Methodology

Abstract

In this chapter, I outline the data sources and empirical approaches employed to explore the enduring economic effects of Christian missionaries in Zambia. Utilizing a diverse range of historical, geographical, and contemporary data, I employ distinct empirical strategies. Furthermore, I address the inherent limitations of the data.

Data

Colonial Data Sources

In this study, I draw quantitative data from various historical sources. From 1890 to 1924, Northern Rhodesia was under the administration of the British South African Company (BSAC). As pointed out by Gann (1969) and Gelfand (1961), under the rulership of BSAC, data collection in the territory was sparse; the Company did not invest resources to bolster its capacity to collect data, especially for the African population. When the British colonial government took over the administration of Northern Rhodesia from the Company, they formally established various departments responsible for providing various public services and collecting data. The departments of African education and health were formalized under the rulership of the British colonial government. Though data collection was not always consistent, data collection under the British colonial government was more pronounced than when the Company administrated Northern Rhodesia.

Missionary Stations

Most studies that have analyzed the persistent effects of missionaries in Africa have utilized missionary station location data from the following missionary atlases: Dennis et al. (1903), Béthune (1889), and (Roome, 1925). As argued by Jedwab et al. (2022), these sources underreport the locations of missionary stations. Additionally, missionary stations were not static but increased incrementally over the colonial period. Figure 5.1 compares the missionary locations reported in the world missionary atlases and the missionaries I digitized from the ecclesiastical reports from 1924 to 1948. From Fig. 5.1, it can be seen that nineteen Catholic and Protestant missionary stations are reported by atlases Dennis et al. (1903), Béthune (1889), and 63 Protestant and Catholic missionary stations reported by Roome (1925). There are 212 Protestant and Catholic missionary stations that are reported in the Northern Rhodesian Ecclesiastical reports; there are fundamentally 91 percent missionary stations omitted in Dennis et al. (1903), Béthune (1889), and 71 percent missionary stations omitted in Roome (1925). Most of the world missionary atlases capture missionaries at a single point in time; this study uses detailed information on missionary locations that I collected from the Colonial blue books’ ecclesiastical records. I geocode Protestant and Catholic missionary stations in Northern Rhodesia from 1924 to 1948.

Fig. 5.1

Comparing missionary stations from world missionary atlases and Northern Rhodesian Ecclesiastical reports. (Source: Drawn by Author)

Education

During the Company’s rule, the provision of education by the various missionary agencies was independent of the British South African Company; hence, missionary societies did not receive any financial aid from the BSAC for their education provision mission (Snelson, 1974). After taking over the administration of Northern Rhodesia from BSAC, the British colonial office established a sub-department of African education. The missionary societies that sought funding from the colonial government to provide education had to adhere to a strict native education code defined in the Native Educational Ordinance of 1937 (Kelly, 1999). The information on the various agencies that received aid from the government was published in the returns to Native education reports. In the African education reports, I identify four different education agents in the data: Protestant missionaries, Catholic missionaries, Native Authorities, and Government. The African school returns report also provide the following information, the number of schools administered by each educational agency, the number of boys and girls enrolled in the schools, the average attendance, school fees, voluntary contributions received by each agency, the total expenditure on education, number of Europeans involved in African education, and the amount of government grant received. Additional information on government expenditure on education and missionary grants is obtained from the Colonial blue books’ revenue and expenditure reports 1924–1948. Data collection during the colonial period was not always consistent; ultimately, from the above list of variables, the colonial government consistently recorded the following information, number and types of educational agents, the total number of males and females in the various schools, grants received from the colonial government, and total schooling expenditure per schooling agent. Using this information, we build a panel data capturing educational access for boys and girls in Northern Rhodesia from 1925 to 1953 (20 years).

Healthcare

The British colonial government had established the colonial medical department to provide healthcare in the territory. The colonial medical department kept records of the medical work conducted in the territory in the colonial medical reports. Though these reports contain a plethora of information, the information was not consistently recorded during the colonial period. For example, the work of the medical missionaries is only reported in the medical reports from 1933, while the government’s medical work is reported from 1924. The medical reports provide information on the number of outpatients and inpatients treated at colonial African and missionary hospitals. The medical reports also provide detailed information on the various diseases diagnosed upon admission in African colonial hospitals and the number of people who died yearly. However, the number of deaths recorded were not consistent throughout the colonial period.

Additionally, the colonial medical records contain information on the location of colonial government and missionary hospitals in Northern Rhodesia. For much of the colonial period, the British government operated about 12 African hospitals, and by 1953, the Protestant and Catholic missionaries had established 89 hospitals in Northern Rhodesia. I geocode the exact locations of the African colonial hospitals and missionary hospitals from 1925 to 1953.

