Effect of large-scale land acquisition by domestic and international entities on farm investment in northern Ghana

Abstract

Rural households remained the access-marginalized demographic in Africa despite massive land rushes by both domestic and foreign enterprises. As a result, investment decisions made by agricultural households are affected. To resolve this issue, a study that examines how exposure to large-scale land acquisition (LSLA) impacts people's agricultural investment decisions is required. Using data from 664 households selected using a multistage sampling technique, we examined the nexus between LSLA under local and foreign organizations and short-term investment (STI) and long-term investments (LTI) in land-improving measures. The results revealed that although LSLA under domestic and foreign entities tends to have a negative and statistically significant impact on the probability of LTI, it has a positive and significant impact on the probability of doing so for STI. The results also showed a simultaneity between LSLA and farm investment. Thus, giving farmers legal ownership safeguards them as they undertake LTI. To encourage household investment and reduce further exposure to LSLA, improvements in legal land ownership can be made.

1. Introduction

Economic literature has long acknowledged the critical role that land, and yield-improving practices play in advancing sustainable agriculture and rural development [1]. For example, in the majority of Africa, these methods have helped to promote sustainable agriculture and rural transformation by increasing output, farm income, and food security while lowering hunger [2,3]. Because of this, the government and policymakers in Ghana [4] and Africa generally [5,6] have made land and yield-improving practices a long-term policy priority.

Secured land access, however, is core to the take-up of such practices among households [1,[7], [8], [9], [10], [11], [12], [13], [14], [15], [16]]. Additionally, it has been seen that households' access to incentives like financing is enhanced by secured access to land. Then, through market purchases, these incentives may be leveraged to preserve investments [12,13,16]. For these reasons, numerous think tanks and governments in the majority of Africa have reexamined their stance on land access and are now encouraging access among rural households. For instance, access to and investment in the land are still necessary for sustainable development under the Comprehensive Africa Agriculture Programme [6].

Despite attempts by African governments to promote fair access and rights, producers, particularly smallholder farmers, encounter numerous obstacles when trying to get land for agricultural production. Large-scale land acquisition (LSLA), which puts Africa's land and peasant access at risk [7,17], is one of these recent concerns. Although these purchases have helped to boost rural employment [[18], [19], [20], [21]], scholars and policymakers continue to disagree over how they affect smallholder adoption and investment choices.

On the one hand, some have claimed that LSLA can increase investment by facilitating the spread of technology to peasants [22]. This idea is predicated on the idea that LSLA investment will enable local peasants to gain work, learn new technologies, and eventually boost household investments in those technologies (ibid.). Opponents countered that because promised jobs and technology transfer are frequently not realized, LSLA does not increase household investments. According to Theting and Brekke [23] in particular, those investments from LSLA frequently fell short of the promise of job development, forcing residents to give up farming and hunt for employment elsewhere. Tinyade [24] and Behrman et al. [25] similarly made the case that the mechanical nature of such investments frequently results in the creation of few jobs. Robertson and Pinstrup-Andersen [26] also made the case that when land is appropriated for large-scale farm development, displaced people are left without a means of securing financial support for investments in land-improving methods.

A third viewpoint, which has evolved in addition to these two opposed ones, asserts that smallholder investment in land and advancement of methods rather determine the success of LSLA Suhardiman [27]. Therefore, farm households prefer to invest in land to oppose LSLA or to fight against being kicked out of LSLA, unless they receive full compensation for their investments, which is typically not the case.

In the wake of the above contrasting opinions, many studies investigated the impact of LSLA on livelihoods (e.g. Refs. [7,19,[28], [29], [30], [31], [32], [33], [34], [35], [36]]). Yet surprisingly, none of these studies specifically looked at how different investors' LSLAs affected households' farm investments, even though information on how these investors' LSLAs affected households' farm investments could be helpful to policymakers. The effects of LSLA by local and foreign businesses were examined in research by Ayamga et al. [7], Ayamga et al. [35], and Abdallah et al. [36], although the decisions about long-term investment (hereinafter, LTI), and short-term investment (hereinafter, STI) were not the objectives of these studies. Additionally, because of the admission of various investors into LSLA, agricultural households have been classed as those who lose land to LSLA either directly or indirectly from domestic and international investors [7]. However, the research lacks information on how LSLA by domestic and international investors affects households' LTI and STI decisions. Furthermore, none of these studies looked at how household LTI and STI affected LSLA. In other words, the simultaneity between LSLA, LTI, and STI has not been examined in prior research.

The primary aim of the study is to quantify how LSLA affects smallholder investment decisions in Ghana, considering the potential endogeneity of both LSLA and farm investment decisions. In particular, the study examines the effects of direct exposure (hereinafter, DE) and indirect exposure (hereinafter, IDE) to LSLA under domestic and foreign investors on STI and LTI in land-improving strategies in northern Ghana. Additionally, the study explores how LSLA in Ghana is affected by both STI and LTI. Farmers have frequently been discouraged from investing in land-improving strategies due to uncertainty around returns on households’ farm investments. Land title registration, which takes time, is one significant programme that can eliminate these issues [37]. To speed up the registration procedure, the government and other interested parties recently launched a land digitization program. The next course of action for the government and other stakeholders involved in the land digitization process will therefore be highlighted by information on the impact of LSLA from local and international investors.

We defined LSLA in this study as a land transaction involving a land area of at least 20.23 ha following Ghana's Lands Commission [38], one of the top agencies in charge of administering, acquiring, and registering land in Ghana. We adopted this criterion because the study was conducted in Ghana. Exposure to LSLA is defined as being close to impacted households, having limited land due to land transfer to domestic or foreign investors, losing cultivated or uncultivated land, and losing land-based resources [17]. Households exposed to LSLA by local or foreign companies lose productive- or forest resources because of the acquisition by domestic or foreign entities [36]. The IDE from LSLA by domestic or foreign firms can take many forms, including living in a community that has been affected, losing fallowed land, or having a small amount of land due to acquisition by local and foreign investors [17]. Losing productive resources because of domestic or foreign firms' purchases is DE from LSLA by domestic or foreign entities.

LSLA is held in Ghana by many investors. The unusual situation in Ghana, like in other nations, is that foreign investors continue to dominate the LSLA [39,40]. Therefore, it is critical to understand how LSLA by both domestic and international investors influences smallholder investments in the land- and yield-improving approaches. Such details might be crucial for Ghana's land usage, acquisition, and production strategies. According to Feder and Onchan [12], a household farm investment decision includes investments in (i) capital equipment, (ii) land and yield-improving methods, and (iii) non-agricultural activities and assets. In this study, we concentrated on land and yield-improving methods like drip irrigation, minimum tillage, crop rotation, residue retention, grass strips, NKP, sulphate of ammonia, and urea. Other methods included local catchments/dugouts, terraces, windbreaks, land leveling, waterways, pipes, ponds, and canals. We specifically divide these methods into two groups based on the lengths of time needed to see a return on investment in such methods: (i) LTI, which refers to investing in irrigation-related techniques like water pumps, wells, local catchments/dugouts, terraces, land leveling, pipes, ponds, canals, drip irrigation facilities; soil and water conservation techniques (SWCT), which refer to minimum tillage, crop rotation, residue retention, and grass strips; and (ii) STI, which refer to investing in inorganic fertilizers like NKP, Sulphate of ammonia and Urea. We assess the impacts of LSLA by domestic and foreign investors on the two methods mentioned above using data from 664 homes during the 2017/2018 farming season.

