Determinants of savings frequency among tomato farmers in Ghana

Abstract

Multinomial logistic regression was employed to identify the determinants underlying the respondents’ frequency of savings—weekly, monthly and seasonally. The results of the study showed that amount saved per period, number of years of education and engagement in non-farm income generating activities significantly influenced farmers’ savings frequencies. The findings are quite significant as they take the decision to save beyond the suggested two-stage sequential process to include a third stage, which is the time horizon of savings. The findings revealed that rural households are predisposed to extend their savings time horizon by holding onto their surplus funds in order to retain some capacity for their present consumption and other needs before thinking of saving. On the other hand, the anticipation of a gloomy future as a result of bad harvest for instance may induce the rural householder to shorten his/her savings time horizon, that is, reduce the time-frame s/he holds onto surplus funds and quickly save such funds. Finally, the study suggests education as a catalyst to create a desirable behaviour of saving “now” (weekly or monthly) rather than procrastinating savings to the “future” that is, saving seasonally.

Keywords:

• Frequency
• savings
• multinomial regression
• tomatoes farmers
• Ghana

JEL Classification:

• Q12
• E2
• O17

PUBLIC INTEREST STATEMENT

Much as savings plays an important role in economic development process, it has been neglected very much in favour of credit in rural communities particularly in developing countries. Against this backdrop, the study sought to determine tomato farmers’ frequency of saving based on the notion that these farmers have the capacity to save and make conscious effort to save. From the perspective of economists, income is partitioned between consumption and savings. Hence, all households regardless of their income levels would have a certain portion of income being used to take care of immediate consumption and a portion channelled into savings. It is generally assumed that individuals’ or households’ partitioning of income into consumption and savings depends on the relative importance of their immediate consumption requirements in relation to their future goals. For this reason, the economic perspective on savings takes into consideration the time dimension of partitioning of income into consumption and savings which brings in the idea of the frequency of saving.

1. Introduction

Much as savings plays an important role in economic development process, it has been neglected very much in favour of credit in rural communities particularly in developing countries. Fundamentally, it is acknowledged that economic variables such as technological progress, education, institutional development, sound domestic policies, proper resource management and favourable external economic environment play important roles in economic growth and development of a nation. However, it is generally believed that sustained growth and development are difficult to be attained and maintained without savings (Akaah et al., 1987; Bautista & Lamberte, 1990). This is because economic growth and development which lean much on the afore-mentioned economic variables cannot be long sustained under conditions of declining savings rates thereby given credence to Lewis’ (1954) famous dictum on savings—raising the savings rate is the central problem in economic development (Bautista & Lamberte, 1990; Gersovitz, 1988).

1.1. Importance of savings to rural households

Rural households including tomato farmers are vulnerable to a large number of uncertainties and risks related to diseases, conflicts and climatic changes especially erratic rainfall frequency which in most cases affect agricultural production—the main stay of the rural economy (Aidoo-Mensah, 2017). However, certain risk mitigation actions can be employed to help overcome or prevent some if not all of these risks. Such actions may include preventative health care systems, free medical care, subsidies on basic goods and services, provision of food hand-outs and public support arrangements such as food for work programmes (Hoogeveen et al., n.d.). Notwithstanding the fact that these risk mitigation measures are important to help rural households to cope with risks and uncertainties, they have the inherent tendency of creating dependency syndrome among rural households (Aidoo-Mensah, 2020). On the other hand, establishment of reliable and appropriate safety nets such as promoting savings habits among rural households can enable them handle some or all these risks and uncertainties on their own with little or no external assistance. Thus, savings can be relied upon as an important tool of improving well-being, insuring against times of shocks, and providing a buffer to help people particularly rural households cope in times of crisis with little or no external assistance (Miracle et al., 1980; Rutherford, 1999; Zeller & Sharma, 2000). According to De Laiglesia and Morrisson (2008), besides increasing investment rates in less developed countries, savings is a fundamental tool in the task of lifting rural households to a more sustainable and faster growth development path. Moreover, savings particularly at the household level is needed to finance capital (both physical and human) formation in order to increase output and wellbeing of rural households in developing countries (Bautista & Lamberte, 1990).

Furthermore, relatively underdeveloped financial systems existing in the rural sector mean that accumulation of financial resources through savings is often the only way to acquire productive capital or wealth that can be passed on to future generations (De Laiglesia & Morrisson, 2008). Apart from its direct contribution to output growth, savings also makes capital accumulation possible enabling the employment of complementary production inputs and serves as a vehicle for the adoption of improved technology (Bautista & Lamberte, 1990).