Historical Geographical Data

When penetrating the inland, the missionaries followed the explorer routes that previous European explorers mapped. Figure 5.2 illustrates the various European explorers that explored Northern Rhodesia; one of the most prominent and renowned explorers in Northern Rhodesia was David Livingstone; he is believed to have opened up Northern Rhodesia to many Protestant missionaries. I draw information on historical missionary explorer routes from Davies (1971) and Nunn and Wantchekon (2011). I digitize the historical railway line in Northern Rhodesia using contemporary data developed by DIVA-GIS. I also digitize historical rivers in Zambia using information from OKI.

Fig. 5.2

Missionary explorer routes in Northern Rhodesia. (Source: Drawn by Author)

Furthermore, I digitize the location of historical mines using the information I collected from Juif and Frankema (2018). I use the information on historical population densities and cities from Klein Goldewijk et al. (2010). Moreover, Klein Goldewijk et al. (2010) provide information on the grazing land size available in the various African countries and how much of the land is rainfed and the cropland size. The data on the African coastline is obtained from the African Marine Atlas (AMA). I acquired Zambia’s geographical boundary data from the New York University (NYU) spatial repository.

This study uses historical malaria ecology data from Kiszewski et al. (2004). I obtain data on soil productivity from the Food and Agricultural Organization (FAO). The information on the various maize plantations in colonial Zambia is obtained from Jenkin (2018). I collect information on the ruling Paramount Chiefs in Northern Rhodesia from the Ministry of Chiefs and Traditional Affairs Zambia. I obtain information on pre-colonial ethnic groups in Zambia from Murdock (1967). This ethnographic atlas shows whether a specific tribe embraced the bride price tradition. It also provides information on the various kinship systems practiced within the various tribes. The Murdock ethnographic atlas provides information on how much various tribes depend on the various agriculture modes.

Current Data

To study the long-term effects of missionaries, I use various micro cross-sectional contemporary data sets. To analyze the enduring effects of missionaries on education, I use the 1990 Zambian individual-level Census Population, Housing and Agriculture data provided by IPUMS. The census provides information on various individual and household characteristics. Crucial for this study’s empirical strategy is the birthplace of the individual. This variable allows me to determine the Euclidean distance from an individual’s birthplace to a missionary station; the study uses proximity to a mission station as a proxy for missionary exposure. To determine the long-term impact of missionary praxis on HIV and related sexual behavior, this study uses three waves of the Zambian Demographic Health Survey, specifically, the 2007, 2014, and 2018–2019 surveys. The Zambian DHS data is a nationally represented sample containing information on individuals from 13,625 households. The DHS data sets employed in this analysis contain biomarkers for sexually transmitted diseases; these are biological measures of an individual’s health condition. The individual’s HIV status information is not self-reported data but actual HIV test results. Upon surveying individuals, the DHS agents¹ collect dried blood samples (DBSs), which are then sent to the laboratory for HIV testing. As of 2007, the DHS provided georeferenced data for various household groupings, referred to as clusters. To ensure respondent confidentiality, GPS coordinates for urban clusters have 2 kilometers of error, and rural clusters have 5 kilometers of positional error. The georeferenced data is crucial for my empirical strategy; it allows me to measure the distance from a respondent’s place of residence to the closest missionary station and hospitals. I use these distances as proxies for missionary exposure. Additionally, I use a data set on current hospitals in Sub-Saharan Africa locations provided by Maina et al. (2019).

Methods

Missionary Expansion

To capture the synchronic and diachronic impact of missionaries over space and time, I amalgamate various first-hand historical data sources and contemporary data sources and analyze the impact of missionary exposure within a quantitative framework.

As a first step, I explore the determinants of the patterns of missionary settlement in Northern Rhodesia from 1924 to 1953. Fundamentally, the impetus of this analysis is to understand why missionaries were choosing certain locations to build mission stations and churches over others. This analysis is paramount for subsequent long-run estimations in this study because, to closely capture the long-term impact of Christian missionaries in Africa, it is pertinent to understand the drivers of missionary expansion in Africa. If the determinants of missionary expansion are not understood and not controlled for when estimating the long-term impact of Christian missionary exposure, the estimation results may be laden with endogenous bias.

To ascertain the determinants of Missionary expansion in Northern Rhodesia from 1924 to 1953, I partition Zambia into 7134 grid cells of the size 11.1 × 11.1 km. With these partitions, I determine the historical-geographical features of each cell. Using ArcMap, I calculate the average historical population density within each cell, the mean grazing land size, rainfed land size, soil productivity, land elevation, and ruggedness. I determine whether there is a mission station, city, maize plant, and Paramount Chief within each cell.

