012. Medieval Universities, Legal Institutions, and the Commercial Revolution. NBER Working Papers No. 17979. Ciccarelli, C., Fenoaltea, S., 2012. Through the Magnifying Glass: Provincial Aspects of Industrial Growth in Post-Unification Italy. Economic History Review, forthcoming.

25

Cipolla, C.M., 1969. Literacy and Development in the West. Penguin Books, Harmondsworth. De Blasio, G., Nuzzo G., 2010. Historical Traditions of Civicness and Local Economic Development. Journal of Regional Science 50, 833-857. Del Panta, L., Livi Bacci, M., Pinto, G., 1996. La Popolazione Italiana dal Medioevo a Oggi. Laterza, Roma. De Mauro, T., 1963. Storia Linguistica dell’Italia Unita. Laterza, Bari. Dincecco, M., Federico, G., Vindigni, A., 2011. Warfare, Taxation, and Political Change: Evidence from the Italian Risorgimento. Journal of Economic History 71, 887-914. Buonazia, G. 1870. Documenti sulla Istruzione Elementare del Regno d’Italia, Parte II. Camera dei Deputati, Firenze. Doepke M, Tertilt M. 2009. Women’s Liberation: What’s In It for Men? Quarterly Journal of Economics 124, 1541-1591. Duby, C., Perrot, M., 1998. L’Histoire des Femmes en Occident. Flammarion, Paris. Duranton, G., Rodriguez-Pose, A., Sandall, R., 2009. Family Types and the Persistence of Regional Disparities in Europe. Economic Geography 85, 23-47. Fernandez, R., 2012. Cultural Change as Learning: The Evolution of Female Labor Force Participation over a Century. American Economic Review, forthcoming.

Fernandez, R., Wong, J.C., 2011. The Disappearing Gender Gap: The Impact of Divorce, Wages, and Preferences on Education Choices and Women's Work. Mimeo, New York University. Fogel, R.W., Engerman, S.L., Trussell, J., 1982. Exploring the Uses of Data on Height: The Analysis of Long-Term Trends in Nutrition, Labor Welfare, and Labor Productivity. Social Science History 6, 401-421. Fogli, A., Veldkamp, L., 2011. Nature or Nurture? Learning and the Geography of Female Labor Force Participation. Econometrica 79, 1103-1138. Galasso, V., Profeta, P., 2011. When the State Mirrors the Family: The Design of Pension Systems. CEPR Discussion Papers No. 8723. Galor, O., Weil, D.N., 1996. The Gender Gap, Fertility, and Growth. American Economic Review 86, 274-387. Gibbins, H. de B., 1891. The History of Commerce in Europe. Macmillan, London. Goldin, C., 2006. The Quiet Revolution That Transformed Women's Employment, Education, and Family. American Economic Review Papers and Proceedings 96, 1-21.

26

Goldin, C., 1998. America’s Graduation from High School: The Evolution and Spread of Secondary Schooling in the Twentieth Century. Journal of Economic History 58, 345-374. Goldin, C., 1991. The Role of World War II in the Rise of Women’s Employment. American Economic Review 81, 741-756. Goldin, C, Katz L.F., Kuziemko, I. 2006. The Homecoming of American College Women: The Reversal of the Gender Gap in College. Journal of Economic Perspectives 20, 133-156. Hausmann, R.D., Tyson, L., Zahidi, S., 2007. The Global Gender Gap Report, World Economic Forum, Geneva. Hoftijzer, P., 2001. Metropolis of Print: the Amsterdam Book Trade in the Seventeenth Century. In O'Brien, P. (ed.), Urban Achievement in Early Modern Europe: Golden Ages in Antwerp, Amsterdam, and London. Cambridge University Press, New York. Incatasciato, B., 1978. Leggere, Scrivere, Far di Conto. Per una Storia della Didattica nella Scuola Elementare. In Tomasi, T., Genovesi, G., Tancredi Torelli, M.P., Incatasciato, B., Ulivieri, S., Catarsi, E. (eds.), L’Istruzione di Base in Italia (1859-1977). Vallecchi, Firenze. MAIC, 1864. Popolazione: Censimento Generale (31 Dicembre 1861). Tipografia Letteraria, Torino. MAIC, 1865. Statistica del Regno d'Italia. Istruzione Pubblica e Privata. Anno scolastico 1862-63. Enrico Dalmazzo Tipografo, Torino. MAIC, 1868. Statistica del Regno d'Italia, Popolazione, Movimento dello Stato Civile del 1867. Tipografia Tofani, Firenze. MAIC, 1872. Censimento della Popolazione del Regno d’Italia al 31 dicembre 1871.Stamperia Reale, Roma. MAIC, 1883. Censimento della Popolazione del Regno d’Italia al 31 dicembre 1881. Tipografia Bodoniana, Roma. MAIC, 1884. Statistica della Istruzione Primaria e Normale per l'Anno Scolastico 1881-82. Tipografia Nazionale, Roma. MAIC, 1902. Censimento della Popolazione del Regno d’Italia al 10 Febbraio 1901. Tipografia Nazionale Bertero, Roma. MAIC, 1906. Statistica della Istruzione Primaria e Normale per l'Anno Scolastico 1901-02. Tipografia Nazionale Bertero, Roma. Pécout, G., 1999. Il Lungo Risorgimento: La Nascita dell'Italia Contemporanea (1770-1922). Edizioni Bruno Mondadori, Milano. Percoco, M., 2010. Path Dependence, Institutions, and the Density of Economic Activities: Evidence from Italian Cities. FEEM Nota di Lavoro No. 110.