Slumps in income or income shocks can have dire consequences for most rural households especially those that are struggling to eke out a living. Even some households whose income may be adequate, on the average may face transitory food insecurity or the threat of it. Hence, savings are needed to sustain adequate consumption levels especially in the event of food shortages among rural households. For the poorest households, one large shock or frequent runs of food shortages can lead to major decrease in food consumption, which can lead to permanent disability, especially of children. Accordingly, the poorer, more risk-averse, and vulnerable a household is, the more important precautionary saving becomes a risk easing measure to such a household (International Food Policy Research Institute, 2002).

According to Fernando (1991), promotion of savings habits among rural households and subsequent mobilisation of savings from them by formal financial institutions would result in an improvement in rural income distribution. Moreover, it is expected that promotion of savings habits among rural households and subsequent mobilisation of savings from them by formal financial institutions would enable such formal institutions to improve their financial viability and overall performance in many possible ways (Vogel & Burkett, 1986; Vogel, 1984). One of such ways is that it will enable participating financial institutions gather valuable information on their existing as well as potential borrowers, leading to lower costs on loans to the rural sector (Fernando, 1991). Such valuable information obtained on rural households in the opinion of Fernando (1991), will have positive effect on loan delinquency and reduce risk premium, hence the cost of lending. Moreover, such an effort of extending valuable savings service to rural households by formal financial institutions can be a good step to integrate rural households into mainstream financial systems.

1.2. The tomato industry in Ghana

The tomato industry is a thriving agribusiness activity predominantly undertaken in the savanna and forest-savanna transitional belts of Ghana. It serves as good nutritional balance to farm families and its production helps to boost their income and hence their standard of living. The crop is grown on a large scale in such areas as Tono and Vea areas in the Upper East region; Akumadan, Kumawu and Agogo areas in the Ashanti region; Wenchi, Awisa, Yamfo, Abesim, Techiman, Ofuman, Derma and Techimantia areas in the then Brong Ahafo region and other areas such as Akim Oda, Nsawam, Suhum, Oyoko in the Eastern region (Adu Dapaah & Oppong-Konadu, 2002). These are all communities located in the savanna and forest-savanna transitional belts of Ghana.

In spite of the fact that the agribusiness activities associated with the tomato industry plays an important role in the financial and nutritional well-being of most farm families in Ghana, production of the crop has not been promising over the years (Adu Dapaah & Oppong-Konadu, 2002). Fundamentally, this has been attributed to the failure of the crop to reach its potential in terms of attaining yields comparable to other countries, its inability to sustain processing plants and also its inability in improving the livelihoods of households involved in tomato industry (Robinson & Kolavalli, 2010). It is therefore not surprising that Ghana continues to import several tonnes of tomato and tomato products into the country and the nation has been observed to be second only to Germany as the largest importer of tomato paste, consuming an average of twenty five thousand (25,000) tonnes of tomato paste in a year at a total cost of about $25 million dollars (Yeboah, 2011).

The emphasis of attempts by various stakeholders in the tomato industry in Ghana in finding solutions to the many problems associated with the industry has mostly been looked from the agronomic perspective. However, there is no gainsaying of the fact that tomato farmers’ quest for survival now and into the future in today’s ever-changing and challenging environment of economic development hinges not only on agronomic issues but also on their ability to sustain their production activities through their earnings, vis-à-vis, their saving (Aidoo-Mensah, 2018).

This is the reason why that rural households in general self-finance their economic activities basically from their earnings and savings. Therefore, a better understanding of the dynamics of the savings behaviour, vis-à-vis, and the frequency of such savings at the household level will help in the formulation of appropriate policies for savings mobilisation, thereby improving upon local capital formation capacity to enhance national development. This is expected to reduce Ghana’s importation of tomato products thereby conserving the nation’s scarce foreign exchange reserves and also provide employment and development opportunities in the rural communities of the country (Yeboah, 2011).

In furtherance to the above, understanding the savings frequency of these farm households can help make the necessary improvements particularly in savings mobilisation and its subsequent investment into crucial sectors of the economy. This is because, evidence abounds that mobilisation of domestic savings at the rural household level can contribute significantly to economic development of low-income countries by way of capital formation for micro-enterprise development and also improve agricultural lending. Thus, a better knowledge of savings frequency at the farmer level could help develop the potential to finance local investments by relying on locally generated resources rather than near-total dependence on external resources with their stringent conditions.

