Household Balance Costs and Benefits

Land titling is costly. In Senegal, a land title costs around 5,000 CFA, or about 8 USD, per hectare. In Côte d’Ivoire, Bassett (Reference Bassett2020, 144) enumerates “over 20 steps to obtain a land certificate and another dozen to obtain a land title,” many of which involve a fee. These steps involve multiple levels of government: the village land management committees (CVGFRs), the sous-prefectures, and the Agence Foncier Rurale (AFOR) in the capital. Such processes are common across the continent.Footnote ¹⁰

However, the costs to titling go beyond monetary fees or time spent in the sous-prefect’s office. Land tenure formalization is not a one-to-one mapping of existing land use onto paper; it creates winners and losers. In much of Africa, agricultural parcels are subject to overlapping ownerships which exist in a state of strategic ambiguity (German Reference German2022; Lund Reference Lund2008). One person may have the right to farm in the dry season, another in the rainy season. A third person may have the right to graze their animals on the parcel. A fourth person may be the descendant of the original inhabitants of the area, who collects customary (but debatably ceremonial) rents. Which of these four owns the parcel? Land titling forces the issue of hierarchy between these partial owners. Formalization may carry a particular risk for “groups such as women, pastoralists, hunter-gatherers, casted people, former slaves, and serfs, who have traditionally enjoyed subsidiary or derived (usufruct) rights to land” (Platteau Reference Platteau1996, 40).

Households pursue titling despite the expense and the risk because land conflicts are an unfortunately common occurrence in much of Africa. In Côte d’Ivoire for example, settlement patterns by Burkinabé and Baoulé migrants led to large scale uprisings with as many as 4,000 casualties in the 1970s (Boone Reference Boone2003, 220). In parts of Northern Ghana, the dispossession of historical elites led to conflicts between the village chiefs and the dispossessed earthpriests (Lund Reference Lund2008). The fact that individuals who are more highly placed within customary institutions feel more secure in fallowing land likewise highlights the risk of expropriation (Goldstein and Udry Reference Goldstein and Udry2008). Formal property rights can help alleviate such concerns. Titling one’s land reduces the risk of losing it.Footnote ¹¹

Households will be more willing to title their land when the value is higher. Where an asset is more expensive, households will go to greater lengths to protect it—including undertaking a costly titling process (Besley and Ghatak Reference Besley, Ghatak, Rodrik and Rosenzweig2010). These arguments echo political economy theories of endogenous institutions, which posit that property rights emerge when the individual benefits to organizing such a system become equal to the individual costs. Shifts in relative prices can shock prevailing equilibria and drive institutional change (Libecap Reference Libecap1989). Rosenthal (Reference Rosenthal1992, 21) illustrates this dynamic clearly in Revolution-era France, where “it was not worthwhile to define property rights to unimproved land clearly, for enforcing such rights would have required monitoring unwarranted by the low value of the land.” In the context of land formalization, a shift in the value of land should drive rural households to seek formal titles for their parcels.Footnote ¹²

Land titling also incentivizes households to invest in their land. Households with more secure land are able to invest with comparatively greater surety that they will receive the returns to these investments (Dillon and Voena Reference Dillon and Voena2018; Goldstein and Udry Reference Goldstein and Udry2008). These investment can be short term, such as fallowing one’s land or investing in fertilizer, or long term, such as planting tree crops which can take four to five years to become productive. The potential returns to investment in the parcel drive titling by incentivizing households to protect their future investments in the land.

Land tenure insecurity is not always the binding constraint to investing in one’s agricultural parcels. Much land in sub-Saharan is arid or infertile (Herbst Reference Herbst2014). Many seemingly lush tropical areas suffer from poor soil. In such areas, the benefits to households from seeking a formal land title will be lesser—more secure land title will not unlock investment where the potential returns to such investments are too low. These dynamics suggest that households will formalize their land only when the value of the land or the returns to investment in agricultural parcels is high enough to justify the costs. More specifically, I hypothesize that:

Chiefs and Precolonial Institutions

Households do not title in a vacuum. Most villages in sub-Saharan Africa have some manner of customary leader—most often a village chief (Baldwin Reference Baldwin2016). These chiefs have an incentive to prevent households from seeking land titles. However, chiefs do not always have the capacity to act on this incentive to prevent titling.

Customary authorities, most often chiefs, are important political actors across sub-Saharan Africa (Baldwin Reference Baldwin2016). Chiefs’ control over land reinforces their authority in other dimensions. Chiefs can use control over insecure land to sanction households who defy the chiefs (Acemoglu, Reed, and Robinson Reference Acemoglu, Reed and Robinson2014). Even without such threats, the role chiefs play in resolving land disputes enhances their perceived authority. Chiefs are often the first step to resolve land disputes (Ribar Reference Ribar2023). When households bring their disputes to the chief they implicitly recognize the chiefs’ authority to arbitrate such disputes. Lund (Reference Lund2008, 10) summarizes this point, that “[r]ecognition of property rights by an institution simultaneously constitutes a process of recognizing the legitimacy of the institution.” By keeping land rights in the customary regime, chiefs maintain a long-term reservoir of political legitimacy. Baldwin and Ricart-Huguet (Reference Baldwin and Ricart-Huguet2023) illustrate this dynamic: households across sub-Saharan Africa perceive their chiefs to be more authoritative where land values are higher because of increased competition over land. If control over land held in the customary system enhances chiefs’ political authority, then removing control of land to a centralized formal land regime will reduce their authority.