In addition, I measure Euclidean distances from the centroid of each cell to the nearest river, railway line, mine, and coast explorer route. I repeat this process for every year from 1924 to 1948. Ultimately, I build a novel data set that captures missionary stations expansion between 1924 and 1948 (24 years), containing 176,925 observations. I repeat the procedure outlined above to determine the factors that influenced missionary hospital expansion in Northern Rhodesia. Additionally, I determine whether there is a missionary hospital within each grid cell from 1933 to 1953 (20 years). Consequently, after determining the annual characteristics of each grid cell, I construct a balanced panel data set with 135,546 observations.

One panel data encapsulates the expansion of mission stations from 1925 to 1948, and the other the expansion of hospitals from 1933 to 1953. In the first step, I analyze the factors that determined the expansion of missionary stations in Northern Rhodesia. To this aim, I construct a dichotomous variable that captures whether a mission station was present within a given grid cell annually from 1925 to 1948. I then use this dichotomous variable as an outcome variable to calculate the probability of a mission station being present within a specific cell, given the various cell characteristics. In the second step, I ascertain the determinants of missionary hospital expansion by first constructing a dichotomous variable that determines whether a missionary hospital was present within a given cell annually from 1933 to 1953. Using this dichotomous variable as my main outcome variable, I calculate the likelihood of establishing a mission hospital in a specific cell region given the cell characteristics. For all the estimations, I included year fixed effects. This paper estimates all the models using an Ordinal Generalized Linear Model (OGLM) with a probit link.

Christian Missionaries and Education

To ascertain the impact of missionary exposure on education, I first build panel data that captures the development and accessibility of formal education in Northern Rhodesia from 1925 to 1954. The panel data captures the total number of schools that the various educational agents operated, the total number of boys and girls that were enrolled in their schools, the amount they spent on education, for the Protestant and Catholic agents’ information on the number of grants that were annually received from the government is available.

In the second step, I amalgamate the individual level 1990 census data with the historical, geographical data and missionary station location data. In ArcMap, I draw 30 km buffers around each place of birth. The buffers allow the study to calculate average values of selected historical-geographical features. Within each 30 kms buffer around the place of birth, I measure the average land elevation. Additionally, I measure Euclidean distances from the individual’s place of birth to the nearest mission station, railway line, river, explorer route, coastline, and mine.

Within an Ordinary Least Squares framework, the study determines the extent of the gender gap between Protestant and Catholic schools. Consequently, I use the Euclidean distance from the individual’s place of birth to the nearest mission station as a proxy for missionary exposure. Using this proxy, I estimate the effect of nearness to a mission station on educational attainment for individuals born in the precolonial and postcolonial periods. In the latter regression, I control for various selected factors that may have influenced missionary expansion; I control for these factors to ameliorate endogenous bias in our model. Additionally, the novelty of the individual-level data allows the study to control for tribal, provincial, and birth cohort fixed effects.

Christian Missionaries and Health

To study the effects of Christian missionaries on health, I firstly use descriptive statistics to analyze the extent of missionary and secular medical outreach in Northern Rhodesia and gauge the trained Africans’ level of participation in the colonial health sector. To measure the extent of outreach between missionaries and the colonial government, I compare the number of missionary hospitals and African colonial hospitals established in Northern Rhodesia from 1925 to 1953. Furthermore, I determine how frequently African outpatients attended missionary hospitals relative to African colonial hospitals. Additionally, I analyze using descriptive statistics the number of inpatients that were attended to at missionary hospitals relative to African colonial hospitals.

In the next step, I analyze the enduring impact of missionary exposure on HIV and related health behaviors in Zambia. To achieve this, I merge the three waves of the Demographic Health Survey, 2007, 2014, 2018–2019. In ArcMap, I draw 30 km buffers around the place of residence for each cluster in the new panel data. I then calculate average values for selected historical-geographical features. I calculate the mean land elevation within each 30 km buffer. Additionally, I calculate the distances from each cluster’s residence to the nearest missionary station, missionary health hospital, river, explorer route, and railway line.

Ultimately, I use the Euclidean distance from an individual’s residence to the nearest historical missionary station and hospital as proxies for missionary exposure. In this estimation, the main outcome variables are dichotomous variables for HIV status, condom use at first intercourse and pre-marital abstinence, and a continuous variable that captures the individual’s age at the first sexual encounter and the number of lifetime partners. In this analysis, I control for selected variables that determined missionary expansion to ameliorate any endogenous bias in the estimations. Furthermore, I also control for provincial and occupational fixed effects. I employ a Probit estimation for models that use dichotomous variables as outcome variables. For models that use continuous variables as outcome variables, I use Ordinary Least Squares (OLS) to estimate the models.