1.3. Problem statement and justification

In recent times, there has been an upsurge of interest among development economists, governments and international donors to increase domestic savings in developing countries. This is because domestic savings are seen as a probable alternative to capital formation for investments, by concentrating attention and efforts on savings mobilisation from the household sector particularly rural households. According to Chowdhury (1987), the focus on household savings is dictated by the fact that the corporate and government sectors are undoubtedly quite minor actors in the context of domestic savings as the case is in developing economies. Therefore, in most developing countries where the rural sector is the most dominant, it can be conjectured that any effort to raise the level of domestic savings may have to begin from there.

However, a large number of developing countries have been observed to be unable to take advantage of the savings potential of the household sector. Many reasons have been attributed to this inability of which lack of proper and in-depth understanding of the operations and dynamics of the frequency of rural household savings has been cited as paramount. Generally, the scarcity of information on the interactions of the rural economy with the larger national economy in most developing countries makes it difficult for government and other stakeholders such as financial institutions whose aim is to mobilize savings (Bersales & Mapa, 2006) to fully understand this sector of the economy. It therefore stands to reason that a good understanding of the frequency of financial savings by rural households particularly farm families may have valuable financial implications for financial intermediation particularly in an effort to foster adequate integration of savings and investment programmes into developing strategies capable of improving resource allocation (Shitu, 2012).

These considerations become more important as it is assumed that rural enterprises which are largely agriculture in nature with their marginal economic activities tend to result in low incomes which are also seasonal in nature, therefore a disincentive to promote savings mobilisation among such households. Moreover, since farm families may be scattered over relatively wide geographical areas with relatively small amounts saved per a given period, it implies high transaction costs for financial intermediaries particularly for formal financial intermediaries (Aryeetey & Nissanke, 2005). The bottom-line of all these considerations is that costs outweigh the benefits of rural savings mobilisation; for that reason it is of no paramount importance to them to promote savings among rural households. It is therefore not surprising that most formal financial intermediaries discriminate against operating in such environments. Therefore, it is envisaged a better understanding of the savings frequencies of rural households particularly farmers will contribute to the formulation of appropriate policies for savings mobilisation aimed at costs recovery for operators in the financial intermediation efforts (Shitu, 2012), thereby improving upon local capital formation capacity at the rural level. Moreover, empirical studies on savings behaviour particularly frequencies of savings among tomato farmers in Ghana appears to be lacking, thus, the study aims at contributing to the savings literature on what influences the savings frequencies of rural households particularly tomato farmers.

2. Materials and methods

Purposive sampling method was utilized to select three regions in Ghana, namely Ashanti, Brong Ahafo and Upper East. Two districts from each of these regions were purposively selected as well. The selection of the 3 regions and their respective districts took into consideration the volume of tomato production based on official statistics from the Ministry of Food and Agriculture (MoFA). For instance the PPMED of MOFA (1997) record indicates that the Ashanti and the Brong Ahafo regions together contribute 43 per cent of the total tomato produced in the country. A third stage of the sampling involved simple random sampling procedure after identifying and listing on paper all tomato farmers in the various operational areas. This was undertaken with the help of the Agricultural Extension Agents (AEAs) in charge of the operational areas in each of the selected districts.

Using an estimation method based on Bartlett et al. (2001), a total of 562 tomato farmers were selected for the study. However, 496 of the selected respondents indicated that they make conscious effort to save as indicated on Table 1 and therefore qualified for the study.

Table 1. Distribution of conscious effort to save among respondents

Conscious effort to save	N	%
Make conscious effort to save	496	88
Do not make conscious effort to save	66	12
Total	562	100

Source: Field Survey, 2018

2.1. Analytical framework

The frequency of financial savings by rural households particularly farm families has serious financial implications for financial intermediaries. This is mainly due to the fact that farm families may be scattered over relatively wide geographical areas with relatively small amounts saved per a given period which implies high transaction costs for financial intermediaries particularly formal ones (Aryeetey & Nissanke, 2005). Moreover, due to the agriculture-based nature of the rural livelihood activities, incomes tend to be irregular and can be subjected to environmental (weather) and market forces (demand and supply) (Osei-Boateng & Ampratwum, 2011). Under such circumstances, majority of the rural households have been found to live with high sense of income insecurity which makes them to hold on to their incomes and only save when they deem it convenient to do so. It is therefore not surprising that most formal financial intermediaries discriminate against operating in such environments.