Chiefs also act as development intermediaries. Baldwin (Reference Baldwin2016) notes that much of chiefs’ legitimacy as political actors comes from their performance on the job: constituents prefer a chief that delivers development goods, such as roads or clinics.Footnote ¹³ Land titles are another form of development good. However, the extent to which chiefs can claim credit for such development goods depends on where decisions are made. Where land tenure administration is centralized, credit-claiming becomes more difficult, creating a second rationale for chiefs to impede titling.

Stronger chiefs are better able to act on their incentives to impede titling.Footnote ¹⁴ Scholars often measure the strength of precolonial institutions by the number of hierarchical layers of governance (Honig Reference Honig2022; Neupert-Wentz and Müller-Crepon Reference Neupert-Wentz and Müller-Crepon2024). Chiefs situated within hierarchical institutions are better able to prevent land titling because they are more empowered to enforce their decrees. Within weakly hierarchical institutions, customary elites may not have the political capital to enforce judgements (Boone Reference Boone2003). Increased within-village hierarchy may also help chiefs to hold households accountable because within-village elites such as lineage heads can convey grievances to or from the chief or act as the chief’s lieutenant.

A variety of literature explores how precolonial institutions affect contemporary outcomes. Honig (Reference Honig2022, 14), for example, notes that

Similarly, Wilfahrt (Reference Wilfahrt2022) notes how the overlap of precolonial institutions and contemporary state organs affects local redistributive politics in Senegal. Neupert-Wentz and Müller-Crepon (Reference Neupert-Wentz and Müller-Crepon2024) show that areas with hierarchical precolonial institutions have greater levels of contemporary political complexity. Precolonial hierarchy is especially pertinent when it comes to chiefly control over land, because much of the chief’s authority over village life is predicated upon their control over land (Acemoglu, Reed, and Robinson Reference Acemoglu, Reed and Robinson2014). Honig (Reference Honig2022) further argues that the presence of strong customary institutions reduces titling directly. Hierarchical institutions produce vertical accountability, in that customary elites can sanction other customary elites, as well as horizontal accountability, because these chiefs are subject to some degree of checks and balances. The result, Honig argues, is that these institutions “hold leaders accountable to institutional goals” and “slow the erosion of customary land tenure” (Reference Honig2022, 9).Footnote ¹⁵ This political authority means that chiefs are better able to act on their incentives to prevent land titling.

Putting these factors together, chiefs in centralized land regimes have an incentive to impede land titling, and stronger chiefs are better able to prevent titling. As a result, I hypothesize that

Devolved and Centralized Land Regimes

Titling reduces the power of chiefs when it removes land administration to a centralized authority. However, titling can buttress, rather than diminish, the power of chiefs in areas where chiefs can capture the process. In countries with centralized land tenure regimes, decisions around land tenure formalization are made at the national level. For example, the Liberian Land Authority presents itself as a “one-stop shop” for land tenure formalization. Land administration and titling occur in Monrovia—far from customary institutions. Another strong example is Rwanda, where the country’s comprehensive land tenure formalization drive was managed entirely by the state. Rwanda selected areas in which to title, household claims were mapped, and then the central land registry office published the information (Ali, Deininger, and Goldstein Reference Ali, Deininger and Goldstein2014). Other countries devolve their land tenure regimes to local authorities. Land administration is a difficult task: agricultural land often exists in hinterlands where state capacity is comparatively limited (Herbst Reference Herbst2014).

Where control of land is devolved, chiefs are able to capture the land tenure formalization process. For example, municipal councils in Senegal issue the rural land certificates (délibérations foncières), but in practice the municipal councils rely on chiefs to guide land titling procedures and resolve disputes. In a survey of 1,164 household heads across rural Senegal, 92% of households said it would be necessary to inform the chief to acquire a délibération foncière, 47% said it would be necessary for the chief to investigate your claim, and 18% said it would be necessary to pay the chief (Ribar Reference Ribar2023). Similarly, village land management committees (comitès villageois de gestion foncière rurale, or CVGFRs) in Côte d’Ivoire investigate land claims, although titles (certificats fonciers [CFs]) are formally distributed by the national land bureau. Village chiefs have no official position in CVGFRs, but they nevertheless almost always head the committee.Footnote ¹⁶

Chiefs can capture devolved land titling in both neocustomary and statist land regimes (Boone Reference Boone2014, 24–5). In statist regimes, “governments administer the allocation and holding of rural property directly”; in the latter, land is governed indirectly through customary institutions.Footnote ¹⁷ Local land tenure arrangements are complex, heterogeneous, and often illegible to the state (Scott Reference Scott1998). As a result, chiefs play a large role in adjudicating land claims even when land administration is officially in the hands of the state. The important factor in my analysis is chiefs’ de facto control over titling, rather than their de jure control.