27

Ricci, A., 2011. Mercanti e Lingua. In Enciclopedia dell’Italiano, Treccani.it. Sacchi, G., 1858. Le Scuole Reali Inferiori ed Elementari nelle Provincie Venete. Annali Universali di Statistica. Economia pubblica, Legislazione, Storia, Viaggi e Commercio 133. Sapori, A., 1972. La Mercatura Medievale. Sansoni, Firenze. Schultz, T.P., 2002. Why Governments Should Invest More to Educate Girls. World Development 30, 207-225. Serristori, L., 1842. Statistica dell’Italia. Stamperia Granducale, Firenze. Shepherd, W.R., 1926. Historical Atlas, 5th edition. Holt, New York. Spufford, M., 1995. Literacy, Trade and Religion in the Commercial Centres of Europe. In Davids, K., Licassen, J., (eds.), A Miracle Mirrored: The Dutch Republic in European Perspective. Cambridge University Press, Cambridge. 229-283. Todd, E., 1990. L’Invention de l’Europe. Seuil, Paris. Vigo, G., 1971. Istruzione e Sviluppo Economico in Italia nel Secolo XIX, Archivio Economico dell’Unificazione Italiana, Serie II, vol. XVIII. ILTE, Torino.

28

APPENDIX

This appendix describes the data underlying our study in more detail. The dataset contains information from 69 Italian provinces at 1871 boundaries. We collected province level data for three points in time (1861, 1881,1901) spanning the first forty years after Italian Unification. Table A1 contains variable definitions and sources. Information on the construction of the dataset follows next. Table A2 presents summary statistics.

Table A1. Variable definitions and sources

Variable Definition Construction Main Source

Schooling rate Gross Primary School Enrolment Rate: pupils who are enrolled in schools over children of the corresponding school age (6-12).

Pupils in primary schools/population of school age (6-12) *100

MAIC (various years) See details in Appendix 3.

Schooling lower Gross Lower Primary School Enrolment Rate over children of the school age (6-12).

Pupils in lower primary schools/ population of school age (6-12) *100


Schooling higher Gross Upper Primary School Enrolment Rate over children of the school age (6-12).

Pupils in upper primary schools / population of school age (6-12) *100


Male schooling rate

Male Gross Primary School Enrolment Rate

Male pupils/ male population of school age (6-12) *100


Female schooling rate

Female Gross Primary School Enrolment Rate

Female pupils/ female population of school age (6-12) *100


Female to male schooling rate

Female Gross Primary School Enrolment Rate over Male Gross Primary School Enrolment Rate (range 0-1)

Female pupils/ female population of school age (6-12)/ Male pupils/ male population of school age (6-12) = female pupils/male pupils * male population of school age (6-12) / female population of school age (6-12)


Female to male ratio upper schooling rate

Gender Parity Index (GPI) in upper primary schools (range 0-1)

Female upper schooling rate/male upper schooling rate weighted by masculinity index calculated as male population 6-12 / female population 6-12


