Absentee tenants and farmland transfers in sub-Saharan Africa: evidence from Malawi

ABSTRACT We use a unique dataset from Malawi that matches tenants and their landlord counterparts to document the role played by absentee tenants, i.e. tenants who reside outside the area where the rented land is located. We found that non-local tenants made up 22% of the tenants in our sample. A significant subset of them had higher off-farm income and significantly more assets than did other tenants. Conversely, we found that 76% of landlords rented land because they needed cash. Our results highlight the fact that some rental transactions reinforce power imbalances and may exacerbate risks faced by poorer landlords.


Introduction
Land has always been a key factor of production for rural households in sub-Saharan Africa. However, land rental markets increasingly constitute an important conduit for rural people to obtain land for deriving food and incomes from farming Place 2009a, 2009b;Lunduka, Holden, and Øygard 2009;Deininger, Savastano, and Xia 2017). Deininger, Savastano, and Xia (2017) report the amount of land being rented annually for agricultural purposes in six countries; in five of these countries, the amount of rented land exceeds that of new cropland put under production annually over the 2010-2018 period according to FAOSTAT. 1 The economics literature on rural land rental markets has typically focused on the efficiency and equity outcomes of rental market participation observed over the short term (e.g. Jin and Jayne 2013;Chamberlin and Ricker-Gilbert 2016), with little attention paid to how land rentals may contribute to longer-term changes in land ownership, farm structure, entry, and exit from the farming sector, and longer-term household welfare impacts.
This study explores the role of land rental markets in providing a means for non-local tenants to gain access to land based on evidence from Malawi. Malawi is a relatively densely populated African country with limited surplus land (Ricker-Gilbert, Jumbe, and Chamberlain 2014), an active rental market (Chamberlin and Ricker-Gilbert 2016), and documented growth in "emergent" farmers, i.e. farm operations initiated by non-local investors, many of whom reside in urban areas and derive wealth from non-farm employment that is utilised for acquiring farmland Jayne et al. 2016). In particular, we inquire into the extent of "reverse tenancy" (i.e. situations in which poor landholders lease land to wealthier tenants) and "absentee tenants" (i.e. tenants who reside outside the area where the rented land is located) in local rental market activity. 2 We then explore what these characteristics may imply for our understanding of the role of rental markets in longer term patterns of agrarian change.
With these primary research objectives in mind, we address several related questions. First, are local rental markets characterised by power imbalances? Specifically, are wealthy non-local people renting in land from poorer landowners? If so, are they renting land as the first step in a transition to owning it later? Is there any evidence to suggest that smallholder farmers may be permanently parting with their land (either voluntarily or not) to tenants? In other words, do rental arrangements facilitate more permanent land transfers (and do smallholders landlords benefit from such arrangements)? These questions are extremely relevant to current academic and policy debates, in Malawi and elsewhere in SSA, about how changes in rural land institutionspolicies, tenurial norms, and marketsare contributing to longer term agricultural development strategies, such as inclusive agricultural growth and improvements in rural household welfare.
Data used in this study come from a unique dataset of matched tenant-landlord pairs collected in four districts of Malawi with active land rental markets during April and May 2016. One of the benefits of our dataset is that the landlord side of the land rental market is more fully observed, compared to other survey datasets from SSA where it is nearly always under-represented (see Deininger, Savastano, and Xia 2017;Ricker-Gilbert et al. 2019 for more details). The fact that we can observe both the supply and demand side of the land rental market allows us to more accurately identify the potential implications of non-local demand for rural land.
Our analysis begins with descriptive statistics about the nature of land rental arrangements and the duration and types of contracts between tenants and landlords. The present article builds upon Ricker-Gilbert et al. (2019) and Ricker-Gilbert, Chamberlin, and Kanyamuka (forthcoming), who use the same dataset to explore other research questions. We start by disaggregating descriptive statistics based on the type of tenant that is engaged in contracts with landlords. These include (i) local tenants who are mainly farmers, (iii) local tenants who are mainly engaged in non-farm activities, and (iii) non-local tenants. We also disaggregate these groups of tenants based on levels of asset wealth. Building upon the descriptive statistics we estimate a set of regression models to understand the factors that are associated with (i) being a tenant who plans to acquire the rented plot from the landlord in the future, (ii) prices paid per hectare for a rented plot, (iii) characteristics associated with being a landlord who rents out land because of short-term cash needs (e.g. "stress renting" or distress rental). All of these models should further inform the discussion on the absentee tenant and landlord dynamics in Malawi and elsewhere in SSA.