This capture of the titling process means that chiefs do not lose their power over land tenure, even when households seek title. As a result, chiefs can use titling to advance their own political agendas. For example, Onoma (Reference Onoma2010, 81) shows how chiefs in Ghana used land tenure formalization to “purchase and ensure political support.” Different chiefs have different incentives: in Côte d’Ivoire, chiefs’ primary agenda is to exclude outgroups from titling. Devolved land titling also opens space for chiefs to use land titles as development goods to increase their perceived legitimacy (Baldwin Reference Baldwin2016). Because land titling under devolved land regime lets chiefs advance their agendas without eroding their power, chiefs within a devolved land regime have an incentive to facilitate titling.

Strong chiefs are better able to act on these incentives—they have greater control over villagers and so can facilitate titling. Acemoglu, Reed, and Robinson (Reference Acemoglu, Reed and Robinson2014, 323), for instance, find that stronger chiefs in Sierra Leone “have more authority to influence whether or not people can farm or sell a piece of land.” It may also be the case that stronger chiefs are more able to capture land titling institutions, but my Ivorian case study suggests that even comparatively weak chiefs are able to control their village land committees.

This argument also does not contract Honig (Reference Honig2022), who argues that strong customary institutions prevent individual chiefs from acting against the collective interests of customary institutions. If titling does not go against the interests of customary institutions writ large—in this case, because chiefs can capture titling and maintain a degree of power over land—then the accountability provided by hierarchical institutions need not impede titling.

Why do some African states devolve land tenure? Centralized land tenure administration was the default condition across colonial Africa. Centralized administration was necessary to alienate productive land from African populations (Hailey Reference Hailey1938, 1649). Starting in the late 1980s, the World Bank promoted a number of decentralization proposals across the developing world (Deininger Reference Deininger2003). A number of MCC compacts, such as those in Benin and Niger, also promoted decentralization reforms. These reforms are largely donor-driven, rather than reflective of domestic politics. Decentralization programs are also a bad candidate for clientelist or targeted policies, because by nature, they affect entire countries. German (Reference German2022) further notes that many of the structural reforms and other projects are more reflective of a World Bank-driven “knowledge-regime” than of local circumstance.

In summary, customary land enhances the power of chiefs. When households title land, it exits the customary system. Where the administration of land tenure is devolved, chiefs can retain some control over land tenure through their capture of local land institutions, and so titling land does not erode the chiefs’ authority. In devolved land regimes, chiefs have an incentive to facilitate land titling so they can use it to advance their political agendas and potentially claim credit for providing a development good. Where land tenure formalization is centralized, titling disconnects chiefs from their reservoir of political authority and so they have an incentive to oppose it. The incentive to impede or facilitate titling does not confer the ability: stronger chiefs—measured through the presence of hierarchical precolonial institutions—are better able to act on these incentives. In contrast to areas with centralized titling, I predict that:

Outcome Variable: Household Titling

The lack of accessible administrative data on land titling has hampered the ability of scholars to study the subject. To sidestep this problem, I combine data from the DHS and LSMS projects. The DHS project collects comparable data on developing countries around the world. In its module on agriculture and landholding, the DHS asks “do you have a title deed or other government recognized document for any land you own?” Like the DHS, the World Bank’s LSMS program is a large scale effort to collect comparable data across the developing world. The LSMS contains a parcel-level roster of agricultural land and asks “[d]oes your household currently have a title or ownership document for this parcel.”

These similar questions allow me to construct the main outcome variable of the article: a binary indicator for whether a household possesses a title for at least one parcel of agricultural land.Footnote ¹⁹ I also extract the age, sex, and education of the household head, which I include as demographic control variables. Finally, I capture a binary indicator for whether the household has a title for their dwelling, which I use as a placebo outcome in Section A.3 of the Supplementary Material.

These data provide observations for 22 countries across 62 survey waves. Six countries have only one survey wave; the remainder have at least two. Survey data come from the period between 2010 and 2021, with only three waves taking place earlier (Uganda in 2005 and 2009 and Tanzania in 2008). While the DHS data have been collected since the late 1980s, questions about land tenure only appeared in round seven of DHS, which started in the mid-2010s.

Land Values and Returns to Investment

Illegible and informal land markets in most of Africa prevent researchers from directly measuring realized values of land.Footnote ²⁰ To overcome these issues, I implement a novel measurement of “attainable value”: the value for which the maximum attainable yield (per hectare) could sell for on the international commodity market. I combine geospatial crop suitability data with global commodity prices to obtain these land values at a 10km-by-10km grid cell level. Adjusting the underlying parameters of the crop suitability models creates two measures of the returns to agricultural investment: the marginal increase in attainable value from using fertilizer, and the marginal increase in attainable value from planting tree crops as opposed to nontree crops.