Frequency of saving effort among rural households is primarily underlined by the type and number of economic activities engaged in by the household. In a situation whereby income is earned on seasonal basis or the household is engaged in only one income generating activity, the frequency of savings may be compromised. Farmers’ savings preferences are hypothesized as a sequential decision-making process made up of 3 stages:

Stage 1 – At this stage, the tomato farmer makes a conscious effort to save which is assumed results in a decision to save. This implies that the tomato farmer has visualised and looked into the future to see what s/he stands to gain in postponing present consumption in order to save, that is, the benefits to be gained in saving.

Stage 2 – At this stage, the farmer has to make a choice of the mode of savings between formal and informal modes of savings. The selection of the mode of savings is underlined by the perceived security/safety of one’s savings. This selection criterion is based on the fact that savings for rural households are perceived as a risky venture (Berjan et al., 2014). Savings for instance in the form of physical assets such as gold may be prone to theft, cattle may be subject to diseases, accidents and theft; savings with rotating savings and credit associations may be subject to unsound management. Hence, rural savers give the highest priority to the security of their savings when deciding the mode of savings (Buchenau, 2003).

Stage 3 – At this stage, which is the frequency of saving, the decision of the farmer who makes a conscious effort to save and who might have settled on the mode of saving is conditioned on 3 forms of savings frequencies – weekly, monthly and seasonally. All things being equal, the choice of frequency of saving is more likely to be tied to the period when they earn their income from tomato cultivation unless they are engaged in other income generating activities. Hence, it is hypothesized that frequency of earning of income plays a key role in the decision on the choice of frequency of saving. In order to investigate the determinants of the decision at Stage 3, multinomial logistic (MNL) regression is employed.

2.2. Model specification

Multinomial logistic regression models are made of a set of coefficients (${\beta }^{\left(1\right)},{\beta }^{\left(2\right)},{\beta }^{\left(3\right)}$) which are estimated as:

(1) $\mathit{P}\mathit{r}\left(\mathit{Z}=1\right)=\frac{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}}{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}+\mathrm{\ell }\mathit{x}{\beta }^{\left(2\right)}+\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}\hspace{0.17em}$(1)

(2) $\mathit{P}\mathit{r}\left(\mathit{Z}=2\right)=\frac{\mathrm{\ell }\mathit{x}{\beta }^{\left(2\right)}}{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}+\mathrm{\ell }\mathit{x}{\beta }^{\left(2\right)}+\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}\hspace{0.17em}$(2)

(3) $\mathit{P}\mathit{r}\left(\mathit{Z}=3\right)=\frac{\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}+\mathrm{\ell }\mathit{x}{\beta }^{\left(2\right)}+\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}\hspace{0.17em}$(3)

However, there exist more than one solution to ${\beta }^{\left(1\right)},{\beta }^{\left(2\right)},{\beta }^{\left(3\right)}$ that leads to the same probabilities for $Z=1,Z=2,Z=3.$ The model is, thus, unidentified. Therefore, in order to identify the model, an important assumption is made, that is, one of the categories is set aside as a base category and is arbitrarily equated to 0. Since there is no ordering among the response variables—weekly, monthly and seasonally any of the categories can be taken as the base category.

Assuming ${\beta }^{\left(2\right)}$ is equated to 0, then the remaining coefficients ${\beta }^{\left(1\right)}and{\beta }^{\left(3\right)}$will measure the change relative to $Z=2$ (that is, savings done on monthly basis). Simply put, comparison is thus, being made between savings done on monthly basis with other choice of savings frequency—weekly and seasonally. Setting${\beta }^{\left(2\right)}=0$, the above equations become:

(4) $\mathit{P}\mathit{r}\left(\mathit{Z}=1\right)=\frac{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}}{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}+1+\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}\hspace{0.17em}$(4)

(5) $\mathit{P}\mathit{r}\left(\mathit{Z}=2\right)=\frac{\mathrm{\ell }\mathit{x}{\beta }^{\left(2\right)}}{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}+1+\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}\hspace{0.17em}$(5)

(6) $\mathit{P}\mathit{r}\left(\mathit{Z}=3\right)=\frac{\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}{\mathrm{\ell }\mathit{x}{\beta }^{\left(1\right)}+1+\mathrm{\ell }\mathit{x}{\beta }^{\left(3\right)}}\hspace{0.17em}$(6)