This study adds to the body of knowledge on LSLA and farm investment in two key areas, aside from being the first study on the impact of LSLA on farm investment as differentiated by investors. First, by examining the variables affecting household exposure to LSLA by both domestic and foreign investors, we add to the body of literature. As a result, we were able to establish for the first time how household factors affect both local and international organizations' LSLA. Second, we used the two-stage conditional maximum likelihood (hereinafter, 2SCML) procedure [41] to handle the potential endogeneity of LSLA and farm investments, a nontrivial issue that has not been adequately addressed in previous research. When an independent variable correlates with the error term in a regression model, it is known as the problem of endogeneity in econometrics [42]. Investors purchase land based on a variety of factors, including soil fertility, slope, depth, and proximity to a road and a market [43]. However, because researchers may not be aware of these qualities, they may also have an impact on households' decisions to invest in the land- and yield-improving strategies, which results in a connection between the LSLA and error term in the regression model for such investments. Furthermore, there is a chance that farmers will simultaneously oppose LSLA or take steps to avoid evictions or other LSLA-related displacements.

The remaining part of the paper is structured as follows. The overview of LSLA in Ghana is presented in Section 2. The approach is then presented in section three along with a conceptual model, estimating techniques, and study data. The findings and discussions are reported in section four. Section five then presents the conclusion and a few policy suggestions.

2. Overview of LSLA in Ghana

Large-scale land acquisition is not new as several attempts at such acquisitions existed in the colonial period. For instance, the 1894 Crown Lands Bill was introduced to gain control over the concession-granting procedure and land with mining locations for the advantage of foreign gold investors. The bill was intended to transfer to colonial authorities all authority held by traditional leaders over “wasted land” (land with mineral and forest resources) [44]. Chiefs and intellectuals, including newspaper editors, opposed the law, claiming that the residents of the Gold Coast could administer their lands. A further illustration is the 1897 Crown Lands Bill, a modified version of the 1894 Crown Lands Bill that reexamined the objective of ownership to administration. The Gold Coasters, however, continued to oppose the new legislation, claiming that it continued to give the colonial government excessive power. LSLA is thus nothing new in Ghana. However, the present rate of acquisition has varied from the past. For instance, between 2004 and early 2009, about 452,000 ha of land deals have been approved in Ghana [45]. This seems to have reduced recently as the total size of approved deals is reported to be 403,907 ha and the aggregate size under contract is below 400,000 ha by 2020 [21]. Out of approved deals, 112,675 ha are reportedly abandoned, 88,827 ha are in operation, 35,070 ha are without projects, 29,975 are in the start-up phase, and 137,360 ha exist without information (Ibid.). In northern Ghana, a total of 23,000 ha, 400 ha, and 1363 ha of land were respectively acquired by Biofuel Africa Ltd in 2008 for Jatropha production [46], Integrated Water Management and Agricultural Development (IWAD) Ghana Ltd in 2011 for food production [47], and Integrated Tamale Fruit Company (ITFC) in 2000 for food mango plantation [48]. However, while IWAD is still ongoing, reports and protests about the potential negative impacts of the project by Action-Aid in 2009 and Regional Advocacy, and Information Network Systems (RAINS) led to the collapse of the project [46,48]. In southern Ghana, 13,000 ha, 13,000, and 500 ha were respectively acquired by Scan Fuel (now Scan Farm Ghana Ltd) for jatropha production in 2008, KIMMINIC for Jatropha and maize production in 2008; and the European Union for food and biofuel production in 2009. Whereas the projects by Scan Farm Ghana Ltd and the European Union are ongoing, the project by KIMMINIC has been temporarily suspended since May 2012 [46].

Aside from the variation in the rate of LSLA, the investors also varied in Ghana. At the national level, for instance, all the aforementioned land deals came from a total of 46 concluded land deals, and out of this, 11 deals involved domestic investors while the rest of the 35 concluded deals involved foreign investors [21]. Further, purposes/drivers for such land deals also varied from the past. Whereas access to minerals and forest resources were the major purposes for which attempts were made to acquire land [44], migration, urban extensions, and ecotourism, the zeal to conserve nature, maximize profit, enhance FDI in food and non-food commodities remain the reasons for the recent upsurge in LSLA [49]. In Ghana, available information revealed food and biofuel production as the two major reasons for which these deals are carried out [21]. For instance, out of the 46 concluded deals, 31 concluded deals involved food and non-food crops with biofuel production and forestry inclusive.

Meanwhile, the occurrence of these deals is not without dispossessions of local occupants across various regions in Ghana. Available records showed that dispossessions included 27 producers in Savelugu-Nanton District [48] and 25 farmers in Mion Districts [18,46] of northern Ghana, 16 and 29 households in Nkoranza and Agogo Traditional Council of Southern Ghana [50]. Although some of these projects are currently abandoned [21,51], there is no known record of reclaims of land by farmers who were affected. The current study looks at the nexus between LSLA, STI, and LTI in northern Ghana.

3. Theoretical background

In theory, two strands of literature can be observed about the ways through which LSLA affects households' farm investment decisions. The first strand [22,[52], [53], [54]] views investments in LSLA as a subset of foreign direct investment (FDI). In the FDI literature [[55], [56], [57], [58], [59], [60], [61], [62]], it is believed that FDI enhances learning opportunities, and knowledge of local firms, thereby contributing to the transfer of technologies to these firms. Thus, in line with the FDI literature, the first strand of studies including Kleemann and Thiele [22], Santangelo [52] and Dessy et al. [54] believed that LSLA can enhance smallholder access to improved inputs and new technologies, thereby improving investment among smallholder farmers. In an attempt to test such a proposition, Dessy et al. [54], for example, developed an occupational choice model to examine the mechanisms through which FDI in farmland affects peasant welfare. They concluded that if proceeds received from foreigners for LSLA by local authorities are invested in subsidizing the cost of inputs, investment in technologies will be improved among local farmers since these technologies will now be affordable to the local farmers. Based on Dessy et al.‘s [54] model, Kleemann and Thiele [22] also developed a theoretical model to study the mechanisms through which LSLA might affect rural populations in Sub-Saharan Africa (SSA). Again, the hypothesis of Dessy et al. [54] on the effect of LSLA on households' investment was reinforced in Kleemann and Thiele's [22] study. Nonetheless, both studies (i.e. [22,54]) lack empirical evidence to support the theoretical propositions. On the other hand, the effect of LSLA was tested empirically by Santangelo [52]. However, the focus of Santangelo [52] was on LSAI's impact on food security and not farm investment.

The second strand of the literature including Abdallah et al. [35], and Kariuki and Ng'etich [63] views LSLA as central to land tenure insecurity. The land tenure insecurity literature has argued that the uncertainty of households regarding whether they will be able to reap all benefits of investments has a bearing on farm households' investments (e.g. Refs. [12,13,[64], [65], [66]]). This argument is particularly true in Feder and Onchan's [12] study where a theoretical model was developed to depict the relationship between land tenure insecurity, and farm investment in Thailand. This model was later adopted by Place and Hazell [64], Hayes et al. [13], and Place and Migot-Adholla [67] in Ghana, Gambia, Kenya, and Rwanda to test the effect of land tenure security on farm investments, and productivity. Although none of these studies made explicit mention of how LSLA affects land improvement, they both suggest that farm investments are influenced by tenure (in)security which is also strongly influenced by LSLA [30,63]. On this basis, Abdallah et al. [35] linked LSLA to farm investments arguing that tenure insecurity's effect on farm investment is inherently an LSLA's effect on farm investment.

4. Conceptual model

In this study, our conceptual model of the link between LSLA and households' investment decisions can be derived from Feder and Onchan's [12] model using the expositions in the literature discussed. Although the policy environment in Thailand where Feder and Onchan's [12] model was developed is different from Ghana, we employed such a model as the basis for our conceptual model for three reasons. First, the model is more flexible and applicable in Ghana as most of the variables proposed are observable from household surveys employed. Second, compared to other models, the propositions of Feder and Onchan's [12] model are easily testable with the 2SCML model of Rivers and Vuong [41] employed to examine the relationship between LSLA and farm investments. Third, LSLA comes with appropriation and displacement which create tenure insecurity that is strongly linked to farm investment.