Malawi is a country where land is a hot political and social topic: land disputes are routinely reported in the popular press, and land policy debates are the subject of intense public scrutiny. The 2016 Customary Land Act 3 formalised rights of the holders of customary land, enabling registration of land and theoretically facilitating land collateralisation and exchange of landholdings via rental and sales markets. While this has been championed by some as a means of strengthening and extending the customary usufruct rights of smallholders, concerns have also been raised about potentially negative impacts on smallholder welfare. For example, some traditional leaders in Malawi have expressed concerns that the 2016 Land Act would perpetuate loss of clan land by legitimising individual selling "which might lack the checks and balances of the current traditional system" (Kampanje-Phiri et al. 2019, 22) which help to prevent "reckless" selling to outsiders. "Our fear is that titles may give absolute ownership to individuals thereby giving people a license to sell land. If this happens, then the poor will lose land" (from focus group discussion with Khombedza Chiefs, 21 September 2018, reported in Kampanje-Phiri et al. 2019, 18). In other words, by facilitating rental and sales transactions, the recently passed Customary Land Act is frequently alleged to encourage poor smallholders to engage in land transactions that may provide short-term gains at the expense of becoming more vulnerable to eventually losing their land. Related concerns have been raised that the new land laws weaken the voice and power of traditional authorities and local clans over the lands in their areas.
Similar concerns have been articulated by scholars of changing land institutions in Africa: a relatively large body of qualitative literature has described "commodification" of land in SSA as fundamentally disadvantageous for smallholders (Amanor 2012(Amanor , 2018Yaro, Teye, and Torvikey 2017;Chitonge et al., 2017;Kumeh and Omulo 2019). The main tenets of the argument are that commercialisation favours those who are powerful (wealthier individuals, commercial investors) which leads to smallholder displacement and marginalisation; increased inequality; decreased social cohesion; erosion of traditional social norms and values, as customary rules are replaced with market rules; failure to deliver jobs and other measures of local economic opportunity. Of these, the first is the most central to our motivation. However, the specific ways in which land commodification leads to displacement are not always clear. While anecdotal accounts of displacement occurring through large-scale transfers or concessions to commercial operations, these accounts are often used to make very general claims about land commodification and displacement that may not apply to (the likely much more common) contexts of small-scale land rental markets. In other words, it is not at all clear whether small-scale rental transactionswhich are rising in importance in Africa and have generally been considered to enhance both equity and land use efficiency within smallholder communitiesmay also contribute to smallholder land dispossession.
However, we do know that land acquisitions by non-local individuals are taking place and are rapidly reshaping the region's farming systems (Jayne et al. , 2019. These "emergent farmers", often operating between 5-100 hectares, make up increasingly important shares of total national production in many countries compared to 15-20 years ago (Jayne et al. 2019), although they are likely under-represented in nationally representative rural household surveys (ibid). Several studies indicate that many of these medium-scale farmers were initially engaged in nonfarm activities, but used their savings to acquire land and start investing . In land abundant countries, like Zambia and Tanzania, land acquisitions are likely to operate through formal or informal sales markets; in land-scare countries, like Malawi, the scope for such transfers is less clear. We know that the emergent farmer sector is growing in Malawi ). The question is what role are rental markets playing in this growth?