The relative probability of $Z=1$ to the base category is given as:

(7) $\frac{\mathrm{P}\mathrm{r}\left(Z=1\right)}{\mathrm{P}\mathrm{r}\left(Z=3\right)}={\mathrm{\ell }}^{x{\beta }^{\left(1\right)}}$(7)

Equation (7)(7) $\frac{\mathrm{P}\mathrm{r}\left(Z=1\right)}{\mathrm{P}\mathrm{r}\left(Z=3\right)}={\mathrm{\ell }}^{x{\beta }^{\left(1\right)}}$(7) may be termed as the relative likelihood and it assumes that $X$ and ${{\beta }_k}^{\left(1\right)}$ are vectors equal to $X_1,X_2\dots ..X_k$ and ${{\beta }_1}^{\left(1\right)},\hspace{0.17em}{{\beta }_2}^{\left(1\right)}\dots \dots .{{\beta }_k}^{\left(1\right)}$ respectively. The ratio of relative likelihood for one unit change in $X_i$ relative to the base category is given as:

(8) $\frac{e^{{{\beta }_1}^{\left(1\right)x_1}}+{{\beta }_i}^{\left(1\right)(x_{\left(i+1\right)}}\dots \dots .+{{\beta }_k}^{\left(1\right)x_k}}{e^{{{\beta }_1}^{\left(1\right)x_1}}+{{\beta }_1}^{\left(1\right)(x_{\left(i+1\right)}}\dots \dots .+{{\beta }_k}^{\left(1\right)x_k}}=e^{{{\beta }_1}^{\left(1\right)}}$(8)

Therefore, the exponential value of a coefficient is the relative likelihood ratio for a unit change in the corresponding variable (StataCorp, 1999).

2.3. Empirical application of the model

Consider the utility to be derived by a tomato farmer $n$ if s/he makes a $j$ choice for the savings frequency $U_{nj}.$ The systematic component of the alternative $j$ is specified as a function of an array socio-economic and demographic characteristic such as respondents’ total income (Income), number of years of education (Yresedn), average amount saved per period (Aveamt) and engagement in non-farm income generating activities (Non-farm). Consequently:

(9) $U_{nj}={\alpha }_{j}Incom{e}_n+{\beta }_{j}Yresed{n}_n+{\gamma }_{j}Aveam{t}_n+{\mathrm{\Phi }}_{j}Nonfar{m}_n+{\epsilon }_{nj}$(9)

Assuming that the errors ${\epsilon }_{nj}$ are independently and identically distributed with an extreme value, the probability that an alternative $j$ is chosen from $j$ alternative sets can be represented by the multinomial logit (MNL) model (McFadden, 1974). Based on the general form of the MNL model, the above can expressed as:

(10) $Prob\left(choice=j|J\right)=\frac{e^{B_j{Z}_{nj}}}{{\sum }_j{e}^{B_j{Z}_{nj}}}=\frac{exp({\alpha }_j{Y}_n+{\beta }_j{A}_n+{\gamma }_j{N}_n+{\mathrm{\Phi }}_j{G}_n+\phi P_n)}{{\sum }_j^J({\alpha }_j{Y}_n+{\beta }_j{A}_n+{\gamma }_j{N}_n+{\mathrm{\Phi }}_j{G}_n+\phi P_n)}$(10)

Where: $n=1\dots \dots .N$ catalogues the number of observations and $j=1\dots \dots .J$ catalogues the frequency of saving (Weekly, Monthly and Seasonally), $Z_n$ represents the vector of explanatory variables (respondents’ total income (Income), number of years of education (Yresedn), average amount saved per period (Aveamt) and engagement in non-farm income generating activities (Nonfarm).

As in other forms of linear regression, multinomial logistic regression uses a linear predictor function $f\left(k,i\right)$ to predict the probability that observation i has outcome k, of the following form:

(11) $f\left(k,i\right)={\beta }_{0,k}+{\beta }_{1,k}{X}_{1i}+{\beta }_{2,k}{X}_{2i}+.\dots \dots \dots \dots +{\beta }_{m,k}{X}_{mi}$(11)

where ${\beta }_{m,k}$ is a regression coefficient associated with the mth explanatory variable and the kth outcome. As in logistic regression, the regression coefficients and explanatory variables are normally grouped into vectors of size M + 1, so that the predictor function can be written more compactly:

(12) $f\left(k,i\right)={\beta }_k.X_i$(12)

Where:

${\beta }_k$= set of regression coefficients associates with outcome k

And

$X_i$= a row vector of the set of explanatory variables associated with observation i.