According to Feder and Onchan [12], a farmer has the option of investing in equipment K, which is not lost to displacement, land improvements M, which are entirely lost during displacement, and non-agricultural assets Z, which are not affected by displacement. The sum of the expected terminal wealth in the presence and absence of eviction is maximized subject to several constraints including cash and credit constraints, agricultural output, yield, and land improvement functions. Application of the first-order conditions for a maximum, and Cramer's rule yield equation (1) which implies that investment in land improvement $M$ is driven by tenure insecurity $TS$, the initial wealth $W_o$, land $A$, and capital $S$.

$$M=M\left(\text{TS},W_o,A,S\right)$$ (1)

However, this study's goal is to look at how exposure to LSLA impacts farm investment. Therefore, our main task in this section is to conceptualize how LSLA enters equation (1). To begin with, we assumed that LSLA is intrinsically tenure insecurity because it makes people less comfortable about their farm investments, and results in evictions [54]. This assumption helps us comprehend how LSLA enters equation (6). If the link between LSLA and tenure security (TS) is contemporaneous, given other factors X, TS = f (LSLA|X) or LSLA = f (TS|X). Furthermore, it is implied from the research (e.g. Refs. [22,54]) that LSLA directly decreases the amount of land accessible for production. So, in addition to how LSLA affects $M$ through TI, the relationship between LSLA and $M$ can also be seen through how much land is accessible for habitation. Based on these expositions, we evaluate the impact of LSLA on farm investment.

5. Methodology

5.1. The research site, data, and description of variables

5.1.1. Study area

The Northern Region (now demarcated into Northern, Northeast, and Savannah regions), which is in Ghana's north, is the study area for this investigation. Most of the area is guinea savanna, and the bulk of the population works in agriculture. It is distinguished by a single rainy season with erratic and irregular rainfall as well as unfavorable elements like high temperatures that have an impact on the region's agriculture. The region makes up over 70,384 square kilometers of Ghana's total land area, although it has a population of only 2,479,461 and a density of 35 people per km square [68].

To regulate land, two tiers of government are used [18]. The paramount chiefs of the Mamprugu, Bimbilla, and Gonja Traditional areas make up the first tier of the hierarchy. Concerning the first tier, the land is managed with numerous nonwritten laws and customs that pose some difficulties, such as multiple sales and a plurality of management. The local government system, where state laws apply to the land, is the second tier. The governmental organizations, however, with whom the property is placed are the issue. The Land Title Registry and the Lands Commission both assert that they are the final decision-makers when it comes to managing land [69]. These characteristics make the region a hotbed for LSLA as they are explored by investors. Thus, the land is acquired by domestic and foreign investors on a large scale from the chiefs for different purposes. However, while some of these deals have been concluded and contracts are signed, some are in operation and others are abandoned for different reasons (e.g., resistance from local people, negative publicity by NGOs, and lack of funds [46,70]).

In the Savannah Region of Ghana, for example, 38,000 ha of land were reportedly transferred by chiefs to Biofuel Africa Ltd (now Solar Harvest) - a foreign company - for the Jatropha plantation [70]. However, the investment has been halted by the Environmental Protection Agency of Ghana due to the non-consultative and nontransparent nature of the deal. In the Northern Region, a total of 23,000 ha was transferred by the local chiefs to the same company (i.e., Biofuel Africa Ltd.) for food and biofuel production [46]. However, the land has been currently abandoned due to contestations from Action Aid Ghana.

In the Northeast Region of Ghana, Ayelazuno [47] also reported 400 ha of land acquired by IWAD-Ghana Ltd for food production. This investment is currently in operation to produce food for local and international markets. Kuusaana [18] also reported land deals of 1363 ha acquired by the Integrated Tamale Fruit Company (ITFC) in the Savalegu district for food production. This investment is currently for mango plantations and is jointly owned by the people of Ghana and the Netherlands on a 70:30 basis, respectively. In the same location (i.e., Savalegu district), about 3237 ha have been acquired by the Government of Ghana for industrial purposes while 50,000 is intended to be acquired by the Base for Industrial and Environmental Technology, Ghana (BIET-Ghana) in the district [51].

5.1.2. Data collection

Regarding the data, we gathered both quantitative and qualitative data for the study. In total, 690 exposed and non-exposed agricultural families were selected from 240,238 agricultural households in the Northern Region's geographic enclave. Using a multistage sampling procedure, the households were chosen during the 2017/2018 cropping season. The districts of Central Gonja, Mampurugu-Muagdure, Mion, North Gonja, Sagnarigu, and Savelegu-Nantong were chosen for the first stage based on the prevalence of substantial tracts of agricultural land covered by commercial agreements. The study districts are depicted on the map of northern Ghana in Fig. 1.

Fig. 1.

A map illustrating the study districts in northern Ghana.

We used verified data from the Northern Regional Lands Commission to choose these districts. According to this data, the six districts that control LSLA represent around 98.87% of all deals recorded in the region. A total of 23 impacted communities were chosen from the six districts for the second stage through a reduction procedure. Scoping of the districts, which is the first stage of the reduction process, involves identifying LSLA-affected localities. A key informant interview guide served as the foundation for the identification of the affected localities. The operational definitions described in Section 1 were used to influence the construction of the interview guide. Key informants such as MoFA agents, and community leaders were asked these questions. The purpose of the key informants was to assist the researcher in accurately identifying the communities impacted by extensive land purchases. Any neighborhood with exposure to LSLA that met the operational criterion was recorded as a community impacted by LSLA. In all, 41 impacted localities were profiled from the scoping exercise. To determine which communities the LSLA by local and international entities better represented, these groups were contrasted further. In the reduction phase, 23 communities were sampled from the 41 affected communities based on the prevalence of LSLA by domestic and foreign firms. It was difficult to distinguish between agricultural households directly affected by domestic or foreign entities, such as those that lost productive, forest resources, etc., and those that were indirectly affected, such as those that lived in affected communities, lost fallowed land or had limited land due to LSLA. This is because there was no sampling frame available for LSLA-affected farm households. To develop a sample frame for each community visited, common areas - where farmers generally gather to play neighborhood games and discuss daily farming duties and other pressing issues - were discovered. The farmers in these areas were then asked for the names of the households that had been exposed to LSLA. Each response supplied was interrogated using the same interview guide employed for the selection of communities. In the study's final stage sample of exposed households, the names that met the operational definition of DE and IDE to LSLA by domestic and foreign companies were compiled into a list. Those names that did not meet the criterion were classified as nonexposed households. For the list of nonexposed households, we further selected households with plots farther away from exposed households to avoid spillover effects. Using the lottery method of random selection and an improvised list from the area, a total of 690 farm families were sampled. After sampling, a semi-questionnaire-based household survey was carried out. The questionnaire was uploaded to the Kobo Toolbox so that enumerators could interview household leaders since they oversee most home tasks and have more knowledge about households. The questionnaire was administered by ten enumerators who have completed training in questionnaire administration through the Kobo Toolbox. The online open-source Kobo Toolbox was created by the Harvard Humanitarian Initiative to collect field data. The Excel formats that are kept by the software can be downloaded and imported into Stata for analysis. Therefore, using the software has the notable advantage of freeing up time from data entry to be used for other duties. For the household survey, questions from the questionnaire were subsequently loaded onto each enumerator's Android phone, which had the Kobo Toolbox installed. The survey gathered information on a range of subjects, including household exposure to LSLA, institutional traits, household and location traits, agricultural investment, and food output. Because some responses were lost in the data due to nonresponses, a study sample of 664 households was employed.