Female to male ratio lower schooling rate

Gender Parity Index (GPI) in lower primary schools (range 0-1)

Female lower schooling rate/male lower schooling rate weighted by masculinity index calculated as male population 6-12 / female population 6-12


Literacy 1861 Percentage of people able to read and write over total

MAIC (1864)

29

population aged 5 or more in 1861

Male Literacy 1861

Percentage of literate males over male population aged 5 or more in 1861

MAIC (1864)

Female Literacy 1861

Percentage of literate females over female population aged 5 or more in 1861

MAIC (1864)

Female to male ratio Literacy 1861

Gender gap in literacy rates in 1861 (range 0-1)

MAIC (1864)

Medieval Commerce

dummy: 1 if main town lies on medieval commerce routes or was a commercial hub, fair or bank seat in late Middle Ages

Shepherd (1926)

University dummy: 1 if the main town of the province is the seat of ancient university or Studium

Height Average height at age 20 of military conscripts

A’Hearn and Vecchi (2011)

Industrialization provincial index of relative industrialization calculated on the basis of census data as the share of value added, excluding construction, over share of the male population over age fifteen.

Ciccarelli and Fenoaltea (2012)

Masculinity (school-age children)

female population of school-age(6-12)/ male population of school-age(6-12)

Census (various years)

Egalitarian nuclear family with early marriage

Dummy: 1 if the egalitarian nuclear family type with early marriage prevailed in the Middle Ages, 0 otherwise

Egalitarian nuclear family with delayed marriage

Dummy: 1 if the egalitarian nuclear family type with late marriage prevailed in the Middle Ages, 0 otherwise

Incomplete stem family

Dummy: 1 if the incomplete stem family type prevailed in the Middle Ages, 0 otherwise

Communitarian family

Dummy: 1 if the communitarian family type prevailed in the Middle Ages, 0 otherwise

30

Methodological notes

Enrolment rates: We collected data for 1861, 1881, and 1901. Due to historical circumstances, we encountered a few problems because of lack of data in particular referring to the first wave. This arises from the fact that the Italian Unification process had different timing and therefore official statistical annuals could not take into account those provinces which entered the Italian Kingdom only in following periods. As a result data on enrolment rates are constructed from various sources and mainly from Statistica del Regno d'Italia, Istruzione pubblica e privata, a statistical series compiled and published by the MAIC, Ministry of Agriculture, Industry and Commerce from 1865 to 1906, and from Vigo (1971). For most provinces enrolment data referred to 1861(the Unification year) refer to 1862-63. For provinces belonging to today’s Veneto data on enrolments for the first wave are from Documenti sulla istruzione elementare del Regno d’Italia, Parte I (Buonazia, 1868) and Vigo (1971) for primary schools and Sacchi (1858) for upper primary schools and refer to 1856. For Mantua data are from Vigo (1971) for primary schooling and refer to 1851 and for upper schooling are from MAIC (1872). For Rome (and Comarca) data refer to 1858 and are taken from Vigo (1971) and for upper primary schooling are from MAIC (1872). To determine enrolment rates, we assumed 6-12 (excluded) to be the right primary school age and therefore we obtain an unadjusted measure of enrolment (Benavot and Riddle, 1988) because we keep constant the reference school age without taking into account the effective length of the schooling cycles. In some cases gross enrolment rates are greater than 100% and this is due to under-aged and/or over-aged enrolment with respect to the considered school age (and this is even more compelling in the first moves toward mass education where late entrance ought have a high incidence). In calculating lower and upper schooling rates we employ the population 6-12 (excluded) because we have no data for the specific grade ages. Even though we acknowledge that this induces a downward bias in the measure of enrolment rates, we believe that the information is still valuable. Literacy rates: Data on literacy are calculated on the basis of 1861 census that reports the number of people able to read or read and write aged 5 or more. Data for Veneto and Mantua are taken from MAIC (1868) and come from wedding registers as the ability to sign one's name in full on a marriage certificate. Data for Rome come from MAIC(1906) and refer to 1871. Industrialization: Data on industrialization at provincial level are taken from Ciccarelli and Fenoaltea (2012, Table 2). The variable is a provincial index of relative industrialization calculated on the basis of census data as the share of value added, excluding construction, over share of the male population over age 15. Data for the first wave (1861) are those of 1871. Height: Due to the historical nature of the study, data about wealth of ancient Italian provinces are not available. For this reason we employ height as a proxy for wealth. In the economic history literature it is common to make use of anthropometric measures because they provide important indirect information on changes in the well-being of the population. Data on height refer to mean height of military conscripts aged 20 at provincial level and are taken from A’Hearn and Vecchi (2011, Table S3). For an exhaustive presentation of the use of this kind of anthropometric data as a plausible proxy for well-being and living standards see A’Hearn and Vecchi (2011, ch. 2). For a discussion about the methodological approach employed to estimate average height see A'Hearn et al. (2009).