I use version 4 of the Food and Agriculture Organization (FAO)’s Global Agro Ecological Zone (GAEZ) model to obtain the “attainable yield” of different crop types. The model takes into account climate, soil and terrain data, as well as phenology and crop calendars to calculate the attainable yield of crops for a 10km by 10km grid.Footnote ²¹ Potential total production is divided by total grid cell area. Prices come from the IMF’s Primary Commodity Price System.Footnote ²² I set prices to constant 2011 USD. The commodities included in these data are: bananas, barley, chickpeas, canola oil, cocoa, coconuts, coffee, cotton, corn, groundnuts, oats, palm oil, rice, rubber, sunflower oil, soybeans, sorghum, sugar, tea, and wheat.

For each crop and grid cell, I multiply the maximum attainable yield (metric tons per hectare) by the commodity prices in a given year (USD per metric ton) to obtain the attainable price (USD per hectare) for each crop. I then take the maximum of this vector. More formally, the maximum attainable value $ \pi $ of grid cell g in year y is defined as

$$ \begin{array}{rll}{\pi}_{g,y}=\underset{c}{\max }({p}_{c,y}\cdot {s}_{c,y,g}),& & \end{array} $$

where p indicates crop price, s indicates the attainable yield, and observations are indexed by g for grid cell, y for year, and c for crop. These data measure the maximum attainable value in constant 2011 dollars per hectare for a given 10km-by-10km grid cell on a yearly basis.

This measure calculates the attainable value of agricultural production per hectare; in other words, in captures the value of agricultural land. However, testing H2 also requires a measure of the returns to potential agricultural investment. To that end, I calculate two additional variables: the returns to using fertilizer (a short-term investment) and the returns to planting tree crops (a long-term investment). The primary land value measure assumes that households do not fertilize their parcels. Fertilizer is not a consistent or linear multiplier for yields. In some locations and for some crops, fertilizer greatly increases yields. Elsewhere, the returns are minimal. By re-calculating the GAEZ attainable yield models assuming fertilizer use, I can take the difference to identify the returns to fertilization.

I also calculate the returns to cultivating tree crops over nontree crops. Tree crops require a high up-front investment: households must purchase saplings or wait for trees to become productive. Coffee trees, for example, take five to seven years to become commercially viable. As a result, tree crops represent a longer-term investment than purchasing fertilizer for the remaining year. Of the 20 crops for which I have price data, I classify bananas, coconuts, cocoa, coffee, rubber, and tea as tree crops. I calculate the maximum attainable value for tree crops, and subtract the maximum attainable value for other crops. Where this difference is negative, I reset it to zero, in recognition that farmers would not make an unprofitable decision.

Farmers do not receive the global commodity prices. However, the validity of this measurement requires only that the prices farmers receive are positively correlated with global commodity pricing. In Section B.2 of the Supplementary Material, I probe this requirement using a subset of data for which households’ planting decisions are available. Households respond to this measure of attainable value: a 1% increase in the average attainable value per hectare of a crop is associated with a 0.13–0.16 percentage point increase in the amount of land that farmers dedicate to that crop. Similarly, an increase of 1% in the fraction of an administrative area in which a crop is the most profitable is associated with a 0.076–0.08 percentage point increase in the fraction of land that farmers dedicate to that crop. In other words, the elasticity of crop planting with regards to the crop’s attainable value is positive, which supports this land value measure capturing the underlying phenomenon of agricultural production.

This metric superficially resembles a shift-share instrumental variables (SSIV) design. However, my measure is not an instrument—attainable yield operationalizes the latent land values directly. The “methodology” section discusses my methodological precautions in greater detail.

The Strength of Customary Institutions

To measure the extent to which chiefs are able to act on their incentives to impede or facilitate titling, I use geo-referenced data from Murdock’s Ethnographic Atlas (Moscona, Nunn, and Robinson Reference Moscona, Nunn and Robinson2020; Murdock Reference Murdock1967), a common reference for differences among precolonial institutions. The Murdock dataset includes 89 variables on 802 different ethnic groups around the world, of which 239 are located in sub-Saharan Africa. The specific variable from the Murdock dataset through which I operationalize chiefs’ capacity is the precolonial institution’s level of hierarchy, which Murdock measures in two numbers:

This article uses the first of these numbers: the number of administrative levels within villages.Footnote ²³ This local hierarchy is more likely to reflect local norms, rather than constellations of power which were sedimented by colonial regimes (Chanock Reference Chanock1991). Acephalous societies, such as the independent villages of northern Ghana, would rank at the lowest level (Nathan Reference Nathan2023). In Senegal the Imamate of Fouta Toto possessed a ruling council, regional chiefs, village chiefs, neighborhood chiefs, lineage heads, and household heads. Such a precolonial kingdom ranks at a four-four (the highest levels for both sets of hierarchy) on the Murdock scale.