5.1.3. Demographic characteristics of respondents

Table 1 presents the demographic characteristics of the study respondents. In terms of ethnicity, the majority, representing 74.7% of the study respondents, is from the Mole-Dagbani ethnic group including the Dagomba, Mamprusi, and Frafra people. This is followed by those from the Guan ethnic group, representing 19.7% of the sample respondents. The Gurma ethnic group including people from Bimoba, Konkomba, and Kotokoli is the least ethnic group of the sample respondents.

Table 1.

Respondents’ demographic characteristics.

Ethnic group	Frequency	Percent
Guan	131	19.73
Gurma	37	5.57
Mole-Dagbani	496	74.70
Total	664	100.00
Marital status
Married	360	54.22
Consensual union	33	4.97
Separated	4	0.60
Divorced	36	5.42
Widowed	33	4.97
Never married	198	29.82
Total	664	100.00
Head's occupation
Unemployed	119	17.92
Agric. And fishery workers	409	58.88
Civil service/salaried worker	17	2.56
Craft and related trade workers	120	18.07
Pastor/Mallam/Religious leader	10	1.51
Herbalist	7	1.05
Total	664	100.00
Religious denomination
Christian	268	40.36
Muslim	355	53.46
Traditional	41	6.17
Total	664	100.00

Source: Field Survey, 2018.

In terms of marital status, the majority of the respondents have experience in marital relationships. Specifically, about 54.2% of the respondents are married. Further, the divorced, widowed, and those in consensual union are about 5% each in the sample. On the other hand, those who have never married represent only 29.8% of the sampled respondents.

With regards to occupation, the sample is dominated by agriculture and fishery workers representing 58.9% of the sampled respondents. The second majority (18.1%) represent those in craft and related trade work, and this is closely followed by the unemployed who represent 17.9% of the total sample. The rest of the sample (i.e., about 1% and 2%) represent those engaged in religious services and herbal activities, respectively.

In terms of religion, Islam is the dominant as about 53.5% are Muslims. This is followed by Christianity with 40.4% of the sample representing Christians. The traditional African religion is the least practiced religion with 6.2% of the sampled respondents. This is in line with the report from the Ghana Statistical Service which showed that Islam is the dominant religion in the Northern, Northeast, and Savannah regions [71]. Overall, the dominance of the Mole-Dagbani, married, agricultural workers, and Muslims in the sample implies that these groups of people are most likely to be affected by LSLA in the area.

5.1.4. Description of variables employed for analysis

Table 2 displays detailed information on the variables and their descriptive statistics. Twenty percent of the households in the study sample were not exposed to LSLA, twenty percent had domestic entities directly expose them to LSLA, twenty percent had domestic entities indirectly expose them to LSLA, twenty percent had foreign entities directly expose them to LSLA, and twenty percent had foreign entities indirectly expose them to LSLA. A little over 38% of the households in the sampled population make LTI, compared to 62% who make STI. The idea that farmers exposed to LSLA are obliged to select STI over LTI is not supported by these findings, though. An amount of GHȼ5,095.65 (USD884.66) is the average household income for the sample; at the time of the study, USD1.00 was equal to GHȼ5.76. According to the statistics, there are more men than women in the sample overall. Since men are often the main providers for their families, policies that affect the creation of resources are more likely to affect the well-being of many homes. It is also crucial to remember that the sampled households are made up of adults or mature groups of respondents. The sample's respondents have an average age of about 47 years. This demonstrates that the sample is within the ages that make the labour force [72]. Additionally, the sample's low educational level is indicated by the average amount of time spent in school, which is 2 years. As a result, this is probably going to have more of an impact on how easily and how much money a household invests in the land. For the qualitative data, there were six focus group meetings. To save space, we decided to publish information about focus group data collection as supplemental data to this study (see Section S1).

Table 2.

Variables definition/measurement.

Variable	Definition/measurement	Mean (SD)
LTI	1 if the household had invested in either irrigation or SWCT and 0 otherwise)	0.38 (0.49)
STI	1 if the invested in any inorganic fertilizer and 0 otherwise	0.62 (0.40)
Household income	Total income (in GHȼ)	5095.65 (21.03)
Fertilizer subsidy	1 if the head participated in the 2017/2018 fertilizer subsidy, 0 otherwise	0.75 (0.43)
Gender	1 if the head is male, 0 otherwise	0.93 (0.26)
Age	Age of head in years	46.97 (2.87)
Household size	No. of members in a household	12.44 (7.28)
Education	No. of formal schooling years of head	1.97 (3.86)
Farm size	Size of the plot (ha)	6.39 (3.78)
Leadership	1 head is in a leadership position; 0 otherwise	0.26 (0.44)
Sagnarigu	1 household is in the Sagnarigu district, 0 otherwise	0.17 (0.38)
Mion	1 household is located in the Mion district, 0 otherwise	0.09 (0.29)
Central Gonja	1 household is in the Central Gonja district, 0 otherwise	0.18 (0.39)
Savelegu	1 household is in the Savelegu district, 0 otherwise	0.36 (0.48)
Yagba-Kubori	1 household is in the Yagba-Kubori district, 0 otherwise	0.17 (0.38)
North Gonja	1 household is in the North Gonja district, 0 otherwise	0.02 (0.15)
Water resources	1 the village has water resources; 0 otherwise	0.41 (0.49)
Road	Proximity to a weathered road (km)	0.35 (0.48)
Credit	1 head has access to credit; 0 if otherwise	0.33 (0.47)
FBO membership	1 head is a member of any farmer-based organization (FBO); 0 otherwise	0.39 (0.49)
Knowledge	1 head has prior knowledge of households affected by LSLA; 0 otherwise	0.61 (0.49)
Good fertile	1 good fertile soil; 0 if otherwise	0.38 (0.49)
Moderately fertile	1 moderate fertile soil; 0 if otherwise	0.45 (0.50)
Poorly fertile	1 poor fertile soil; 0 if otherwise	0.17 (0.38)
Land institution	1 head has access to formal land institutions; 0 otherwise)	0.34 (0.48)
Non-exposure	1 head did not lose land; 0 otherwise	0.20
Exposure to LSLA by domestic entities
DE to LSLA by domestic entities	1 DE of head to LSLA by domestic entities; 0 otherwise	0.20
IDE to LSLA by domestic entities	1 IDE of head to LSLA by domestic entities; 0 otherwise	0.20
Exposure to LSLA by foreign entities
DE to LSLA by foreign entities	1 DE of head to LSLA by foreign entities; 0 otherwise	0.20
IDE to LSLA by foreign entities	1 IDE of head to LSLA by foreign entities; 0 otherwise	0.20

Notes: STI and LTI are short- and long-term investments. See the Introduction for detailed definitions of DE and IDE to LSLA by domestic and foreign entities.