The vast quantitative rental market literature has been silent on this question so far. The focus of this literature has been on documenting rural land market growth (Holden, Otsuka and Place 2009a;Deininger et al. 2017), and identifying short-run impacts on efficiency and equity outcomes (Holden, Otsuka and Place 2009a;de Janvry and Sadoulet 2011;Jin and Jayne 2013;Chamberlin and Ricker-Gilbert 2016). With respect to market growth, case studies of different areas in SSA generally indicate rapidly rising land market participation rates, particularly in densely populated areas Holden, Otsuka, and Place 2009a; de Janvry and Sadoulet 2011; Jin and Jayne 2013; Chamberlin and RickerGilbert 2016. There is corresponding evidence that land rental prices are rising rapidly in many parts of SSA, particularly in response to energy and food price increases, and in locations close to urban centres, where land is relatively scarce (Wineman and Jayne 2018;Tione and Holden 2020;Abay, Chamberlin, and Berhane 2021), all of which are consistent with increasing demand from urban investors. Many studies have found evidence of increased efficiency in land and labour use by transferring land to more efficient producers and those with greater labour endowments (Jin and Jayne 2013;Chamberlin and Ricker-Gilbert 2016). Equity gains from land market expansion have been associated with net transfers of land from those with more land ex ante to those with less land ex ante (ibid.). However, these studies have either used cross-sectional data or panel data observations spanning four to six years to evaluate relatively short run effects. Little attention has been paid to the role of rental markets in longer-term growth trajectories of farmers, or the entry/exit of individuals from the farm sector. That said, there is considerable descriptive evidence that landlords are often substantially poorer than their tenants in non-land assets (Bellemare 2009(Bellemare , 2012Holden, Otsuka, and Place 2009a;Chamberlin and Ricker-Gilbert 2016;Ricker-Gilbert et al. 2019, forthcoming)a situation sometimes referred to as reverse tenancy, where poor landlords rent to wealthier tenantsand there is some evidence of distress rentals, i.e. renting out land to meet short-term consumption needs at the expense of long-term welfare (Gebregziabher and Holden 2011). While rental contracts in the region are largely short term in nature, if large power imbalances characterise rental transactions, and rental transactions may increase the odds of disputes where such power imbalances may influence outcomes, then it becomes natural to question the role of rental exchanges in the longer-term aspirations of farmers (either to enter or exit farming) and whether or not such exchanges may lead to permanent transfers of land.
The contribution of the present study is to explore the linkage between these disparate areas of inquiry, exploring the previously unexamined role of rural rental markets in the on-going processes of farm structure change and the attendant longer-term strategies of entry and exit from the farm sector. Our results indicated that non-local tenants made up about 22% of the tenants in our sample and there was a sub-set of relatively wealthy non-local tenants who had off-farm incomes that were significantly higher than the local tenants who are mainly engaged in farming. Non-local tenants were less likely to be related to their landlords than were local farming tenants and non-local tenants were more likely to have written contracts with their landlords than were local farming tenants. However, wealthy non-local tenants were less likely than wealthy local tenants to want to acquire the rented plot and they did not pay higher or lower prices for renting land. Wealthy local tenants were about 26 percentage points more likely to want to acquire their rented plots than were poor local tenants on average. On the landlord-side of the market, we found that nearly 76% of households rented out land because they needed cash, and that landlords who rented out land for that reason were less likely to be related to the village chief and had lower savings than landlords who rented out land for other reasons.

Land rental markets in Malawi
Malawi has two major land tenure regimes: customary land and private land. Land rentals are found within both tenure systems, while sales are mostly confined to private land that is titled. Although mechanisms for transferring land from customary to private status are defined by the Malawi Land Act, and the number of such transfers has grown in recent years, the vast majority of smallholders still operate farms under customary tenure (Chirwa 2008).
Under the private tenure system, titled lands may be bought and sold without restriction. Conversely, under customary tenure, usufruct rights to land resources within a chiefdom are allocated by the chief or their representatives (i.e. local headmen or indunas) and are generally granted primarily to clan members, although the chief may extend such grants to outsiders. Customary usufruct rights are generally heritable, but are also subject to potential reallocation by the chief. The buying and selling of customary land has no legal basis, although it may be carried out clandestinely (often invoking traditional mechanisms for transferring permanent usufruct rights via the chief). 4 The renting of customary landholdings, on the other hand, is something that the chief may endorse (often tacitly rather than explicitly), although it is generally also not codified under customary tenure rules. Fixed cash rental arrangements are the dominant contractual form when smallholders rent in and rent out land in Malawi (Holden, Kaarhus, and Lunduka 2006). These contractual arrangements are usually undocumented with no written contract.