Precolonial elites used the institutional discontinuity brought about by European rule to supplement their own authority (Berry Reference Berry2001; Boone Reference Boone2014; Lund Reference Lund2008). In many cases, what today constitutes “custom” reflects the powers that customary elites could convince the colonial state they possessed (Chanock Reference Chanock1991). In other cases, customary elites and chiefs were entirely invented by colonial powers (Nathan Reference Nathan2023). Nevertheless, what matters for my analysis is not the specific powers claimed by the village chief, but rather the extent to which the chief can act on their incentives to facilitate or impede land titling. Within-village precolonial hierarchy increases this ability because it provides the chiefs with more avenues to sanction households who act against the chief as well as subordinates—such as lineage heads—through which to administer the village.

There is undoubtedly noise in the relationship between precolonial hierarchy and contemporary institutions.Footnote ²⁴ However, a variety of literature suggests that, on average, contemporary political institutions are stronger in areas where precolonial institutions were more hierarchical (Honig Reference Honig2022; Neupert-Wentz and Müller-Crepon Reference Neupert-Wentz and Müller-Crepon2024; Wilfahrt Reference Wilfahrt2022). This source of error should be uncorrelated with the independent or dependent variables, so this source of error would attenuate my results toward zero rather than bias them. As a result, one can consider this article’s results as a lower bound on the relationship between land values and titling.Footnote ²⁵

Ultimately, the hierarchy of precolonial institutions is an indirect measure of the strength of contemporary customary institutions. However, it has a crucial advantage over alternatives: it is comprehensive. This measure covers the universe of precolonial polities in Africa. These data are geospatial, so they produce variation at the levels where my other data sources vary. Section D of the Supplementary Material provides additional evidence of the correlation between these Murdock data and the strength of current precolonial institutions.

Devolved and Centralized Land Regimes

Finally, I code whether land regimes in a given country are centralized or devolved. This distinction matters because chiefs are better able to capture the land tenure process where decision-making is devolved. I classify a land regime as devolved if decisions about whether a household can title a given parcel are made at the national level (centralized) or another level (devolved). This variable operationalizes the extent to which chiefs are able to capture land titling in a way that preserves my ability to make cross-national comparisons.

I specifically code this variable based on where the decision is made, rather than the location at which land titles are certified or recorded. For example, in Senegal, déliberations foncières are adjudicated, issued, and recorded by municipal councils, a third-level administrative devision. Côte d’Ivoire is a more complicated case: village-level CVGFRs decide who can obtain a certificat foncier, the sous-prefect verifies that correct procedures were followed, and the national land agency ultimately certifies the title. While administrative procedures take place at all three levels, this is a devolved land regime because villages make the decisions. The process is similar in Zambia: the Office of the Commissioner of Lands ratifies decisions made by chiefs (Honig Reference Honig2022). In contrast, the National Land Agency of Rwanda maps parcels, adjudicates titling, and issues certificates—a centralized land regime (Deininger and Goyal Reference Deininger and Goyal2024, 60–2). Section E of the Supplementary Material gives a complete list of countries and more detailed narratives which explains exactly how and why I classified these countries. No countries changed land regimes during the period of study.

Methodology

I analyze these variables using a series of linear probability models with fixed effects at the country-survey wave level. I also include a set of demographic controls, including the household head’s education level, sex, age, marital status, and whether the household is urban or rural. My geographic controls are measured at the second-level administrative division, and include geographic area, population density of the district, an interaction of area and population density, average terrain ruggedness (Carter, Shaver, and Wright Reference Carter, Shaver and Wright2019), caloric yield (Galor and Özak Reference Galor and Özak2016), average road density, and density of highways (Meijer et al. Reference Meijer, Huijbregts, Schotten and Schipper2018). I transform the road density variables using the inverse hyperbolic sine to improve normality.

I model the moderating effects of land regime and informal institutions on the relationship between land values and titling through a series of triple interaction models. The triple interactions are necessary to capture the dependencies between land values/returns to investment, hierarchy, and land regimes. These regressions take the form:

$$ \begin{array}{r}{y}_{idnw}={\beta}_1{V}_d+{\beta}_2{H}_d+{\beta}_3{V}_d\cdot {H}_d+{\beta}_4{H}_d\cdot {C}_n\\ {}+{\beta}_5{V}_d\cdot {H}_d\cdot {C}_n+{X}_i+{Z}_d+{\psi}_{nw,}& & \end{array} $$

where V indicates the land value variable (I estimate separate equations for each of three land value/agricultural investment variables), H represents the fraction of the district covered by hierarchical precolonial institutions, and C is a binary indicator for whether the country devolved its land administration. X and Z are vectors of household-level and district-level controls (respectively), i indexes observations by individual, d by district, n by country, and w by survey wave.Footnote ²⁶ To facilitate interpretation, I also calculate marginal effects for all three land value variables. Tables 1–3 all include these control variables; Tables A1–A3 in the Supplementary Material replicate these analyses without controls, with similar results.