5.2. Estimation strategies

Based on the conceptual model above we specify the following general equation for household investment decisions:

$$I_{ik}^{*}={\delta }_i{E}_{ik}+{\vartheta }_i{X}_{ik}+{\mu }_{ik}\left(k=\text{LTI},\text{STI}\right)$$ (2)

the sort of investment made by $i^{th}$ household is $I_i$; $E_i$ is exposure to LSLA; $X_i$ supply-side factors; ${\delta }_2$ and ${\vartheta }_i$ are the coefficients of $X_i$ and $E_i$; and ${\mathrm{\mu }}_{\mathrm{i}}$ is a random term. Equation (2) can be converted into a binary probit equation for an investment option using the following mapping from the latent variable to its observed realization, denoting long- and short-term agricultural investments as LTI and STI, respectively:

$$I_{ik}=\{\begin{array}{c}1if{I}_{ik}^{*}>0\\ 0if{I}_{ik}^{*}\le 0\end{array}\left(k=LTI,STI\right)$$ (3)

Equation (2) suggests that ${\mathrm{\delta }}_{\mathrm{i}}$ is the effect of exposure to LSLA on household farm investment decisions. This is accurate if we assume that there is no correlation between the index of exposure to LSLA and the error term in equation (2). However, state, and traditional authorities may transfer the plots based on their high-quality characteristics (which are frequently invisible). Such a non-random method of choosing plots for transfer to investors may result in a systematic distinction between homes that have been exposed and those that have not. In the conceptual model, it has also been demonstrated that spending money on both STI and LTI strategies can reduce LSLA. In this sense, the estimation of equation (2) without accounting for endogeneity will result in inaccurate estimates. To consistently assess the impact of exposure to LSLA under local and foreign businesses, the following systems of the simultaneous equation are specified:

$$E_i=a_0+a_i{X}_i+{\pi }_i{I}_i+{\epsilon }_i$$ (4)

$$I_i={\delta }_0+{\delta }_1{E}_i+{\delta }_2{X}_i+v_i$$ (5)

Where $E_i$, $X_i$, $I_i$ are as described previously, $a_i$, ${\pi }_i$, and ${{\delta }_i}^{\prime }s$ are their respective coefficients; $a_0$ and ${\delta }_0$ are the intercepts; ${\epsilon }_i$ and $v_i$ are residuals. A two-stage least square method has been suggested to estimate the aforementioned system of equations, but it calls for either the dependent variable or endogenous variable to be continuous. Since both exposures to LSLA and the various farm investments were each captured as binary variables, the 2SCML model [41] is employed. The 2SCML assumes normality for the first-stage residuals and the endogenous variable and thus, incorporates these variables into the second-stage probit equation instead of only the predicted values of the endogenous variables. We chose the 2SCML strategy because, it performs better than most alternative methods [41], especially with a high sample size. The 2SCML also offers several practical endogeneity checks. The following sets of equations were employed for analysis:

$$E_i=a_0+a_i{X}_i+{\eta }_i$$ (6)

$$P_i^I={\delta }_0+{\delta }_1{E}_i+{\lambda }_i{\hat{E}}_i+{\delta }_2{X}_i+{\omega }_i$$ (7)

${\hat{E}}_i$ is the residual term from equation (6), where $P_i^I$ denotes the likelihood that a household $i$ will make an STI or LTI. The two-stage conditional least squares (2SCML) method was then used to estimate these equations. To produce residuals, instrumental factors, and household variables are regressed on $E_i$ in the first step of the 2SCML approach. The second stage probability of farm investment $P_i^I$ in equation (7), is then estimated with the inclusion of ${\hat{E}}_i$, $E_i$, and $X_i$. In this study, the land institution is used as an instrument to determine the first stage probit model (measured as 1 if the family has access to a formal land institution, including the Lands Commission, etc.; 0 otherwise). It makes sense that farmers who have access to land institutions learn more about the land and related challenges. Information regarding land and associated issues can increase the rate at which land tenure rights are registered, certified, and secured, which may reduce the likelihood of encountering LSLA. The common t-statistic for coefficients ${\lambda }_i$ is a valid test of the null hypothesis that the exposure variables are exogenous in the investment equations, which is one of the key characteristics of 2SCML [42]. The significance of ${\lambda }_i$ shows that exogeneity is disregarded and that endogeneity is corrected by including the residuals.

The LTI and STI may be interdependent in addition to selection bias. As an illustration, a household that invests in long-term land repair will have fewer resources for short-term projects that require immediate attention (loss of labour and income effect). Estimates of the effect of LSLA on improvement decisions may be biased if this possible dependency between LTI and STI is ignored. Therefore, it is most likely that the error terms of these equations will be connected because of the potential substitutability or complementarity between the investment possibilities. The error terms jointly follow a multivariate normal distribution in the multivariate model, where the multiple investment decision is feasible, with a 0 conditional mean and variance normalized to 1 (for parameter identification), where: $\left({\mu }_{LTI},{\mu }_{LTI}\right)\sim MVN\left(0,\sum \right)$ and $\sum$ is the symmetric covariance matrix. The parameters and error correlations can then be estimated using the maximum likelihood technique [73]. In this study, multivariate probit modeling is used in the second stage estimation of the impact of exposure to LSLA on STI and LTI under each of the domestic and foreign companies [i.e., equation (7)]. In the results, positive and negative correlations respectively imply complementarity and substitutability, and these can be inferred from the correlation coefficients between the farm investment decisions [74]. The 2SCML is used to evaluate the effects of the investment decisions on exposure to LSLA:

$$I_i={\delta }_0+{\delta }_2{X}_i+{ϵ}_i$$ (8)

$$P_i^E={\delta }_0+{\delta }_1{I}_i+{\gamma }_i{\hat{I}}_i+{\delta }_2{X}_i+{\xi }_i$$ (9)

Where ${\hat{I}}_i$ is the residual term from equation (8), and $P_i^E$ in equation (9) denotes the likelihood of exposure to LSLA. Estimation of equations (8), (9) call for the employment of a probit model for the first stage estimate and the introduction of the ${\hat{I}}_i$ and $I_i$ into a second-stage model of $P_i^E$. Since exposure to LSLA was recorded as a categorical variable and dissimilar,1 the second stage of the influence of farm investment decisions $I_i$ in equation (8) on exposure to LSLA $P_i^E$ in equation (9) employed multinomial logit regression. With prior information on other homes affected by LSLA (measured as 1 if any household member has prior knowledge of other communities affected by LSLA; 0 otherwise), we identify the probit model in equation (8). Farmers who are aware of other LSLA-affected households in other localities are more likely to invest in land-improvement strategies to strengthen the security of their land's tenure. Consequently, the risk of eviction from agricultural investment is decreased by first-hand knowledge of the types of land purchased by investors. For instance, Suhardiman et al. [27] found that farmers who invested in rubber plantations and had prior knowledge about LSLA from family and connected networks increased the security of their remaining land. Even though the instruments appear to be valid intuitively, we test for validity using Lee's [75] overidentification test statistics, which have degrees of freedom equal to the number of instruments that were removed and χ2 distribution. According to Lee [75], the insignificance of the instruments in the estimation of the alternative versions of equations (7), (9) is then considered evidence validity of the instruments. Again, coefficient ${\gamma }_i$ is a valid test of the endogeneity of the investment variables in the exposure equations [42]. Exogeneity is rejected and their inclusion corrects for endogeneity if the coefficient ${\gamma }_i$ is substantial.

6. Results and discussion

The purpose of the study is to examine the nexus between LSLA and smallholder investment decisions in Ghana, considering the potential endogeneity of both LSLA and farm investment decisions. The following sections present the findings.

6.1. Exposure to LSLA

The regression findings of the factors that affect both DE and IDE to LSLA for domestic and international firms are shown in Table 3. As previously noted, LSLA exposure was recorded as a polychotomous variable for each of the domestic and foreign businesses. As a result, in the second stage of the 2SCML, the results were produced using an MNLM where LTI and STI, and their residuals from the first-stage linear probability model (Table A1 of the appendix) were added to the MNLM.

Table 3.

Multinomial logit estimates of determinants of exposure to LSLA.

Variable	Under exposure to LSLA by domestic entities				Under exposure to LSLA by foreign entities
	DE		IDE		DE		IDE
	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
LTI	−0.93	0.06***	−0.28	0.06***	−0.41	0.03***	−0.15	0.04***
STI	−0.17	0.02***	−0.82	0.14***	−0.24	0.01***	−0.66	0.08***
Residual_LTI	−0.09	0.02***	−0.15	0.04***	−0.41	0.17**	−0.17	0.09*
Residual_STI	0.12	0.02***	−0.22	0.07***	−0.21	0.01***	0.31	0.03***
HH_income	−0.08	0.13	0.21	0.21	−0.01	0.01	0.01	0.12
Leadership position	−0.04	0.01***	−0.20	0.12*	−0.04	0.02**	−0.11	0.06**
Gender	−0.01	0.01*	−0.11	0.03***	−0.41	0.17**	−0.02	0.12
Age	−0.26	0.17	−0.04	0.27	0.21	0.17	0.08	0.12
Household size	0.12	0.44	0.17	0.77	−0.33	0.41	−0.13	0.12
Education	−0.05	0.00***	−0.09	0.00***	−0.35	−0.22	−0.12	−0.16
Farm size	−0.03	0.12	0.03	0.19	−0.28	0.35	0.19	0.24
Land institution	−0.01	0.01*	−0.07	0.00***	−0.12	0.07*	−0.28	0.12**
FBO membership	−0.15	0.09*	−0.55	0.17***	−1.18	0.69*	−0.28	0.15*
Road	0.18	0.30	0.48	0.55	−0.45	0.57	0.44	0.39
Credit	−0.50	0.11***	−0.17	0.04***	−0.11	0.74	−0.35	0.50
Water source	0.11	0.23	0.43	0.55	0.29	0.02***	0.21	0.03***
Good fertile	0.52	0.04***	0.16	0.00***	0.38	0.05***	0.64	0.33*
Moderate fertile	1.20	0.04***	0.38	0.14**	−0.76	0.33**	0.47	0.28*
_cons	−1.04	0.08***	−0.40	0.08***	−1.99	1.07*	−0.56	0.13***
Joint significance of location variables: χ2 (6)	52.53***
χ2-statistic for joint significance of residuals	3.13***
Joint significance of instruments	0.98 (0.83)
No. of observations	472				531

Notes: ***, ** and * indicate statistical significance at 1%, 5% and 10%.

Source: Author's Computation from Field Survey, 2018

The residuals are all statistically significant, as shown in Table 3. The residuals are also collectively substantially significant, according to the Wald chi-square test, which also showed this. As a result, the exogeneity of LTI and STI is rejected. These imply that if we had not considered endogeneity, the results of the investment variables would have been biased. Thus, endogeneity is corrected by including the residuals of both LTI and STI. The assumption that households' knowledge of other households affected by LSLA can only influence DE and IDE of LSLA through LTI and STI is not rejected by the two statistics for the joint significance of the instrument.

Results for variables that represent LTI and STI are particularly intriguing. As shown in Table 3, the probability of DE and IDE of LSLA under domestic and foreign entities appears to be lower for households with LTI and STI in northern Ghana because these variables have negative signs as expected and are significant at 1% in the equations for DE and IDE of LSLA under both domestic and foreign entities. These results are conceivable given that remuneration for plots with tenure assurance is pricey and may prevent investors from buying such plots.

Transparency, respect for human rights, and assuring the sustainability of benefits, and the environment are also part of the acceptable code of conduct for purchasing land on a large scale (see, for instance Refs. [38,76]). This would be costly, especially for plots with tenure security, and it might also discourage investors from buying them. Nevertheless, these results are consistent with the idea that households that invest in land-improving technology increase the security of the land's tenure and are less likely to experience DE and IDE from LSLA. The findings support the idea that LSLA affects investment, as well as investment affects LSLA in northern Ghana. This result adds credence to theoretical and empirical studies that suggested eviction rates could be reduced by investing in land-improving methods (e.g. Ref. [77]).

Furter, the probability of DE and IDE to LSLA under domestic and foreign entities appears to be lower among males, leaders, and highly educated people, according to the results. This further supports studies (e.g. Ref. [35]) that found that intrahousehold power dynamics are crucial for limiting LSLA exposure in households.

Land institutions also showed the anticipated negative sign and significantly affected both DE and IDE under domestic and foreign entities, suggesting that households with access to formal land institutions are less likely to experience either DE or IDE from LSLA under domestic and foreign entities. This is conceivable given that information from land institutions can lower the rate of eviction by facilitating land registration. This finding supports the findings of Cotula [78], Lay and Nolte [53], and Giovannetti and Ticci [79] who found a positive influence of institutional quality on LSLA but contradicts Arezki et al. [80] who found mixed results for the influence of institutional quality on LSLA.

FBO membership is also credibly supported and has a significant impact on both domestic and foreign entities' DE and IDE to LSLA, suggesting that households that engage in group activities are less likely to experience either type of exposure. The ideal situation is that social groups like Action Aid-Ghana and the RAINS are at the forefront of the struggle against acquirers given the reappearance of LSLA following the energy, food, climate change, and financial crises of 2007–2008. These organizations adopt catchphrases like “environmental protection watchdogs” and “guardians of livelihoods of the poor” to emphasize their concerns about the consequences of LSLA for food security [46,81]. In their battle against LSLA, the voices of the marginalized adopted some of these ideas. Therefore, it should come as no surprise that FBO participation has a major impact on a household's DE or IDE to domestic or foreign entities' LSLA.

Although access to credit is highly correlated with solely DE and IDE to LSLA under domestic businesses, it is also negatively signed in all equations. The primary explanation for the lack of a significant correlation between credit access and exposure to LSLA under foreign firms is that those who obtain credit lack the ingenuity to challenge deals made by powerful foreigners who are mostly corporations, governments, World Bank, International Monetary Fund (IMF), and Financial Institutions [45], with the support of conventional and state authorities.

On the other hand, only DE and IDE to LSLA under foreign businesses are highly related to the village with access to water resources, which is likewise positively marked as expected. According to this, households in villages having access to water resources are more likely to be exposed to LSLA directly and indirectly under foreign businesses than under domestic entities. The fact that domestic entities primarily depend on rainwater for production after the acquisition is one explanation for this conclusion. This is consistent with studies that showed that water availability influences investor participation in LSLA [43,[82], [83], [84]].

In line with expectations, soil fertility is also favorably correlated with both DE and IDE to LSLA under both domestic and foreign organizations. This implies that households with productive soil plots are more likely to be exposed to LSLA directly and indirectly through domestic and international entities. Those who claim that LSLA only happens in marginal areas disagree with this conclusion [46].

6.2. Farm investments

Table 4 presents the findings of the impact of DE and IDE from LSLA on household farm investment under domestic and foreign enterprises. The LTI and STI may be interdependent (compliments or alternatives), as was noted in section (5.2). As a result, using the 2SCML, the second stage multivariate probit model included the DE and IDE variables as well as their associated residuals from the first stage linear probability model. The results of the first stage are shown in Table A2 of the appendix.

Table 4.

Multivariate probit estimates of determinants of farm investments.

Variable	Under exposure to LSLA by domestic entities				Under exposure to LSLA by foreign entities
	LTI		STI		LTI		STI
	Coef.	SE	Coef.	SE	Coef.	SE	Coef.	SE
DE	−0.87	0.23***	0.43	0.13***	−0.75	0.22***	0.05	0.02**
IDE	−0.76	0.01***	0.99	0.41**	−0.29	0.11**	0.08	0.02***
Residual_DE	0.51	0.21**	−0.66	0.34*	−0.08	0.02***	0.04	0.02**
Residual_IDE	−0.45	0.03***	0.33	0.01***	−0.07	0.02***	0.05	0.02**
HH_income	0.27	0.03***	0.30	0.02***	0.09	0.02***	0.02	0.03
Fertilizer subsidy	0.05	0.12	0.12	0.03***	0.04	0.22	0.05	0.01***
Gender	0.54	0.05***	0.65	0.02***	0.55	0.14***	0.63	0.26***
Age	0.10	0.13	−0.28	0.23	−0.20	0.22	0.15	0.19
Household size	0.01	0.12	0.02	0.40	−0.01	0.04	0.01	0.06
Education	0.11	0.04***	0.01	0.00***	0.01	0.00***	0.01	0.00***
Farm size	−0.41	0.98	−0.34	0.61	−0.21	0.42	−0.22	0.52
Compensation	−0.61	0.57	−0.11	0.31	0.10	0.12	0.03	0.02
FBO membership	0.13	0.05**	0.27	0.02***	0.78	0.05***	0.58	0.05***
Road	−0.27	0.35	−0.12	0.22	−0.30	0.34	0.17	0.26
Credit	−1.21	0.05***	−0.61	0.02***	−0.21	0.04***	−0.49	0.02***
Water source	0.12	0.92	0.07	0.15	0.10	0.34	0.03	0.17
Good fertile	−0.02	0.03	−0.21	1.00	−0.40	0.20	−0.01	0.05
Moderate fertile	−0.06	0.63	−0.22	0.50	−0.10	0.12	−0.20	0.22
_cons	0.66	0.08***	−0.43	0.04***	5.10	0.08***	0.81	0.07***
${\rho }_{LTISTI}$	−0.87	0.08***	−0.84	0.10***	−0.97	0.11***	0.87	0.14***
Joint significance of location variables: χ2 (6)	61.69***		92.72***		32.24***		38.06***
χ2-statistic for joint significance of residuals	3.23***		2.45***		2.27 ***		4.23***
Joint significance of instruments	1.49 (0.47)		7.44 (0.11)		5.69 (0.55)		10.63 (0.10)
No. of observations	400				508

Notes: ***, ** and * indicate statistical significance at 1%, 5% and 10%.

Source: Author's Computation from Field Survey, 2018

As can be seen in Table 4, every estimated correlation coefficient is negative and significant at 1%, indicating that unobserved variables involved in each investment option are significant and negatively related. This finding supports the idea that modeling investment decisions jointly as opposed to separately is more effective. These imply that STI in inputs like NPK, urea, and ammonia and LTI in irrigation or SWCT are interchangeable. Furthermore, the exclusion restriction test is not rejected by Chi-square statistics for the joint significance of the instrument.

Additionally, residuals for DE and IDE are statistically significant, indicating simultaneity bias and the possibility that the coefficients of DE and IDE would not have been the same without the use of 2SCML [41]. The Chi-square statistics for the joint Wald tests on the vector of these residuals from the first-stage estimations are also displayed in Table 4. These results show that the null hypothesis that the residuals are collectively equal to zero is rejected for each investment equation. These support the findings of the various t-statistics that point to simultaneity bias. The results showed that DE and IDE to LSLA under both local and foreign businesses are significant at 1% in the equations for STI and LTI. This raises the question of how exposure to LSLA affects households' farm investment. These factors, while harming LTI, have a beneficial impact on STIs. Responses from focus group discussions indicated that the area has been subsidizing STIs like NKP, which is why investment has increased. The results are consistent with the expectation that DE and IDE to LSLA under domestic and foreign corporations will reduce the household's probability of making LTI while increasing its probability of making STI. The findings add to the evidence that farmers will favor STI over LTI if they are evicted by LSLA [12,85].

In addition, income has a positive sign and a considerable impact on LTI and STI. This implies that wealthy households stand the chance of increasing both LTI and STI under LSLA by domestic and international businesses. These findings agree with that of Navarro-casta et al. [86] in Peru where income was found to play an important role in farm investments.

The household involvement in the fertilizer subsidy program, however, is favorable and significantly associated with only STIs like NPK, urea, and sulphate of ammonia. This shows that participation in the programme will increase investment solely in STI. The fertilizer subsidy programme does not cover the cost of LTIs, and farmers lack the necessary resources to carry out such investments, which is the fundamental explanation for the insignificant effect of participation on LTI. This is consistent with the study of Abdallah et al. [35] in which fertilizer subsidy was found to be significantly related to NPK and Urea in the area.

The gender of the household head had a positive and significant impact on both LTI and STI, indicating that male-headed households are more likely to invest in LTI and STI than their female-headed counterparts. These results conform with the findings of Navarro-casta et al. [86] in Peru where gender played a significant role in the investment in LTI and LTI.

Additionally, there is a significant relationship between education and FBO membership and the types of farm investments. These results suggest that households with social networks and higher levels of education are more likely to increase their probability of making both LTI and STI. The findings are conceivable because, according to Lapar and Pandey [87], schooling is believed to improve a household's capacity for gathering, processing, and applying any available information on farming and investing. Households with higher levels of education have more access to information about land-improving technology and are consequently more likely to invest in these techniques. In a similar vein, access to information about investments and production is generally improved by a farmer's social network. Along with providing knowledge, these organizations also engage in group labor, where members take turns contributing the labor needed to fund LTI and STI.

According to Boahene et al. [88], credit promotes resource mobilization and investment in both LTI and STI. Therefore, it was predicted that households with access to credit would engage in both LTI and STI. Unexpectedly, access to finance is important but did not show the expected sign. This shows that credit-accessible households are less likely to make both LTI and STI. Following up during focus group discussions revealed that most farmers lack land and instead use credit for various purposes away from mobilizing money for farming. Additionally, all the indicators of soil fertility showed negative trends, but they did not significantly affect investments. This partially explains why investments in boosting soil fertility do not increase on highly fertile plots.

According to the empirical analysis, a specific household's LTI and STI vary dramatically depending on its DE and IDE to LSLA through local and foreign firms. Thus, this supports the theoretical hypotheses that increased tenure insecurity is caused by a scarcity of land or a rate of evictions, which forces farmers to prioritize STI over LTI (e.g. Refs. [12,85]). The findings, however, go against recent research that suggested LSLA would encourage local communities to invest more in farm technologies [22]. The findings also indicate the significance of household and institutional characteristics in farm investment by identifying income, fertilizer subsidy, gender, education, FBO membership, and credit access as significant determinants of STI and STI. Our findings further support the idea that LTI will reduce a household's probability of DE and IDE from LSLA under domestic and foreign businesses [35]. As a result, the study's hypothesis, and the theoretical claim of simultaneity between LSLA and farm investment are both supported. Individuals' LTI and STI also help to reduce tenure insecurity or the rate of evictions. In addition, households' vulnerability to LSLA under domestic and foreign organizations is significantly influenced by factors like gender, leadership position, educational attainment, access to land institutions, FBO participation, the availability of water resources, credit availability, and soil fertility. Consequently, these demonstrated the significant contributions of household, institutional, and regional variables to LSLA.

7. Conclusions

This study looked at how LSLA and household agriculture investments are related in northern Ghana. The study specifically examines the connections between households' investments in LTI and STI and DE and IDE to LSLA under domestic and foreign corporations. A total of 664 homes were selected for the analysis using a multi-stage selection procedure, and analyses were conducted using the 2SCML. A conceptual model created for the interaction between LSLA, and farm investment served as the basis for this investigation. Based on Feder and Onchan's [12] model of the nexus between land tenure insecurity and farm investment, which contends that a lack of available land worsens tenure insecurity and forces farmers to place a higher priority on immediate agricultural investments than LTI, this model was created. The findings showed that both DE and IDE to LSLA under local and foreign businesses deplete LTI in irrigation and SWCT but boost STI. The findings also showed that both investment types reduced the probability of households' DE and IDE to LSLA from domestic and international firms. Overall, the findings point to a bidirectional relationship between agricultural investment and DE and IDE from LSLA by both domestic and international companies. While LSLA depletes investment, the opposite causality is also conceivable, where investment lowers the possibility that the household will be subjected to LSLA. Therefore, households in northern Ghana are more likely to select STI over LTI because of acquisitions by domestic and foreign organizations. Additionally, households in northern Ghana that invest in both LTI and STI are more likely to escape being evicted by both domestic and foreign organizations. These results have serious implications for sustainable development in Ghana. This is because the long-standing wheels of sustainable development including increased production and farm income, food security, poverty, and hunger reduction, depend on farm investments. Thus, the depletion of some farm investments may lead to decreased production and farm income, thereby lowering food security and efforts to achieve sustainable development. This may also affect efforts towards poverty and hunger reduction and thus, dumped the efforts for sustainable development.