Emerging farmers in Malawi
In recent years, emerging farmers have become a major force within Malawi's farming sector. In a recent study of three districts, Anseeuw et al. (2016) found that the area of land acquired by emerging farmers in these three districts nearly doubled between 2000 and 2015. Data on farm histories from the same study revealed that about half of these emerging farmers grew from smallholder roots; the other half began as urban-based professionals who acquired land for farming later in life. Most of these acquisitions involved initial transfers of customary land (>70%), although about a third of the customary lands acquired by such investors was later transferred out of customary tenure via titling (ibid). While it is not clear how representative these numbers are for the country as a whole, they are indicative of processes that may well be very widespread. One question that remains poorly understood to date is how much of the land acquired by emerging farmers initially came from renting in land from smallholders, and was later purchased or otherwise permanently acquired by the emerging tenants. 5

Data
Our analysis is based on data collected by the Lilongwe University of Agriculture and Natural Resources (LUANAR) through the Center for Agricultural Research and Development (CARD) in collaboration with the International Maize and Wheat Improvement Center (CIMMYT) and Purdue University. We collected data in four districts with relatively high levels of land rental market activity: Lilongwe, Salima, and Nkhotakota in the Central region, and Zomba in the Southern region. 6 Nkhotakota and Salima were selected to represent rural areas, while Lilongwe and Zomba represented peri-urban areas. Our sample of 150 farm households in each district (50 tenant, landlord, and autarkic households, respectively) gave us a total of total sample size of 600 households. Within each district, we worked with the District Agricultural Development Officer (DADO) to identify Extension Planning Areas (EPA) with high rental market activities. Within each of the selected EPAs we randomly selected one village.
Within each selected village, we identified smallholder farm households who participated in the land rental market through a Focus Group Discussion (FGD) with the Village Headman, Lead Farmers, and members of both the Village Development Committee (VDC) and the Vulnerability Assessment Committee (VAC). These key informants described village history and an account of land issues, including land rental activities in the village. On average, the FGD comprised about 10 individuals of which 50 % were women and community-level issues regarding landownership, land availability and use, drivers of land renting, and prevailing farm gate prices of cash crops were discussed. At the end of each FGD, we then randomly sampled individual farming households from the village list of all households. The list served as a sampling frame for our survey. Households involved in renting in land (tenants) or renting out land (landlord), and those that neither rent in nor rent out land (autarkic) during the 2015/2016 season were sampled for the interviews. Each sampled landlord was matched to his or her tenant pair during the interview. Thus, if a tenant household was sampled, its corresponding landlord was automatically sampled for the interview and vice versa. This process was repeated until a sample size of 10 matched pairs was reached (i.e. 20 households) in each village. Furthermore, 10 autarkic households in each village were randomly selected from the list as control households. Thus, a total of 30 households were sampled per village. We attempted to track any member of the tenant-landlord pair who resided outside the village, although most pairs ended up being local. Nevertheless, we oversampled, especially in larger villages, to account for nonresponses and both tenant and landlord absenteeism. In each district, five villages were selected with 30 farming households per village adding up to 150 households in five villages at the district level.
A key variable in this analysis is the designation of a tenant as local or non-local. Landlords define tenants as local or not depending on whether or not they come from the same village as the landlord. Another key categorisation is whether or not the landlord engages in "stress renting"i.e. renting out land to satisfy immediate cash requirements, possibly under rental arrangements which would not be optimal in the absence of such short-term cash needs. This designation is based on the landlord's response to the question "Why did you rent out this parcel?" being "needed cash/money".

Methods
Our analysis focuses on the three main research questions related to tenant and landlord dynamics in Malawi, as discussed in the article's introduction. We start with a descriptive analysis comparing tenants and their landlord pairs across the wealth distribution and break the sample down by tenants who have greater than the 50th percentile of the asset wealth distribution (which means they have the equivalent of greater than US $ 198) and those with below the 50th percentile.
After the descriptive analysis, we estimate three linear probability models (LPM) to better understand the factors associated with different types of tenants and landlords. The first model assesses the factors associated with whether or a tenant (i) intends to acquire the plot that he or she currently rents in as follows: where A = 1 if the tenant says that he or she eventually plans to acquire the rented plot through purchase or other means, and zero otherwise. The variable L = 1 if the tenant is rich and local. Rich is defined as being above the 50th percentile of the asset distribution of the tenants in the sample, while local is defined as the tenant residing in the same community as the landlord and the rented plot (as defined by the landlord). The corresponding parameter to estimate is defined by b 1 . The variable N = 1 if the tenant is poor and non-local, defined as having assets below the 50th percentile of the asset distribution of the tenants in the sample and not residing in the same community as the landlord and the rented land, according to the landlord. The parameter to estimate is b 2 . The variable R = 1 if the tenant is rich and non-local with b 3 as the parameter to estimate. The coefficient estimates onb 1 -b 3 are compared to an excluded category of tenants, those that are poor and local. We also compare the three coefficients to each other to test if the different categories of tenants have a different propensity to acquire land. This allows us to test the hypothesis of whether or not rich non-local tenants are more likely to acquire the rented land than are other tenants in the sample. Tenant household social demographics such as (i) relationship to the chief, (ii) number of members in the household, (iii) if the head is female, (iv) age of the household head, and (v) education of the household head are denoted by the vector D, with b 4 as corresponding parameter vector to estimate. Household wealth proxies include pre-rental land ownership, and if a household reported that they were in need of credit are all denoted by the vector W, with b 5 as the vector to estimate. The tenant-specific error term is denoted by v.