Thus, the empirical model is given as:

(13) $f\left(k,i\right)={\beta }_0+{\beta }_1\left(Income\right)+{\beta }_2\left(Yrsedn\right)+{\beta }_3\left(Aveamt\right)+{\beta }_1\left(Nonfarm\right)+{\epsilon }_i$(13)

2.4. Definition and measurement of variables

Income – The income of the respondents for the study included their total income from their tomato production activities during both the major and minor tomato seasons in 2021 as well as income from other crops, animals and non-farm activities.

Yrsedn – Years of education indicate the number of years a tomato farmer has had formal education.

Aveamt – This is the average amount a tomato farmer saved at any given time during 2021. This was measured in Ghana Cedis (GH¢).

Nonfarm – Engagement in non-farm is a dummy variable given as one (1) if a tomato farmer engaged in non-farm income generating activity and zero (0) if otherwise.

3. Results and discussions

3.1. Motives for non-saving

As to why they do not make any conscious effort to save, the 12% who indicated that they do not make any conscious effort to save gave varied reasons for their non-saving habit. As indicated in Table 2, “Income too small” was ranked as the most important reason for non-saving. This is not surprising as the most often cited reason for non-saving has been low levels of income, and in particular insufficient disposable income (Kempson et al., 2000). In reference to the respondents who by the nature of their dominant economic activity, tomato cultivation, do earn their income on seasonal basis, small size of income is almost become a seasonal norm. This is because available data suggest that over the past two decades, the tomato sector in Ghana has been stagnant and possibly declining, both in terms of area cropped and yield possibly due to low benefits accruing to the farmers in terms of price levels as farm-gate prices are essentially becoming lower and variable with time (Robinson & Kolavalli, 2010).

Table 2. Ranking of motives for non-saving

MOTIVES FOR NON-SAVING	MEAN	RANK
Income too small	6.151515	1
Too many financial commitments	5.166667	2
Recurring emergencies like sicknesses etc.	4.818182	3
I have a very busy schedule	3.378788	4
No convenient place to save	3.08	5
You cannot deposit small amounts	2.76	6
Inconvenient hours of opening and closing of financial institution	2.651515	7
Number of observations (N)		66
Kendall’s W		0.402203857
Chi-square		159.2727273
Df.		6
Asymptotic significance		0.00000

Source: Field Survey, 2021

“Too many financial commitments” was ranked as the second most important reason for non-saving. It is generally said that the demand of everyday living with its attendant many financial commitments deprive many households and individuals to plan their future by adequately laying aside some money as savings (Dezyk & Slater, 2003).

Many households or individuals in the developing countries with little or no access to insurance to help cope with unpredictable and recurring emergencies such as sicknesses, fires, etc., have been observed to find it difficult saving because the occurrence of these afore-mentioned events tends to have great toll on their finances (Kempson & Finney, 2009). In line with this, the respondents who do not make conscious effort to save ranked “Recurring emergencies like sicknesses etc” as the third most important reason for non-saving.

3.2. Degree of association among non-savers on their reasons for non-saving

In order to examine the degree of agreement among the non-savers on their reasons for non-saving, the Kendall’s coefficient of concordance was employed. Kendall’s W-value of 0.402203857 as seen in Table 2 indicates a reasonable degree of concordance among the 7 items rated by the respondents, and therefore the null hypothesis that there is no agreement among the ratings is rejected at any reasonable level of significance (p < 0.01).

3.3. Motives for saving

As to why the tomato farmers deem it important to save, the 88% who indicated that they make conscious effort to save gave various reasons why it is important for one to save. A ranking analysis method (Table 3) was used in order to understand why these respondents think it is important to save.