The results also have strong policy implications for the adoption of land-improving techniques, land tenure and property rights, interests of investors and well-being of local communities, and land tenure security. Thus, policymakers should recognize and promote the value of diversified land management strategies. Supporting both short-term and long-term land improvement techniques can contribute to sustainable land use and resilience against eviction pressures. Policymakers can acknowledge and recognize the value of households employing both short-term and long-term land improvement techniques. This recognition can inform policy frameworks that encourage a balanced approach to land management. The government can strengthen land tenure and property rights to provide greater security for households, encouraging long-term investments in land improvement. The government can also strengthen and enforce land tenure security to protect the rights of households investing in land improvement. Secure land tenure can provide a foundation for long-term investments and reduce the risk of eviction. Policymakers can review and update existing land laws and regulations to ensure that they protect the rights of local communities, especially in the context of large-scale land acquisitions. Several guidelines exist at the local (e.g. Refs. [38,89]) and international level [90] for large-scale land acquisition. Thus, monitoring and regulatory frameworks should be strengthened by the government to ensure that large-scale land acquisitions adhere to sustainable and socially responsible practices stipulated in the guidelines. Regulations should consider the importance of diversified land management in mitigating eviction risks. Government and policymakers should encourage responsible and sustainable partnerships between local communities and domestic or foreign organizations involved in land acquisitions. Partnerships should prioritize the well-being of local communities and the adoption of diversified land improvement techniques. Promotion of community-led development initiatives that empower local communities to make decisions about land use, investment strategies, and resource management can also be initiated by government and policymakers. This can help ensure that development aligns with the needs and aspirations of the communities. Nonetheless, the results must be interpreted with caution since the study's focus was only on the northern part of Ghana. Moreover, the study used cross-sectional data so the results cannot be interpreted to mean that the effect is over time.

CRediT authorship contribution statement

Abdul-Hanan Abdallah: Writing – review & editing, Writing – original draft, Validation, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Michael Ayamga: Writing – review & editing, Validation, Supervision, Methodology, Funding acquisition, Conceptualization. Joseph A. Awuni: Writing – review & editing, Validation, Supervision, Project administration, Methodology, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix

Tables A1.

First-stage estimations of drivers of farm investment.

Variable	Under exposure to LSLA by domestic entities		Under exposure to LSLA by foreign entities
	LTI	STI	LTI	STI
Leadership position	0.41 (0.29)	0.54 (0.27)**	0.06 (0.29)	0.34 (0.34)
Gender	0.00 (0.01)	0.00 (0.01)	0.01 (0.01)	−0.01 (0.01)
Age	0.09 (0.03)***	0.11 (0.03)***	0.08 (0.03)***	0.05 (0.03)
Household size	0.01 (0.02)	0.01 (0.02)	0.00 (0.03)	0.02 (0.03)
Education	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)
Farm size	0.05 (0.20)	0.16 (0.18)	0.26 (0.19)	0.19 (0.24)
Knowledge	0.35 (0.27)	0.52 (0.25)**	0.54 (0.26)**	0.19 (0.31)
FBO membership	0.01 (0.02)	0.03 (0.01)**	0.02 (0.02)	−0.01 (0.02)
Road	−0.11 (0.03)***	−0.03 (0.03)	0.04 (0.03)	0.04 (0.04)
Credit	−0.05 (0.28)	0.14 (0.27)	0.03 (0.29)	0.31 (0.32)
Water source	0.00 (0.00)*	0.00 (0.00)**	0.00 (0.00)**	0.00 (0.00)
Good fertile	0.01 (0.01)*	0.01 (0.01)**	0.01 (0.01)*	0.02 (0.01)**
Moderate fertile	0.21 (0.24)	0.22 (0.22)	0.07 (0.23)	0.07 (0.28)
Sagnarigu	−1.89 (0.41)***	−0.01 (0.32)	0.81 (0.30)***	0.46 (0.35)
Mion	−1.08 (0.25)***	−0.45 (0.21)**	−0.50 (0.23)**	−0.86 (0.30)***
Central Gonja	0.29 (0.20)	0.13 (0.18)	−0.27 (0.19)	0.00 (0.23)
Savelegu	−2.39 (0.28)***	−2.20 (0.27)***	−2.28 (0.27)***	−1.76 (0.31)***
Yagba-Kubori	−1.63 (0.30)***	−0.46 (0.25)*	−0.24 (0.24)	−1.60 (0.38)***
Constant	0.84 (0.60)	0.32 (0.58)	0.07 (0.59)	−0.52 (0.78)

Notes: *** Denotes significance level at 1%; ** Denotes significance level at 5%; and * Denotes significance level at 10%. Standard errors are in parentheses.

Tables A2.

First-stage estimations of drivers of LSLA.

Variable	Under exposure to LSLA by domestic entities		Under exposure to LSLA by foreign entities
	DE	IDE	DE	IDE
Leadership position	0.37 (0.37)	0.55 (0.43)	1.30 (0.42)***	1.36 (0.40)***
Gender	0.02 (0.01)**	0.02 (0.01)**	0.05 (0.01)***	0.03 (0.01)**
Age	0.08 (0.03)**	0.16 (0.03)***	0.23 (0.04)***	0.28 (0.04)***
Household size	0.03 (0.03)	0.01 (0.04)	0.03 (0.03)	0.00 (0.05)
Education	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)***	0.00 (0.00)***
Farm size	−0.10 (0.24)	0.43 (0.30)	−0.21 (0.33)	0.25 (0.37)
Land institution	−1.21 (0.40)***	−0.07 (0.34)	−0.54 (0.42)	−0.19 (0.42)
FBO membership	0.00 (0.03)	0.00 (0.03)	−0.04 (0.04)	−0.04 (0.04)
Road	0.05 (0.04)	0.02 (0.05)	0.07 (0.06)	0.08 (0.06)
Credit	0.32 (0.36)	0.35 (0.39)	0.53 (0.41)	1.48 (0.53)***
Water source	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)	0.00 (0.00)
Good fertile	0.02 (0.01)**	0.01 (0.01)*	0.01 (0.01)*	0.01 (0.01)
Moderate fertile	0.07 (0.29)	−0.41 (0.36)	−0.20 (0.47)	−0.47 (0.51)
Sagnarigu	0.97 (0.38)**	0.89 (0.43)**	1.40 (0.52)***	−0.32 (0.76)
Mion	−0.83 (0.33)**	−0.50 (0.39)	−2.90 (0.84)***	−2.37 (0.77)***
Central Gonja	0.45 (0.25)*	−0.43 (0.28)	0.05 (0.33)	−0.55 (0.35)
Savelegu	−1.86 (0.33)***	−2.07 (0.35)***	−1.46 (0.42)***	−1.93 (0.49)***
Yagba-Kubori	−0.22 (0.30)	−0.09 (0.35)	−0.43 (0.39)	0.24 (0.34)
Constant	−2.19 (0.81)***	−2.69 (0.88)***	−5.80 (1.10)***	−5.25 (0.94)***

Notes: *** Denotes significance level at 1%; ** Denotes significance level at 5%; and * Denotes significance level at 10%. Standard errors are in parentheses.

Appendix A. Supplementary data

The following are the Supplementary data to this article.