This paper is the final project for Prof. Paul Shaman’s course Econometrics I (STAT 520) in Fall 2016. A full electronic version is available at jtcohen6.github.io/projects/stat520final.

The direct inspiration for this work is the paper of İ. Semih Akçomak, Dinand Webbink and Bas ter Weel (hereon “Authors”), which appears in Volume 126 of The Economic Journal (June 2016, DOI: 10.1111/ecoj.12193. The less-direct inspiration is a personal interest in Dutch history and culture, as well as the possibility of my enrollment in graduate study at Leiden (Classics & Ancient History) come Fall 2017.

1 The Problem

The Authors have assembled a wide-ranging, well-read, and overall impressive historical study, examining the formation of a 14th-century Christian order as a means to assess (quantitatively!) conventional wisdom surrounding the rapid rise of Northern European prosperity. Was it the ‘Protestant ethic’? Dutch shrewdness? Private enterprise? The Hanseatic League? Less glamorously, the Authors propose that it was due to a religious order that encouraged literacy and urban living.

Geert Groote was a 14th-century Dutch Roman Catholic deacon who, frustrated with misguided emphasis in mainstream Catholicism, decided to found a religous order, the Brethren of the Common Life (BCL), in Deventer (his hometown). He was fairly cosmopolitan, having attended the University of Paris, and his fundamental idea was similar to the Luther’s Reformation: Christians should read, study, and learn from the Bible themselves. Unlike Luther, Groote was unaristocratic, and his organization developed a true grassroots infrastructure teaching Catholics to read and write. They built monasteries, sponsored schools, and also set up boarding hostels so that poor children from the Dutch countryside could attend school in the cities. One possible result of all this? Increased literacy and urbanization, over the centuries, in those cities where the BCL was present.

The Authors argue that ‘human capital’ – a vibrant, concentrated, and literate population – enabled the Netherlands to begin ‘mercantilizing’ earlier than England, on par with Portugal and Spain and disproportionate to its size. The Netherlands had fewer than one million people in 1500, and yet it published more books per capita than Italy. Book publication, male literacy, and urban population growth are the Authors’ proxy variables to describe economic potential.

2 The Data

There are three sets of data provided by the authors:

growth: fairly detailed data (population sizes, book production share, literacy in 1860, distances, religious-activity dummy variables) for 67 Dutch cities

literacy: more-detailed literacy data (1600, 1675, 1750) and distances for 33 Dutch cities

hinterland: geographic and population data for 91 Dutch, Belgian, and German cities

2.1 Independent Variables

The Authors’ investigation was ultimately one of historical causality: What was the (quantitative) impact of the Brethren of the Common Life? For a religous movement to be ‘affiliated’ with a city, they decided that it must have a monastery, school, hostel, or other movement-affiliated institution located within 5 kilometers, or about an hour’s walk.

BCL_presence: Was there a BCL institution located within 5 km?

BCL_number: How many BCL institutions were located within 5 km?

While BCL_number may give a better sense of the ‘intensity’ of the movement’s involvement in a given community, it is also more liable to bias by a municipality’s population, wealth, and resources. A bigger city could accommodate more institutions, or a smaller city may lack a public school and support BCL-founded education.

Regardless, analysis will be performed for both independent variables.

2.2 Dependent Variables

There are three dependent variables, attempting to show a mixture of human and economic capital. These might be be related or prerequisite to a common understanding of commercial development, as it occurred in the Low Countries during the 16th and 17th centuries.

Book_share is the percentage of total Low Country bookeditions (2,378 in the Netherlands and Belgium between 1470 and 1500) produced in a given city before 1500. The number of book editions is taken from Incunabula Short Title Catalogue of British Library.

Literacy (male, 1600): percentage of bridegrooms from a given city (33 available) who signed name on marriage registration instead of stamping fingerprint; 12,152 total observations at the Amsterdam Archives. Literacy (male, 1860) is data collected from 19th-century birth registers in the Historical Sample of the Netherlands (HSN), aggregated by authors from provinces to municipalities.

City growth (1400-1560) for 67 cities with 1,000 or more residents in 1400. Most population estimates are the work of recent scholars, from whom the Authors took the most relevant and compelling figures. Given the disparities between municipalities and centuries–rapid growth overall and especially in urban areas following the agricultural and industrial revolutions–it makes sense to use the logarithm of population percentage change.

2.3 Exogenous Variables

There are a number of variables that, while (ideally) uncorrelated with the independent variable, may have influence over the dependent variable. It is necessary to include these in the more-robust regression procedures in order to control for their confounding impact, thereby isolating the partial effect of BCL affiliation.

Schools_before_1400: Proxy for a city’s educational infrastructure prior to and independent of BCL affiliation

*Population_1300; _1400; growth_1300_1400*: Control for the impact of city size and prior urbanization trends upon the growth 1400-1560

The presence of other major religous orders. The five largest Dutch monastic movements were Modern Devotion (which included BCL), Tertiarians, Franciscans, Beghards and Beguines, and Cistercians.

Other proxy variables for economic potential and wherewithal: location on a river or seaport; distance to a Roman colony or road (trade route); membership or affiliation with the Hanseatic League (international trade). Distances to various cities also factor in as needed; for instance, any regression attempting to explain literacy should include Distance_to_Amsterdam, since all literacy records before the 19th century are from the Amsterdam Archives and might be biased toward those cities closer to Amsterdam.

2.4 Summary Statistics

The following table includes means, standard deviations, and two-sample t-test P-values for the dependent, instrumental, and exogenous variables in each of the two crucial samples: those Dutch cities unaffiliated with the Brethren of the Common Life (BCL=0), and those affiliated (BCL=1). [Cf. Authors’ Table 4, p. 841] The t-tests are two-tailed; that is, the null hypothesis for each variable is the difference between BCL and non-BCL cities should be trivial in either direction.

2.5 Discussion: Book Share

Book production is a good national metric for cultural capital and economic development. Estimates of per capita gross-domestic product in Western European nation-regions, reflecting scholars’ direct claims about economic wellbeing, are plotted against per capita book edition printing in the same nation-regions. Here for 1500:

The two per capita variables, representing human and economic capital, have a correlation of 0.901.

*Notes:* Per the authors [p. 834], population and per-capita GDP numbers are taken from Maddison Project. Unfortunately, no GDP estimates are available for France before 1820 (at which point they are rather detailed), even though France and French-speaking Switzerland contain 5,437 book editions and 15.3 million people (over a quarter of Western Europe) by 1500. The book publication figures group together Germany with German-speaking Austria and Switzerland, although GDP figures are only available for the first. The same is true of Spain and Portugal, although only the former has a GDP estimate for 1500; Portugal’s tends to be higher than Spain’s (from 1550 on, and probably earlier due to its earlier Reconquista), but the Spanish estimate (probably too low) was used for both.

Of course, only a handful of municipalities are responsible for most of the printing in a given nation. This unfortunately means that book_share is not a particularly useful metric for representing human-capital differences inter urbis.

Only five of those cities with nonzero book production also appear in the Amsterdam Archives data offering literacy in 1600. Comparing book_share and literacy among those five: