Determinants of land tenure security among small-holder farmers in rural Kenya: An ordered probit analysis

Abstract Land tenure security is key to rural transformation, diversification, increased agricultural production, and environmental sustainability. However, its determinants are still unclear, especially in Sub-Saharan Africa (SSA), where land tenure insecurity is widespread. The study evaluated the determinants of land tenure security using the ordered probit model. Using pre-tested semi-structured questionnaires, data were collected from 366 respondents in Narok county, Kenya using a multistage sampling procedure. From the results, household heads in the high land tenure security category were older, more educated, owned relatively larger land sizes, and had higher maize productivity with somewhat smaller household sizes than those in the low and medium land tenure security categories. Furthermore, age, marital status, education level, number of years the household had stayed on the land, land fertility, land acquisition through purchase, land size, the distance from the home to the parcel and household size were the main determinants of land tenure security. Pro-women and youth policies should be enacted to enhance land tenure security. Moreover, laws that protect marital property rights should be put in place and ease the land purchase procedure by reducing transaction costs and weeding out corruption.


Introduction
The last 50 years have seen land tenure reforms gain importance in the priority lists of governments and other development partners (Place, 2009). Today, land reforms, particularly land tenure security, are considered one of the main ways to increase agricultural productivity and food security among nations (Ghebru & Lambrecht, 2017). Additionally, it is one of the key contributors to a country's growth, development, social cohesion, and poverty alleviation, among other development indicators (Ayamga et al., 2015;Valkonen, 2021).
Land tenure security has also received attention in national and global development agendas. Among them are Kenya's Vision 2030, Sustainable Development Goals (SDG), and the United Nations, Paris Agreement. The Kenyan constitution introduced the National Land Policy to constitutionally recognize land matters, tenure system issues, land use management, and land administration. As a result of the policy, several land laws were legislated to protect the tenure rights of the citizens. These laws include the Government Land Act (Cap 280), Land Titles Act (282), Registration of Titles Act (281), Trust Land Act (Cap 288), and Registered Land Act (Cap 300) (Kieyah & Nyaga, 2010).
Before colonization, the communal land tenure system was the most commonly used in Kenya. Land tenure insecurity in Kenya started just after Kenya became a British colony when the land was converted to crown land (Kieyah & Nyaga, 2010). Local owners were forcefully displaced from their land and reallocated to the white settlers (Wakoko, 2014). After independence in 1963, all the land was under the presidency to undertake the resettlement of its citizens. However, the process was marred by corruption from politicians and local leaders (Khamisi, 2018).
The Kenyan constitution recognizes three types of tenure systems, namely, private (land rights held by an individual), communal or customary (land owned by a particular community as ancestral property), and public or state (the land occupied by schools, hospitals, forests, game reserves national parks) (Djurfeldt, 2020; Food and Agriculture Organization of the United Nations, 2017). Due to the increasing population, diminishing livelihoods, and pressure from the political class to formalize and privatize, the community land has been subdivided into plots held by private persons (Kyalo & Chiuri, 2010). The highlighted challenges informed the need to enhance land tenure security of farming communities to increase productivity, forming the basis for this study.
S. T. Holden et al. (2013) and Muchomba (2017) define land tenure security as the confidence that the individual rights over land will be recognized and protected by other people during an invasion by private investors, expropriation, and distribution by the government. Land tenure security dimensions include; de jure (tenure sanctioned by government power), de facto (tenure security which is common knowledge), and perceived tenure security (individual subjective evaluation of their tenure scenario) (Quizon, 2017;Van Gelder, 2010). If land tenure security is realized through clearly enforced rights, it can provide an incentive for investments, ease access to credit, and transfer of land, reducing poverty and spurring economic development (Bambio & Agha, 2018;Place, 2009;Valkonen, 2021). Land tenure security increases the participation of women in decision-making, thus giving them a voice in the decision-making processes (Muchomba, 2017;Mwesigye et al., 2020). Furthermore, it protects women's land from dispossession, especially in a patriarchal society (Owoo & Boakye-Yiadom, 2015). Land tenure security is considered a key ingredient in environmental protection and safeguarding natural resources (Robinson et al., 2018). In previous studies, Besley (1995) argues that land tenure security improves agricultural investment, increasing agricultural productivity.
Neo-classical theories have fronted the individualization of land tenure rights as a necessary condition for economic growth and development. There is a perception that the privatization of land rights can efficiently protect land interests. Against this backdrop, many studies have restrictively defined land tenure security using procedures from the formal land market, such as land titling. Land titling is defined as possessing a legal title deed or certificate over land (Lawry et al., 2017;Ma et al., 2016;Owoo & Boakye-Yiadom, 2015).
While acknowledging the significance of formal land documentation such as a title deed, the question of what measure should be used in areas that largely use the customary land tenure system should not be overlooked. The use of land titling as a proxy for land tenure security may have an insignificant effect in situations where informal and customary tenure systems exist (Arnot et al., 2011;Place & Otsuka, 2002). This study suggests that using several rights (rights to use and rights to transfer land) possessed by an individual may be an appropriate indicator for land tenure security. Therefore, this study argues that the more rights an individual possesses over a certain parcel of land, the higher the land tenure security. In addition, the drivers shaping the possession of these rights have not been adequately studied, especially in SSA countries. Few studies have examined land tenure security drivers (Ayamga et al., 2015;Ghebru & Lambrecht, 2017;Linkow, 2016). However, none has analyzed land tenure security as a categorical variable derived from the possession of land rights. Therefore, this study attempts to address the aforementioned knowledge gap by examining Kenya's land tenure security drivers.

Objective of the study
The main objective of this study is to analyse the determinants of land tenure security among smallholder farmers in Narok County, Kenya. The study adds to the body of knowledge by analysing the land tenure security variable as a categorical variable derived from possession of different rights over the land.
The paper is organized as follows; the first section provides a brief background of the study, the works of previous authors, the research gap to be addressed and the study's objective. In the second section, the methodology used in the study is discussed with a background of the study area and sampling procedure. The third section contains a discussion of the descriptive and econometric results. Finally, the conclusion and recommendations are provided in the fourth section.

Study area and sampling procedure
The study was conducted in Transmara East and Transmara West sub-counties, among the six subcounties of Narok county, Kenya ( Figure 1). The county lies between latitudes 0° 50"and 1° 50" and longitudes 35°28 " and 36° 25 " (County Government of Narok, 2018). The county has a population of about 1,057,873 persons (529,042 males and 528,805 females), making a gender ratio of approximately 1:1. In the two sub-counties of interest, the average farm size is 5 acres for the small-holder farmers and 20 acres for the large-scale farmers (County Government of Narok, 2018). Land ownership in Narok county is categorized into three (3); community, trust, and private land The study used a statistical formula by Fleiss et al. (2013) and adapted by Bolarinwa (2020) to obtain a sample size of 366 respondents. The sample size per sub-county was calculated based on the population size of small-holder maize farmers in the sub-county resulting to 187 and 179 respondents in Transmara West and Transmara East sub-counties, respectively. The study used a multistage sampling procedure to choose the study respondents. The first stage involved the purposive selection of Narok county due to the frequent land-related conflicts experienced in the area (Kariuki et al., 2016). Transmara East and Transmara West were purposively selected in the second stage because they are the main small-holder maize-producing counties with similar climatic conditions but different socio-economic, institutional, and land-related characteristics (County Government of Narok, 2018). The third stage involved the purposive selection of two wards in each sub-county which are the main maize producers and have the most reported incidence of land tenure security in Narok county. These wards are Mogondo and Ololmasani in Transmara East sub-county and Kilgoris Central and Angata Barikoi in Transmara West sub-county. The last stage used systematic random sampling to select every 5 th person on an alphabetically arranged list of 2000 eligible small-holder maize farmers obtained from the ward agricultural office of the county government of Narok. The sample size per ward was based on proportionate to the size of the small-holder maize farmers in the respective ward. A pilot study was conducted on 10% of the total sample size, and the results were used to computed the reliability of the data. The Cronbach's alpha of 0.721 implied that the data was reliable. Data on socio-economic, institutional, and landrelated characteristics were collected using semi-structured questionnaires installed in Open Data Kit (ODK) data collection software. The data were then coded and analyzed using Stata 15 computer statistical software (Stata Corp, 2014). The study was only limited by the low level of education among the respondents however, thorough probing during data collection exercise was done.

Measurement of the land tenure security variable
The study used land tenure security as the dependent variable, while independent variables included socio-economic, institutional, and land-related factors. Land tenure security can be measured in different ways. Carter and Olinto (2003) measured land tenure security (LTS) using the land tenure system associated with the land: private, communal, or public. However, according to Cattaneo, the years the household is expected to reside on the land before eviction can also be used to measure LTS. On the other hand, S. Holden and Yohannes (2002) used a binary variable that is secure = 1 or insecure = 0, while Place and Otsuka (2002), in their study, measured LTS using the method of acquiring the land. They, however, acknowledge that; the measure is not explicit enough to capture all the aspects. Rao et al. (2016) used the probability of eviction to measure land tenure security, while Owoo and Boakye-Yiadom (2015) and Ma et al. (2016) used a title deed or land certificate ownership to measure land tenure security. The finding backs this measure that a strong correlation exists between possession of a legal title and land tenure security (Brasselle et al., 2002). However, possessing a title deed may not necessarily mean having the security of tenure over land (Place & Otsuka, 2000). Illegal squatters in government land have relatively secure land tenure since they face a lower probability of evictions, yet they lack legal title over the land. Arnot et al. (2011) state that when governments are unstable, possessing a legal title may not mean anything.
Due to these inefficiencies in the singular measures of land tenure security, the study employed a composite measure consisting of various rights over the land. However, this measure may be location specific thus it may only be used in areas with similar or close socio-economic characteristics. Security land tenure can be assessed using three dimensions: user rights, transfer rights, and the autonomy given to the holders of rights, specifically the transfer rights (Brasselle et al., 2002). To achieve this purpose, ten different types of rights were defined based on the two broad categories (right to use and transfer). Unlike Place and Otsuka (2002), who assigned equal weight to all the rights, this study adopted Brasselle et al. (2002) approach that appreciates each right's different weights. In this approach, households were asked whether they have permanent, transitory (temporary), or none of the user rights.
The user rights cover the following; (i) choice of crop to grow, (ii) land fallowing and cultivation at the end of the fallow period, (iii) making land developments, (iv) disposal of crop produce, and (v) prevent people's livestock from grazing on the land. Additionally, respondents were asked if they required any approval to enjoy the following rights of land transfer (vi) give land along customary lines, (vii) transfer land as an inheritance, (viii) lease land in exchange for cash, (x) sell land, and (xi) mortgage the land. All these elements capture the existing concerns relating to the land LTS that may affect the investment. Three (3) categories, low LTS, medium LTS, and high LTS, were derived from the data to distinguish the different levels of land tenure security, making the LTS a categorical variable. This approach eases the interpretation of results and simultaneously captures the different roles of LTS alternatives for rural dwellers (Ajefu & Abiona, 2020; Brasselle et al., 2002). The study also adopted with amendments the categorization criterion of the land tenure variable as used by Brasselle et al. (2002). Table 1 presents the frequency distribution results of the respondents' various land rights (to use and transfer) in the study area. Results indicated that, among the sampled households, choice of crops to grow, land development, and the right to prevent grazing were the most common rights; hence difficult to be used to create categories. Furthermore, few respondents held less than one right (permanent or transitory). The most prevalent right in the transfer rights was the inheritance right, followed by the right to lease land, the right to give land along customary lines, the right to sell the land, and lastly right to mortgage. Thus user rights (ii and iv) and all the transfer rights were used to create categories since they exhibited sufficient variations in the results.
Category 1 (Low LTS) if they did not hold any transfer rights and didn't hold more than two user rights (whether permanent or transitory) or did not hold the latter two rights (or one of them) or had at least two permanent or transitory user rights in addition to rights (i), (iii), and (v). Category 2 (Medium LTS) if apart from the rights of inheritance and to give land along the customary line, they held the right to lease land (with or without approval) and at least two permanent user rights in addition to rights (i), (iii) and, (v). Category 3 (High LTS) if apart from the rights to inherit, to give land along the customary line, to rent or lease land, to sell land, they had a mortgage right on the land (with or without approval) and at least two permanent rights of use in addition to rights (i), (iii) and, (v).
The categorical nature of the dependent variable (land tenure security) necessitates using the ordered probit model, which estimates the parameters of the underlying distribution instead of the response itself (Daykin & Moffatt, 2002). The model can analyse ordinal multinomial dependant variables. There is a latent continuous metric underlying the ordinal responses (Wollni et al., 2010). The structural form of the model is as presented in equation 1; where Y � denotes the variable that indicates the level of land tenure security, which is in an ordinal form coded as 1, 2, 3, … K. Xis a vector of the various explanatory variables (demographic, institutional, and land-related factors) and ε represents the error term or the unobserved factors that may determine land tenure security. Intervals are as shown in equation 2; where δthe threshold or cut-off parameter to be estimated β β lacks intercept terms as it is normalized to zero to allow the parameters to be "free" parameters (Daykin & Moffatt, 2002). As suggested by Daykin and Moffatt (2002) and Wollni et al. (2010), it is assumed that εit is normally distributed, and the probabilities of the observed ordinal responses are given by equation 3; whereφ is the standard normal cumulative distribution function. Log-likelihood function is used to estimate parameters δand β as expressed in equation 4; Since the coefficients from the ordered probit do not provide precise interpretation, marginal effects were obtained (Mallick & Rafi, 2010). Authors such as Al Mamun (2018), Bezabih et al. (2011), Dokken (2015), and Mallick and Rafi (2010), used ordered probit to obtain unbiased and consistent results in their research. The socio-economic, institutional, and land-related characteristics used as explanatory variables were adopted from previous studies (Ali et al., 2011;Lambin & Meyfroidt, 2011;Linkow, 2016;Lovo, 2016;S. T. Holden & Ghebru, 2016). These variables include; age, sex, education level, and marital status of the key decision-maker on farming matters, land size, adult equivalent, land concentration, years of cultivation, fertility of the land, market access, road access, leadership, land acquisition, land dispute, and group membership as presented in Table 2. Table 3 presents descriptive and inferential results of continuous variables used to analyse the determinants of land tenure security. The results show that, overall, the mean age of the household heads was 43.63 years. Household heads belonging to the low land tenure security category were statistically significantly younger (41.06 years) at a 1% level than those in the medium land tenure security category (41.92 years) and high land tenure security category (45.62 years). This finding indicates that there was a significant difference in terms of age among the three categories of land tenure security. Low tenure security was associated with young people underscoring the importance of experience, knowledge and skills in securing land which is associated with older people.

Descriptive and inferential statistics
Overall, the average years of schooling of the sampled household heads was 9.47. The results also show a statistically significant difference at 1% in terms of education level among the three categories of land tenure security. Household heads belonging to the high land tenure security category were more educated, with a mean schooling of 10.06 compared to 8.75 years for low and medium land tenure security categories. As suggested by Rao et al. (2020), better-educated individuals tend to be more updated with information; hence they can secure their land more quickly than their less-educated counterparts.
The overall mean household size for the sampled household was 4.54, with results indicating that households in the low land tenure security category having significantly more household members (4.90) than those in medium and high land tenure security categories at 4.42 and 4.32, respectively. The household size tends to determine household resource utilization, and therefore households with more members are more likely to allocate resources to pressing needs such as educating children and purchasing food rather than investing in securing land rights (Djurfeldt, 2020).
The results show that households in the high land tenure security category had significantly stayed for more years (18.13) on the land compared to households in the low and medium land tenure security categories at 14.38 years and 16.46 years, respectively. This finding suggests that households who stayed at the land for longer periods tend to be more land tenure secure than those who stayed for shorter periods. The more a household stays in a particular parcel of land, the more they tend to carry out investments such as planting trees, thus increasing their land tenure security. Joel and Bergaly (2020) note that land investment is a function of time taken on the land, influencing land tenure security.
Overall, the average land size for the sampled respondents was 1.65 hectares. There was a statistically significant difference in land size at a 1% level among the three categories of land tenure security. Household heads in the high land tenure secure category had more land (1.90 hectares) than those in low and medium land tenure secure categories, with 1.31 hectares and 1.48 hectares, respectively. Having more land incentives one to secure it to protect it from land grabbers (Ghebru & Lambrecht, 2017).  Note: Standard deviations in parenthesis *, **, *** represents significance level at 10%, 5%, and 1%, respectively The average time to walk from the homestead to the parcel of land was 3.69 minutes, indicating that the parcels were relatively nearer to the homesteads. The results further indicate a significant difference at 1% in terms of the walking time to reach the parcel of land from the homestead. It took more time to reach parcels belonging to households in the high land tenure category (4.32 minutes) compared to those in the low and medium land tenure security categories at 2.78 minutes and 3.67 minutes, respectively. The more time it takes to reach the parcel, the further the distance, hence the incentive to secure the land rights to prevent private developers from grabbing it. Nearer parcels of land are usually more secure hence owners tend to be less motivated to secure them (Djurfeldt, 2020).
The results also show that households in the medium land tenure security category were significantly (1% level) more productive (2657.01) in producing maize than those in the low and high land tenure security categories at 2304.86 and 2588.17, respectively. Land tenure security tends to incentivise land owners to invest in productivity-enhancing technologies such as irrigation and using fertilizers. Additionally, in the case of ownership of a land title deed, holders can use it to secure credit facilities to invest in new farming technologies, which may increase productivity (Rashid, 2021). Table 4 presents the results of the descriptive and inferential statistics of the categorical variables used in the analysis. Overall, most of the respondents were males (54.10%). However, concerning the three categories of land tenure security, there was a statistically significant relationship at a 1% level between land tenure security and the gender of the household head. In the medium and high land tenure security categories, most household heads were males at 54.17% and 62.87%, respectively. In contrast, women were the majority in the low land tenure security category (58.57%). In most rural communities, the patriarchal system is rampant; hence, men have better access to resources and information, making it easier to secure their land than their female counterparts thus disadvantaging them from participating in income generating activities to increase their income (Rahman & Amit, 2022).
Overall, the majority (77.60%) of respondents did not acquire their land through purchase which would probably indicate the underdevelopment of the land markets in the study area. Regarding the three categories of land tenure security, most respondents acquired their land through purchase. There was also a statistically significant relationship between the method of land acquisition and land tenure security at a 1% level. Land acquisition through purchase in Kenya follows a well-laiddown procedure, and hence owners are most likely secure compared to those who have acquired land through other means, such as inheritance and leasing, among other ways (Valkonen, 2021).
The majority (71.86%) of the respondents noted that their land was fertile. In all three land tenure security categories, most respondents had fertile land at 63.57%, 79.17% and 76.73% for the low, medium and high land tenure security categories, respectively. Land fertility was significantly related to land tenure security at a 5% level as per the chi-square results. The fertility of the land is an important aspect in determining the value of the land. Less fertile lands are less valued due to their low productivity. On the other hand, high-fertility land are deemed productive, and hence owners are more likely to secure their rights over such parcel of land.

Determinants of land tenure security
The ordered probit model chi-square results (Table 5) of 122.48 with 16 degrees of freedom and a pvalue of 0.000 shows that it is significant at 1%. Thus the independent variables included in the model sufficiently explain the variations in the dependent variable. The model threshold parameters; α 1 and α 2 are significant at 5% and 10%, respectively, underscoring the reliability of the model results. Due to the problem of interpretation of ordered probit results, Wollni et al. (2010) and Mallick and Rafi (2010) suggest post-estimation marginal effects of the ordered probit model to ease the estimation of the model. Multicollinearity test of variables used in the analysis as presented in table 6 show that, the variables are not perfectly correlated and thus fit to be used in the model. Note: *, **, *** represents significance level at 10%, 5%, and 1%, respectively Sex, marital status, years of schooling of the household head, land acquisition, land fertility, period the household had stayed on the land, land size, parcel access, and household size are statistically significant at different levels. Being a male household head reduces the probability of the household having low and medium land tenure security by 45.5% and 1.9%, respectively. However, it increased the probability of having high land tenure security by 47.3%. The study area is a patriarchal community; men are regarded as the custodians of the family property, such as land and therefore possess more rights than women. Similarly, Whitehead and Tsikata (2003) found that women, especially in customary land tenure systems, face difficulties in possessing land rights. Moreover, due to women's weaker position concerning socio-economic status in the community, they tend to behave insecurely in land tenure (Ghebru & Lambrecht, 2017).
Marriage reduces the likelihood of having high land tenure security by 44.5%, while it increases the likelihood of having low and medium land tenure security by 40.7% and 3.8%, respectively. Married people are more likely to have children, so they may allocate the land to their young ones, making it difficult to be trusted by the community. Contrary to research by Ghebru and Lambrecht Note: Standard errors in parenthesis *, **, *** represents significance level at 10%, 5%, and 1%, respectively (2017) and Valkonen (2021), who noted that marriage helps household heads attain a high social and economic status and thus secure their land easily as compared to unmarried household heads.
Although the results show that advancement in education decreases the probability of households having low and medium land tenure security by 1.7% and 0.1%, respectively, it increases the probability of having high land tenure security by 1.8%. Higher levels of education likely expose an individual to opportunities that can be used to secure land. Formal education enables one to get timely, updated, and correct information on land-related policies that may improve land tenure security (Rao et al., 2020;Rashid, 2021).
The method of land acquisition was statistically significant at 1% at all levels of land tenure security. Acquisition of land through purchase increases the likelihood of households' to be high land tenure secure by 29.7%. In contrast, it reduces the probability of having low and medium land tenure security by 26.8% and 2.9%, respectively. Land purchase in Kenya follows a formal process, and therefore holders tend to have more secure rights than those who acquire land through other means, such as inheritance. In areas where land markets are relatively developed, land acquired through purchase tends to the more tenure secure since the purchase process mainly involves government institutions (Djurfeldt, 2020).
Ownership of fertile land increases the probability of having higher land tenure security by 15.3% as opposed to those in low and medium land tenure security categories at 14.7% and 0.6%, respectively. Due to the expected benefits derived from fertile lands, such as high productivity, holders tend to protect the land from illegal exploration and grabbing. Fertile lands usually attract a higher value in either formal or informal land markets. This motivates the owner to secure it for fear of losing the land (Coulibaly, 2021). On the contrary, Joel and Bergaly (2020) found that fertile land requires fewer investments for optimal production and hence less incentive to secure the land.
The higher number of years a household stayed in the land increases the probability of being in the high land tenure security category by 0.7% while reducing the probability of being in the low land tenure security category by 0.7%. The longer someone stays in a parcel of land, the more likely they make short-and long-term investments in the land. The investments may increase the tenure security of the land. Similarly, Brasselle et al. (2002) and Joel and Bergaly (2020) found a positive relationship between land investments and improvements such as planting trees, fences, soil conservation structures, and land tenure security.
An increase in land size by a hectare increases the likelihood of a household being in the high land tenure security category by 5.1%. In comparison, it reduces the likelihood of being in the low and medium land tenure security categories by 4.8% and 0.3%, respectively. Households with large tracks of land are more likely to be influential and, therefore, able to control the governing land system in the community. Ayamga et al. (2015) agree with the finding of this study and note that the feeling of being land tenure secure could result from abundant land. However, Alemu (1999) and Ghebru and Lambrecht (2017) argue that households with large pieces of land may find it difficult to manage and protect them from encroachment.
The more time (minutes) spent walking from the homestead to the land parcel increases the probability of a household's security land tenure by 4.6%. The more minutes, the more the parcel is exposed to land grabbers and illegal explorations; therefore, it incentivises securing the land to protect it. A plot of land relatively far from one's homestead is more likely to be reallocated to other users hence the motivation to secure the land (Sitko et al., 2014).
Households with more members are less likely to belong to the high land tenure security category land by 4.3%. Household members in most rural households are the providers of agricultural labour; therefore, the availability of sufficient manpower reduces the amount of idle land, demotivating securing the land. This conflicts with the finding by Rashid (2021), who argued that large households motivate land tenure security due to the increased demand for land.

Conclusion and recommendations
The study sought to determine land tenure security drivers using the ordered probit regression model. Findings indicate that high land tenure security household heads were older, more educated, owned relatively larger land sizes, and had higher maize productivity with relatively smaller household sizes than in low and medium land tenure security households. In addition, age, marital status, years of schooling of the household head, period the household had stayed on the land, land fertility, method of land acquisition, size of land, distance from the household to the parcel, and household size were the main determinants of land tenure security. The study findings have implications for the relevant policies on strengthening land tenure security in rural areas.
In order to improve land tenure security of households, more pro-youth and women policies should be developed to increase the community's social status. Laws that protect marital land should be enacted to enhance the tenure security of married people. Acquisition of land through purchase can potentially increase households' land tenure security. Therefore, the government and other stakeholders should focus on developing policies to ease the land purchase process, reduce transaction costs, and weed out corruption.

Areas of further research
Further research should focus on other measures of land tenure security such as use of land title deed.