Table 1.

Strong Customary Institutions Moderate the Relationship Between Land Value and the Uptake of Land Titles

Note: The dependent variable is whether a household possesses a land title. The independent variables are the maximum attainable value; the fraction of an administrative unit that is covered by a hierarchical precolonial institution; and whether the country devolved its land regime. The unit of analysis is the household. Land value data vary at the second-level administrative division. Demographic controls include the age, sex, and education of the household head; geographic controls include area, population density, an urban/rural indicator, road density, highway density, and terrain ruggedness. Table A4 in the Supplementary Material shows coefficients for these controls. Data are from the DHS and LSMS projects. All regressions use OLS with survey weights and country-wave fixed effects. Standard errors are clustered at the country-wave level.

Table 2.

Strong Customary Institutions Moderate the Relationship Between Returns to Fertilizer and the Uptake of Land Titles

Note: The dependent variable is whether a household possesses a land title. The independent variables are the increase in maximum attainable value from fertilizing a parcel; the fraction of an administrative unit that is covered by a hierarchical precolonial institution; and whether the country devolved its land regime. The unit of analysis is the household. Land value data vary at the second-level administrative division. Demographic controls include the age, sex, and education of the household head; geographic controls include area, population density, an urban/rural indicator, road density, highway density, and terrain ruggedness. Table A5 in the Supplementary Material shows coefficients for these controls. Data are from the DHS and LSMS projects. All regressions use OLS with survey weights and country-wave fixed effects. Standard errors are clustered at the country-wave level.

Table 3.

Strong Customary Institutions Moderate the Relationship Between Returns to Planting Tree Crops and the Uptake of Land Titles

Note: The dependent variable is whether a household possesses a land title. The independent variables are the difference in maximum attainable value between planting tree crops and planting other crops; the fraction of an administrative unit that is covered by a hierarchical precolonial institution; and whether the country devolved its land regime. The unit of analysis is the household. Land value data vary at the second-level administrative division. Demographic controls include the age, sex, and education of the household head; geographic controls include area, population density, an urban/rural indicator, road density, highway density, and terrain ruggedness. Table A6 in the Supplementary Material shows coefficients for these controls. Data are from the DHS and LSMS projects. All regressions use OLS with survey weights and country-wave fixed effects. Standard errors are clustered at the country-wave level.

Table 4.

Chiefly Authority Increases Titling for Autochthones but Not Allochthones

Note: The dependent variable is whether a respondent has a formal land title for at least one agricultural parcel. The independent variables are indices of responses to: [f]or each of these activities [that chiefs in Côte d’Ivoire sometimes mandate], I would like you to tell me how much of the village you think would do what the chief asked them to do. Demographic controls include education, sex, age, ethnicity, the respondent’s relationship to the household head, and wealth. Geographic controls include an indicator for PAMOFOR, distance to department capital, cocoa suitability, coffee suitability, and terrain ruggedness. Table C7 in the Supplementary Material shows coefficients for these controls. All regressions use OLS with inverse sampling probability weights. Standard errors are clustered at the administrative village level.

Threats to Inference

I measure land values by interacting the attainable yield per hectare for a variety of different agricultural products with their commodity prices on the global market and taking the maximum. This estimation strategy is a form of weighted exposure to common shocks design (Borusyak and Hull Reference Borusyak and Hull2020): the crop prices are the shocks, and the total attainable yields per hectare are the weights. These shocks themselves are exogenous; no individual farmer or country can affect the a crops global commodity price. One potential threat to inference in this case is that observations’ weights (i.e., their crop suitabilities) are likely not entirely exogenous to land titling. Soil quality may have other causal pathways to land titling rates; for example, Baldwin and Ricart-Huguet (Reference Baldwin and Ricart-Huguet2023) show that land quality affects the power of traditional leaders. In such cases, a nonrandom exposure to common shocks research design can lead to omitted variable bias (Borusyak and Hull Reference Borusyak and Hull2020). I include the average shock across all years for observation i to control for this bias.

A second threat to inference is that a country’s land tenure regime, centralized or decentralized, is endogenous to other variables. Albertus (Reference Albertus2021, 169), for example, points out that farmers who are “fixed geographically and lack property rights while facing obstacles to acquiring necessary agricultural inputs and credits are the stuff of clientelist party fantasy.” As a result, states may centralize land regimes in order to withhold titling. Many African leaders are also large-scale landowners, which gives them economic incentives to promote or withhold titling (Onoma Reference Onoma2010). Countries may also make the decision to devolve land titling on the basis of geography: devolution would be especially useful in countries with large hinterlands which complicate governance (Herbst Reference Herbst2014). Mali and Niger, for instance, both have sparsely populated hinterlands and devolved land tenure. However, small countries, such as Ghana and Lesotho, also devolved land tenure.Footnote ²⁷ The decision to devolved tenure could also involve different social alliances and constellations of power: states are more likely to devolve power to rural elites who support their political regime (Boone Reference Boone2003).