31

Family types: We build on the classification proposed by Todd (1990) and we code provinces according to the prevailing family type in the Middle Ages that we adapt to the Italian specific regional differentiation. Therefore we distinguish four types of families: incomplete stem, communitarian, egalitarian nuclear with delayed marriage, and egalitarian nuclear family with early marriage (Del Panta et al., 1996). Ruling dynasties: This set of (binary) dummy variables are intended to capture the long-run effect of different policies and reforms enacted by different ruling dynasties in the first half of the nineteenth century: the Savoys-Carignano, who ruled in the North Western areas of Italy, i.e. Piedmont, Liguria, and the island of Sardinia; the Habsburgs, who directly controlled the Lombardy-Venetia in the North East, as part of the Austrian Empire, as well as indirectly the Granduchy of Tuscany (Augsburg Lorena) and other minor Duchies nearby (the Duchy of Modena under Este of Austria, the Duchy of Parma and Piacenza and the Duchy of Massa Carrara under Augsburg Lorena); the Pope, in Rome and the Center South; and the Bourbons in the South and in the Duchy of Lucca. We build on Pécout (1999). Medieval commerce: This variable is a dummy taking value 1 if the main city was on Medieval commercial routes or was the seat of a fair or bank and 0 otherwise. To have a graphical idea of the distribution of commercial routes and fairs see Figure 2 taken from Shepherd (1926). University: We create a dummy variable that equals 1 for the presence of an ancient University or Studium (in particular Law’s Schools) in the Middle Ages and 0 otherwise. The sources are Atlante Storico De Agostini and Nuovissimo Atlante Storico Mondiale.

32

Table A2. Summary statistics

Variable Obs Mean Std. Dev. Min Max Schooling rate 207 52.66921 24.63512 9.745636 117.3 Schooling rate at lower level 207 48.15225 22.82749 8.600049 113.7153 Schooling rate at upper level 207 4.469968 3.430144 0 25.5019 Male schooling rate 207 56.88289 24.77012 13.00213 119.8 Female schooling rate 207 48.24242 25.93999 3.767413 114.7 Female to male ratio schooling rate 207 0.83175 0.208213 0.058806 1.515913 Female to male ratio upper schooling rate 207 0.531752 0.262629 0 1.287918 Female to male ratio lower schooling rate 207 0.824465 0.211084 0.04235 1.410763 Industrialization 207 0.918889 0.32904 0.43 2.23 Height 207 163.2855 2.067145 158.1 167.4 Egalitarian nuclear family with early marriage 207 0.304348 0.461246 0 1 Egalitarian nuclear family with delayed marriage 207 0.231884 0.423059 0 1 Incomplete stem family 207 0.130435 0.337598 0 1 Communitarian family 207 0.333333 0.472547 0 1 Savoys 207 0.115942 0.320932 0 1 Augsburgs 207 0.391304 0.489225 0 1 Pope 207 0.173913 0.379954 0 1 Bourbons 207 0.318841 0.467157 0 1 Female to male ratio literacy 1861 207 0.493192 0.183965 0.184799 0.891584 Literacy 1861 207 23.45184 13.08672 8.429249 58.6748 Male literacy 1861 207 30.33475 14.35486 13.19357 67.13915 Female literacy 1861 207 16.58197 12.38707 3.027826 50.95628 Medieval commerce 207 0.492754 0.50116 0 1 University 207 0.376812 0.485762 0 1 Masculinity 207 .9644141 .0229951 .891104 1.053347

The Education Gender Gap: Evidence Following the Italian...

Documents

Transcript of The Education Gender Gap: Evidence Following the Italian...