The second equation investigates the factors associated with the price per hectare that a tenant pays for the rented plot: where P represents the log of price per hectare that the tenant pays for the rented plot. The other variables are the same in Equation (2) as they are in Equation (1) with a 1 − a 5 as parameters to estimate. The coefficient estimates onâ 1 -â 3 tests the hypothesis of whether or not rich non-local tenants pay a different price for rented land than do other tenants. The tenant-specific error term is denoted by e. Finally, we estimate an equation for landlord l, to understand factors associated with his or her decision to rent his or her plot because of the need for cash, a characteristic of "stress-renting" as follows: where S = 1 if the landlord states that he or she rents land because they need cash and zero otherwise. The variables D and W are similar as in Equations (1 and 2), but D includes whether or not the landlord was born in the village where the rented land is located and W includes household savings from all sources; d 1 − d 2 are parameters to estimate. The landlord-specific error term is denoted by 1. Standard errors in Equations (1-3) are made robust to heteroskedasticity. This analysis is mainly descriptive in nature so we do not claim causality in our results. Nonetheless, the results from these models generate important information about absentee tenants and their landlords. Figure 1 shows that most tenants are considered to be locals with farming as their main occupation (57%) according to their landlords, while 21% are locals engaged in non-farming activities as their main occupation. Finally, 22% of the tenant sample are non-local tenants, primarily residing in different communities from their landlord. Figure 2 shows that when looking at the annual revenue in 2015/2016 of these different types of tenants, the local tenants who were mainly farmers had a mean revenue of US $651, while local tenants who were mainly engaged in nonfarm activities had nearly double the revenue of $1207. Non-local tenants had an average income that was a bit lower than local non-farming tenants, but much higher than local farmers, with an average revenue of $1086. 7 Figure 3 further disaggregates the three categories of farmers by their standing on the tenant wealth distribution. We see that for those tenants above the 50th percentile of the asset distribution (meaning they had more than US $198 in 2015/2016), the local tenants who mainly engaged in nonfarm activities and non-local tenants had significantly higher non-farm revenues than local farmers. This suggests that activities outside of farming are a major source of revenue for these households. It is also consistent with the presence of relatively wealthy people who have gained land through non-  farm means coming in from outside rural communities, as found in the literature on emerging farmers (Jayne et al. , 2019. Figure 4 shows the percentage of different types of tenants that say they plan to purchase the land that they currently rent from their landlords. It is interesting to note that wealthy local tenants who are farmers and wealthy local tenants who are mainly engaged in non-farm activities are significantly more likely to plan to acquire rented land compared to poor farmers in these categories and compared to non-local farmers, both rich and poor. This suggests that the determinant of desire to purchase land is dependent on being located close to the rented plot and also the wealth  of the tenant. Still, it is interesting to note that >50% of non-local tenants still planned to purchase the plot that they were renting from their landlord in 2015/2016. Figure 5 shows the main reasons why rich non-local tenants rented in land during the 2015/2016 season compared to all other types of tenants. By far the main reason for renting-in land by all types of tenants was as a way to acquire land for farming. However, 80% of rich non-local tenants gave this response, while 73% of other tenants gave this response as their main reason for renting. Relatedly 7% of rich non-local tenants said that the main reason they rented was to expand the area that they farmed, while 18% of other tenants cite expansion as the main reason. This may suggest that a slightly higher percentage of rich non-local tenants are using land as an initial pathway into farming, while a higher percentage of other types of tenants are already engaged in farming and use renting to expand their area cultivated. This finding is also consistent with the emerging farmer literature.