Many of these variables—politics, geography, and political coalitions—may co-vary with land tenure regimes.Footnote ²⁸ I control for the direct effect of these variables through country-wave fixed effects. I also control indirectly for both population density, administrative area size, and their interaction in all regressions which should control for being in a hinterland. Devolved land rights is also a national variable, whereas local power blocs or hinterlands are subnational. One important note is that these control variables do not eliminate the possibility that it is devolved land regimes which drive these results. Variables which correlate with devolved land tenure—such as state capacity or bureaucratic weakness—could confound these results when interacted with land values and strength of chiefs. Ultimately, the control variables and robustness checks minimize the probability of unobserved variables confounding my results, but do not eliminate it.

The concept of ownership is not constant across sub-Saharan Africa. Land ownership is often plural, in that multiple actors may have at least some claim over a particular parcel. Increased costs of titling due to plural ownership of land could reduce titling rates in certain areas and bias the results. The detailed subset of LSMS data collected as part of the EHCVM helps identify the extent of collective ownership. These surveys contain data on 148,885 separate parcels across 42,287 landholding households. Among these households, 70.6% individually own their parcels, 11% collectively own their parcels, and 18.4% own some parcels collectively and own some parcels individually.Footnote ²⁹

Because my outcome variable is whether a household possesses at least one land title, so this statistical bias would exist only for the 11% who only hold land collectively. None of these 11% of households report owning their collectively held land; instead, they report holding it through loans or sharecropping agreements. I exclude these households because I limit my analysis to landholding households. While the DHS data collection does not ask the mechanism through which a household holds land, I have no reason to believe that collective ownership is more common in countries where DHS data collection took place than in countries where LSMS data collection took place. I cannot entirely exclude the possibility that differing norms of collective ownership create bias in my results, but statistics from a subset of the data suggest that this problem applies only to a small percentage of respondents.

Finally, another potential threat to inference comes from the potential endogeneity of donor funded land tenure formalization programs. Donors may target programs to areas where land values are high; in turn, these programs will likely increase rates of land titling. As such, donor funded land tenure formalization programs are a mediator variable, rather than a confounding variable (Morgan and Winship Reference Morgan and Winship2015). It is possible that adding this variable into my analysis would attenuate my results. Unfortunately, there is no comprehensive database of land tenure formalization programs with sufficient geographic granularity to explore this mediator. However, donor-funded land programs would only mediate the direct relationship between land value variables and titling. Land tenure formalization programs often subsidize the administrative costs of titling, but they do not change chief’s incentives to impede or facilitate land titling. As such, including donor-funded land programs would be unlikely to affect the overall marginal effects I display in Figure 2.

Figure 2.

The Marginal Effect of Land Variables by the Presence of Precolonial Institutions Depends on the Prevailing Land Regime

Note: This figure shows the marginal effects of three land variables on the probability a household has a title, broken out by the percent of the administrative division covered by a hierarchical customary institution and whether the country has a devolved or centralized land regime. Results are from column 6 of Tables 1–3.

Strong Chiefs Lead to More Titles

Côte d’Ivoire devolved land tenure to CVGFRs, so my theory would predict that villages with authoritative chiefs would have more titles. I measure chiefly authority using seven survey questions: “[f]or each of these activities [that chiefs in Côte d’Ivoire sometimes mandate], I would like you to tell me how much of the village you think would do what the chief asked them to do.” The seven activities were: (1) Participate in village cleanup day; (2) Give up a piece of land for a school; (3) Spend a day repairing a road; (4) Spend a day planting trees; (5) Come to participate in a traditional dance; (6) Give money to support a traditional ceremony; and (7) Give up a piece of land for a mosque. For each activity, respondents answered: nobody, very few people, some people, most people, or everybody. I create two indices. One captures chiefs’ political authority and includes items one to four. The other captures chiefs’ authority in a more traditional sphere and includes items five to seven.Footnote ³³ Chiefs’ political power is the metric which aligns with my overall theory: strong chiefs are those with higher political power. I include traditional authority for a comparison.

Overall, approximately 38% of households possess at least one formal land title. Table C6 in the Supplementary Material shows the relationship between the chiefly authority indices and household titling.Footnote ³⁴ The outcome variable is a binary indicator for whether a household possesses a formal land title for at least one of their parcels. The independent variables are the two indices of chiefly authority. Columns 1–3 show no relationship between chiefly authority and titling. Columns 4–6, however, show that chiefs political power increases titling, but only for autochthones.Footnote ³⁵ Among autochthones, a one standard deviation ( $ \sigma =0.52 $ ) increase in chiefs’ political authority is associated with a 0.13 percentage point increase in titling, an increase of 34% over the baseline rate. In contrast, a one standard deviation in political authority is associated with a 0.09 percentage point decrease in titling among allochthones, a decrease of 22% relative to the baseline rate of titling. These results paint a clear picture: in this devolved land regime, strong chiefs lead to more titles for the in-group (autochthones). However, in the Ivorian context, chiefly capture of land tenure administration means strong chiefs actually decrease titling rates among the outgroup (allochthones), in line with the chiefs’ agendas.Footnote ³⁶