Table 3. Ranking of motives for saving

MOTIVES FOR SAVING	MEAN	RANK
To take care of future consumption	11.06855	1
To serve as insurance against emergencies	9.737903	2
To raise capital for investment	9.582661	3
To raise capital for expanding my business	9.913306	4
Social security	8.945565	5
Meet basic household needs	8.949597	6
To start a business	7.653226	7
To control expenditure	7.381048	8
Financial freedom	2.453629	9
Safe custody of wealth	3.419355	10
To buy an expensive item	4.177419	11
Social reasons (Weddings, bride price etc.)	4.096774	12
To pay back loan	3.620968	13
Number of observations (N)		496
Kendall’s W		0.597081285
Chi-square		3553.827809
Df.		12
Asymptotic significance		0.00000

Source: Field Survey, 2021

The result as depicted in Table 3 indicates that “Taking care of future consumption” is the most important motive respondents think is the reason one has to save. The choice of “taking care of future consumption” as number one motive to save is in line with life-cycle hypothesis of consumption (or LCH model). The model defines consumption pattern of an individual from one’s early life till retirement and ultimately death. In other words, individuals are assumed to plan a lifetime pattern of consumer expenditure based on expected earnings over their lifetime.

According to the model, early in one’s life consumption expenditure may exceed income as the individual may be making major purchases like buying a new home, starting a family, and beginning a career. At this stage in life, it is hypothesized that the individual will borrow from the future to support these expenditure needs. In mid-life however, these expenditure patterns begin to level off and are supported or perhaps exceeded by increases in income. At this stage the individual repays any past borrowings and begins to save for her/his retirement. Upon retirement, consumption expenditure may begin to decline however income usually declines dramatically. At this stage of life, the individual dis-saves or lives off past savings until death (Modigliani, 1986; Ruby, 2003).

Moreover, the choice of “taking care of future consumption” as number one reason to save is in line with neoclassical economic theory which generally considers consumption to be the ultimate end of economic activity. Hence, households/individuals are deemed rational when their ultimate motive to save is focused on taking care of consumption now and in the future.

It has been observed that many people lack the financial resilience to keep up with demands on their finances particularly in the event of unexpected occurrence. For many of such individuals and households in developing countries where formal insurance systems are not well developed or in some cases totally absent the only means to take care of such shocks is to save towards the shocks. It is therefore not surprising that the respondents ranked “To serve as insurance against emergencies” was ranked the second most important motive to save.

In recent decades, it has been observed that the numerous initiatives of governments as well as development agencies worldwide more particularly in developing countries to provide some form of access to financial services to rural households has not achieved the expected positive impact (Rabobank, 2005). Thus, rural households in most cases have resorted to the creation of sources of funds for investment into such items as farm machinery to ease drudgery through their own personal savings. Therefore, it is not out of place that “To raise capital for investment” was ranked as the third most important motive for making conscious effort.

3.4. Degree of association among savers on their motives for saving

In order to examine the degree of agreement among the savers on their motives for saving, the Kendall’s coefficient of concordance was employed. Kendall’s W-value of 0.597081285 as seen in Table 3 indicates there is about 60% agreement among the respondents in the ranking of the 13 items explaining the motives for saving and therefore the null hypothesis that there is no agreement among the ratings is rejected at any reasonable level of significance (p < 0.01).

3.5. Frequency of saving

Among the 496 respondents who make a conscious effort to save, the majority of them, constituting about 60% indicated that they save seasonally, especially when the price of tomato is good as seen on Table 4. This is not surprising since available data suggest that over the past two decades, the tomato sector in Ghana has been stagnant and possibly declining, both in terms of area cropped and yields possibly due to low benefits accruing to the farmers in terms of price levels as farm-gate prices are essentially becoming lower and variable with time (Robinson & Kolavalli, 2010).

Table 4. Distribution of respondents’ frequency of savings

Frequency of Savings	FREQ	%
Weekly	58	12
Monthly	139	28
Seasonally	299	60
TOTAL	496	100

Source: Field Survey, 2018

3.6. Average amount saved per period

The seasonal nature of the respondents’ income generating activity is such that they can afford to make small deposits at a time. This situation is worsened by the high perishability of tomato, poor market access, and competition from imports which in some cases render some farmers unable to sell their tomatoes, which are then left to rot in their fields (Robinson & Kolavalli, 2010). Thus, in some instances in Ghana, tomato farmers particularly in the three northern regions have been found to be saddled with huge debts as a result of their inability to market their produce (Mari & Knottnerus, 2007).

Table 5 indicates that the majority of the respondents (63.8% in Ashanti Region; about 60.5% in Brong Ahafo Region and 96.2% in Upper East Region) can manage to save up to or less than GH¢500 at a time.