Results
One relevant question that comes from the results shown in Figure 5 is what do tenants do with the land they rent in. The data from across all tenant managed plots suggest that maize is the most widely grown crop, as expected (Appendix Table A1). However, Hybrid maize is 10 percentage points more likely to be grown on a tenant's owned plot than on his or her rented plot. Conversely, tenants are about 13 percentage points more likely to grow groundnuts on their rented plots. This suggests that tenants generally are more likely to use their owned plots to secure their maize supply and use the rented plots to expand area cultivated into cash crops like groundnuts. This also interesting and surprising from a soil fertility perspective because maize depletes nitrogen from the soil while groundnuts are nitrogen loading. Table 1 shows the land rental arrangements for the different types of tenants, disaggregated by asset wealth. Row (ii) shows that non-local tenants both poor and rich, are nearly three times less likely to be related to their landlord than are local tenants, though most are of the same ethnicity (row iii). Consistent with this, we see in row (iv) that non-local tenants are much more likely to have written rental contracts with their landlords compared to local tenants. Interestingly, rows (v) and (vi) show that across the board all rental arrangements are relatively short-term, less than two years on average. Conversely, all types of tenants plan to continue renting the plot for a meaningful amount of time in the future (on average). In fact, the average tenant says that he or she plans to rent the land for nearly five additional seasons (row vii). Table 2 presents a simple t-test of mean differences to identify the characteristics that are differentiating between (i) wealthy, non-local tenants; (ii) other types of tenants as defined earlier; and (iii) landlords. First, wealthy non-local tenants have two more members in their households on average than other tenants or landlords (seven vs five members), and this different is statistically significant at the 1% level. This suggests that these wealthy, non-local tenants have more labour available to potentially work on farm and they have more mouths to feed. In addition, the average wealthy non-local tenant has 10.44 years of education compared to 7.53 for all other tenants and 4.75 for landlords (means different at 1% level). This is consistent with the notion of wealthier, more  educated elites, coming into communities to rent land. Wealthy, non-local tenants also have much higher cash savings and value of durable assets, both at the household level and on a per-capita basis than do other types of tenants and landlords. Table 3 presents the reasons why landlords rented out their plot during the 2015/2016 season. The main reason given by nearly 76% of landlords was the need for cash at planting time. This is strong suggestive evidence that most landlords are "stress renting" their land rather than engaging and investing in other activities. Lack of labour is the second most common reason given by greater than 16% of respondents, which is also evidence of constraints preventing landlords from cultivating their plots. Only 2.33% of respondent landlords said they rented out their plots because they wanted to engage in more profitable activities. This suggests that there is little to no evidence that renting out land is acting as a pathway to a better livelihood for landlords in this sample ( Figure 6). Table 4 investigates possible power dynamics between non-local tenants and their landlords by comparing the demographics of the landlords who rent to wealthy non-local tenants compared to demographics of landlords who rent to other types of tenants. The t-test of means suggests that there is some evidence to support the notion that wealthy non-local tenants rent from landlords who need cash and may be stress renting their land. The first row of the table suggests that wealthy non-local tenants are 23 percentage points more likely to rent from landlords who needed cash (p-value = 0.05). This may provide some suggestive evidence of exploitation by nonlocal tenants. However, their landlords are no more likely to need credit, or have significantly lower savings or have significantly less land than other landlords. The only other factor that is statistically significant in Table 4 is the fact that female landlords are 22 percentage points less likely to rent to wealthy non-local tenants. Potentially weak tenure security and lower level of financial resources of female landlords compared to rich non-local tenants may be a reason why women are hesitant to rent out their land to these outsiders.   Table 5 shows the regression results for the model presented in Equation (1) for the factors affecting whether or not the tenant says he or she intended to acquire the rented plot from his or her landlord. The consistent result across the table is that rich-local tenants who have more assets than the sample median (> US $198) and reside in the same community as the landlords are about 26 percentage points more likely to purchase the rented plots than poor local tenants.  Notes: Model estimated via linear probability (LPM); *, **, *** indicate that the corresponding means are different from each other at the 10%, 5%, and 1% levels, respectively; p-values in parentheses; models include constant and district fixed effects; standard errors robust to heteroscedasticity; rich tenants have assets great than the 50th percentile of tenant asset distribution in the sample; non-local tenants do not reside in the same community where the land is rented, according to the landlord; 12 of the 173 tenants in the sample did not answer the intention to acquire land question so were not included in this analysis.