Chiefs Capture Land Management and Exclude Allochthones

How do chiefs capture titling? One mechanism is that chiefs dominate the CVGFRs and fill them with co-ethnics. CVGFRs investigate land claims and issue dossiers for potential CFs (Bassett Reference Bassett2020). The national land bureau (AFOR) ultimately records and issues CFs, but important decisions are made locally. I asked the chiefs of each village (1) whether the village was mostly autochthonous or allochthonous and (2) whether the CVGFR was mostly autochthones or allochthones. Table 5 shows that even in villages which are equally split or mostly allochthonous, autochthones dominate the CVGFRs.

Table 5.

Chiefs Staff Village Land Management Committees with Autochthones

Note: Data from a 2024 survey of customary elites in the Haut-Sassandra and Indénié-Djuablin regions of Côte d’Ivoire.

Chiefs may also unevenly enforce land titles. If titles are less useful where chiefs are more powerful, it further suggests that chiefs can capture and control the titling process. In other words, strong chiefs may be able to overrule or override land titles. Table 6 shows whether respondents agree that: (1) having a title helps you keep your land if the government wants to take it; (2) having a title helps you be compensated if the government does take your land, and (3) having a title is useful in case of a dispute against your peers. If chiefs enforce titles unevenly, we would expect to see a negative coefficient on the interaction between allochthones and chiefly authority, because the chief’s agenda is to discriminate against allochthones.

Table 6.

Autochthones More than Allochthones Think Strong Chiefs Make Titles More Useful

Note: Dependent variables are answers to “[d]o you think somebody with a certificat foncier would be (1-2) more likely to keep their land if the government attempted to take it; (3-4) to be compensated fairly for the land, were it taken” and (5-6) to succeed in a land dispute? All answers use a five-point Likert scale. Demographic controls include education, sex, age, ethnicity, the respondent’s relationship to the household head, and wealth. Geographic controls include an indicator for PAMOFOR, distance to department capital, cocoa suitability, coffee suitability, and terrain ruggedness. Table C8 in the Supplementary Material shows coefficients for these control variables. All regressions use OLS with inverse sampling probability weights. Standard errors are clustered at the administrative village level.

A clear divide emerges in Table 6 between the in-group (autochthones) and the out-groups (allochthones). Across all three categories, allochthones think titles are more useful. The chiefs’ authority has no relationship with the perception that titles are useful to be compensated for land expropriated by the government, which makes sense if households think such decisions are made at a higher political level. Across the remaining outcomes, a one standard deviation increase in chiefs’ traditional authority is associated with a 0.17–0.26 standard deviation decrease in the perceived utility of titles among autochthones. Among allochthones, the same increase in chief’s authority is associated with a larger 0.50–0.53 standard deviation decrease in the perceived usefulness of titles. All respondents think strong chiefs reduce the usefulness of titles, but this result is much stronger among allochthones, the out-group. Unlike previous results, it is chiefs’ customary authority, rather than their political authority that makes a difference. Land titles are less of a constraint on the behavior of more powerful chiefs which allows them to capture land management and discriminate against allochthones.

Chiefs are also able to extract rents from their control over titling. While chiefs have no official role in titling, 70% of the 801 respondents in my survey expected to have to pay chiefs to formalize a land title. Only 57% thought they would have to pay their CVGFR, which is actually empowered to title land and collect fees. More broadly, on average allochthones and autochthones thought they would have to pay 3.8 customary actors—actors without a de jure position in the titling process—in order to acquire a title.Footnote ³⁷ These statistics show that chiefs’ role in titling process not only buttresses their political authority, but allows them to extract rents.

An alternative explanation for the broader results is that households title less when they have confidence that the chief can protect their property rights. Such trust would decrease the marginal increase of perceived land tenure security from titling one’s land, and therefore would reduce uptake of land titling. In Table C5 in the Supplementary Material, however, I show that trust in one’s chief does not increase either household titling or the extent to which households perceive land titles as valuable. This result accords with other recent literature, which suggests that the legibility provided by land titles is a valuable resource regardless of title’s legal weight (Ferree et al. Reference Ferree, Honig, Lust and Phillips2023).

By holding land values and the national land regime constant while varying chiefs’ power at a highly granular level, this case study sheds light into how customary authorities constrain the land titling process. Strong chiefs facilitate titling within their village—but only for the chiefs’ in-group. Moreover, allochthones perceive titles to be less effective, suggesting that enforcement of land titles may be unequal. While the cross-national analysis reveals broad trends which are consistent with the theory, this case study shows that the observable implications of this theory in Côte d’Ivoire align with the intermediate steps of the broader theory.