Conversely, rich non-local tenants are not more likely than poor local tenants to want to acquire the rented plot. In addition, the f-tests at the bottom of Table 1 indicate that they are between 29 and 33 percentage points less likely to want to acquire the rented plot than are rich local tenants and are no more likely to want to acquire the rented plot than are poor non-local tenants. Table 6 shows the regression results for the model presented in Equation (2), estimating factors affecting the log of land rental prices per hectare. Results indicate that there is no statistically differential effect on rental prices paid by different types of tenants in the sample. The main factors that had a statistically significant association with land rental prices were number of tenant household members and age of the tenant household head in years, both of which were positive. Table 7 shows the regression results for the model presented in Equation (3), estimating the factors that affect the probability that a landlord rents out his or her land because he or she needs cash. Results from the table show that landlords who were relatives of the local chief in the area where the plot was rented were 11 percentage points less likely to rent out the plot because he or she needed cash on average (p-value = 0.084 in column 3). This could suggest that the chief's kin are protected in hard times by their family network or they are wealthier to begin with. Related to this, and as un surprisingly, households who have more cash savings were significantly less likely to rent out land due to the need for cash. This finding is consistent with the idea of distress rentals and sales which have been documented in recent empirical literature (Gebregziabher and Holden 2011;Kusunose and Lybbert 2011;Ghebru and Holden 2014;Ricker-Gilbert et al. 2019). The fact that landlords part with their largest asset when they need cash raises questions about price used for tenants with more than one rented-in plot; *, **, *** indicate that the corresponding means are different from each other at the 10%, 5%, and 1% levels, respectively; p-values in parentheses; models include constant and district fixed effects; standard errors robust to heteroscedasticity; rich tenants have assets great than the 50th percentile of tenant asset distribution in the sample; non-local tenants do not reside in the same community where the land is rented, according to the landlord; 19 of the 173 tenants in the sample did not answer the land rental price question so were not included in this analysis.
how landlords are trading long-term welfare for short-term cash income. This is particularly true in the context of this study as we found earlier evidence that a significant percentage of wealthy tenants plant to continue renting-in land beyond the current season and/or to eventually buy it.

Conclusions
Our goal in this article has been to describe and estimate the role played by absentee (non-local) tenants in rural land rental markets operating in customary tenure areas, using a unique dataset of matched tenants and landlords from Malawi collected in 2015/2016. In doing so, we linked to recent literature showing the prevalence of reverse-tenancywhere tenants are wealthier than their landlords in rental markets in SSA (Bellemare 2009(Bellemare , 2012Jin and Jayne 2013;Chamberlin and Ricker-Gilbert 2016;Ricker-Gilbert et al. 2019), as well as the emerging literature on emerging farmers acquiring land for entry into farming or expansion of existing lands Jayne et al. 2019). We also sought to understand the relationship between non-local tenants and their landlords. Non-local tenants made up about 22% of the tenants in our sample. There was a majority of relatively wealthy non-local tenants who had higher off-farm income and significantly more assets than did the majority of local tenants who were mainly engaged in farming. Non-local tenants had higher education than other tenants, and were less likely to be related to their landlords than were local farming tenants. Non-local tenants were more likely to have written contracts with their landlords than local farming tenants. The rich non-local tenants were also 23 percentage points more likely to rent from a landlord who rented out land to meet needs for cash, consistent with distress renting (Gebregziabher and Holden 2011;Kusunose and Lybbert 2011;Ghebru and Holden 2014;Ricker-Gilbert et al. 2019). This suggests some possibility for exploitation in the rental relationship. In addition, 80% of rich non-local tenants said that they rented land as a means to acquire land for farming. This is consistent with the literature on emerging farmers coming from outside the community to acquire land Jayne et al. 2016Jayne et al. , 2019. One of the contributions of this article has been to provide evidence that rental markets may be an important initial stage in eventual ownership transfers to such outsiders. However, while more than 50% of wealthy non-local tenants said that they wanted to eventually acquire the plot that they were renting, such longer-term ownership objectives were even more prevalent among wealthy tenants who were considered to be local by their landlords (i.e. residing in the same community): 83% of the wealthy local tenants who mainly engage in non-farm activities, and 79% of wealthy local tenants who are mainly farmers wanted to eventually acquire permanent control of their rented plots. Regression results support this finding: wealthy non-local tenants were significantly less likely than wealthy local tenants to want to acquire the rented plot and they did not pay higher or lower prices for renting land. In addition, wealthy local tenants were about 26 percentage points more likely on average to want to acquire their rented plots than were poor local tenants, through purchase or other means.
This suggests that while wealthy non-locals are part of the population of people who potentially want to purchase land initially acquired through rental markets, most of the desire to eventually purchase rented plots comes from the local population of wealthy residents. This makes intuitive sense, as these local people are established in the community and can monitor the labourers who work their land at a lower cost, and the wealthy locals may have an easier time claiming to have the right to purchase the plot from the tenant if they are seen by local leaders to be actively engaged in its management. In addition, we found that tenants are more likely to use rented plots to grow cash crops like groundnuts than they do on their owned plot. Therefore, it could be that non-local tenants are trying to rent-in land to make income in the short term, rather than make longer-term purchases. This is consistent with large-scale commercial producers in the developed world who often rent-in land to avoid tying up the cash that would be needed to purchase it outright. For example, more than 50% of farmland in the United States is rented.
It is clear from our study that tenants are significantly wealthier than landlords along all dimensions besides landholding (see Table 2). It also appears that more than 75% of landlords rent out land because they need cash. As such, our findings also raise some potential questions about the role of land rental markets as a vehicle to promote pro-poor growth, as has been advanced previously in the literature (de Janvry and Sadoulet 2011). However, our results are not conclusive in this respect because we do not observe the livelihood situations of the landlords compared to their prerenting status. However, this issue is worth highlighting and deserves further attention in a context such as rural Malawi where land rights are informal and potentially subject to exploitation by wealthier and more powerful locals and outsiders.
We also found that landlords who rented out land because they needed cash were less likely to be related to the village chief (signalling weaker social capital) and had lower savings levels (signalling weaker financial capital) than landlords who rent out land for other reasons. These findings are consistent with those in Ricker-Gilbert et al. (2019) who document relatively weaker capital endowments of landlords relative to tenants, and argue that equity outcomes attributed to land rental transferswhich have often been evaluated solely in terms of the rental-mediated equilibration of household labour/land ratios when analysing the tenant side of the marketshould be considered more broadly by taking into account non-land assets.
Our results further support the need to understand the incentives and constraints that landlords face in their decision to lease out their most valuable asset. Additional research may help further clarify the relationship that poorer landlordsincluding those who are distress rentershave with tenants who are often much wealthier and potentially more influential in the community than they are (see wealth disparities between tenants and landlords in Table 2 and evidence of stress renting landlords in Table 6). While we can only speculate at this stage, given the limitations of available data, it is clear that power imbalances between tenants and landlords have the potential to contribute to ownership transfers at later stages. This is evidenced our finding that tenants are much more likely to want to continue renting or to permanently acquire the rented plots than their landlords are (Table 1). For example, if disputes over ownership arise (which are very common in Malawi, see Deininger, Savastano, and Xia 2017) litigants with relatively less power (in human, social, political, and financial capital) may face higher risks of unfavourable outcomes in traditional adjudication processes. In fact, it is possible that the 2016 Customary Land Act may have exacerbated this issue. Before the Act was passed, it was more difficult for tenants to possess land because they had to negotiate with the entire clan to acquire the land, not just the individual tenant. But the 2016 Land Laws vest land tenure rights with the nuclear household (cutting out the broader clan and local traditional authorities), making it easier for wealthy tenants to translate their relative power over a poor individual landlord into acquiring his/her land.
Our data suggest that non-local tenants are significant participants in Malawi's growing land rental market. Though we were limited by a relatively small sample size, we have identified that these actors exist and future research should work to obtain a more fully representative sample of these non-local tenants and survey them in more detail. As populations grow, food prices continue to rise, and land will inevitably become a more valuable commodity in SSA. This will naturally attract attention and interest in land from people inside and outside rural areas. Recent work has focused on understanding the longer-term aspirations of smallholder farmers in SSA (Verkaart, Mausch, and Harris 2018). However, this work has largely ignored the aspirations of people from outside local communities who acquire farmland. Our work indicates that this is a likely an important source of change in the composition of the farming sector. It is essential for policy makers to understand these issues to help ensure that customary tenure systems function in a way that ensures broadly shared agricultural growth and protection of individuals' access to land and livelihoods.