Table 5. Distribution of respondents’ average amount saved per period

Average Amount saved per given period (GH¢)	Ashanti Region (N=116		Brong Ahafo Region (N=223)		Upper East Region (N=157)		All households (N=496)
	N	%	N	%	N	%	N	%
<= 500	74	63.8	135	60.5	151	96.2	360	72.6
501–1000	16	13.8	59	26.5	6	3.8	81	16.3
1001–1500	25	21.6	29	13.0	0	0.0	54	10.9
>1501	1	0.8	0	0.0	0	0.0	1	0.2
TOTAL	116	100.0	223	100.0	157	100.0	496	100.0

Source: Field Survey, 2021

3.7. Empirical analysis of frequency of saving

In order to investigate the determinants underlying the respondents’ choice of their savings behaviour which are weekly, monthly and seasonally, a multinomial regression analysis was done with the results as reported on Tables 6 and 8. The multinomial regression model is specified as below:

$FreqSav={\beta }_0+{\beta }_1\left(Income\right)+{\beta }_2\left(Yrsedn\right)+{\beta }_3\left(Aveamt\right)+{\beta }_1\left(Nonfarm\right)+{\epsilon }_i$

Table 6. Model fitting information

Model	2 Log Likelihood	Chi-Square	df	Sig.
Intercept Only	898.907
Final	605.557	293.350	8	.000

Source: Field Survey, 2018

Table 7. Likelihood ratio tests

Effect	-2 Log Likelihood of Reduced Model	Chi-Square	Df	Sig.
Intercept	605.557	0.000	0
Respondents’ total income (Income)	613.628	8.070	2	.018
Number of years of education (Yrsedu)	622.974	17.417	2	.000
Average amount saved per period (Aveamt)	664.852	59.294	2	.000
Engagement in non-farm income generating activities (Non-farm)	720.868	115.311	2	.000

Source: Field Survey, 2021

Table 8. Parameter estimates of the results of frequency of saving

FREQUENCY OF SAVINGS^a		Coefficient (B)	Std. Error	Wald	df	P-value	Exp(B)	95% Confidence Interval for Exp(B)
								Lower Bound	Upper Bound
WEEKLY	Intercept	−4.825	1.202	16.110	1	.000
	Income	.000	.000	4.656	1	.031**	1.000	1.000	1.001
	Education	.064	.036	3.182	1	.074**	1.066	.994	1.144
	Amount saved	−.005	.001	26.859	1	.000***	.995	.993	.997
	[Engagement=1]	3.592	1.027	12.228	1	.000***	36.312	4.849	271.917
	[Engagement=2]	0^b			0
MONTHLY	Intercept	−5.838	1.148	25.861	1	.000
	Income	.000	.000	6.577	1	.010**	1.000	1.000	1.001
	Education	.116	.029	15.675	1	.000***	1.123	1.060	1.190
	Amount saved	−.003	.001	28.633	1	.000***	.997	.996	.998
	[Engagement=1]	4.749	1.019	21.723	1	.000***	115.433	15.670	850.336
	[Engagement=2]	0^b			0

a. The reference category is: SEASONALLY. b. This parameter is set to zero because it is redundant.***significant at 1%, **significant at 5%, *significant at 10%

Where the dependent variable, that is, frequency of saving (FreqSave) is a choice among weekly, monthly and seasonally. The choice of multinomial regression was based on the fact that the dependent variable used in the model is nominal or categorical. That is, it cannot be ordered in any meaningful. Moreover, there are more than two categories—weekly, monthly and seasonally. Thus, whilst binomial logistic regression has a dichotomous (two) dependent variables, multinomial logistic regression extends the approach for situations where the independent variable has more than two categories.

From Table 6, the distribution reveals that the probability of the model chi-square (293.350) was 0.000 and this is less than the significance level of 0.05 (i.e. p < 0.05). Thus, the null hypothesis that there was no difference between the model without the independent variables and the model with the independent variables was rejected. This suggests that all the independent variables jointly improve the model over the intercept-only model (that is, with no variables added). In other words, the addition of all the independent variables generates a statistically highly significant model.

On the strength of the multinomial logistic relationship given by the pseudo R² measure (Nagelkerke R²), the model explained about 53.2% of the variation in a respondents’ choice of the frequency of savings.

3.7.1. Likelihood Ratio Tests

The results of the multinomial regression model as indicated on Table 7 show that Number of years of education, Amount saved per period and Engagement in non-farm income generating activities are statistically significant at 1% level of significance because p = 0.000 for each of these variables. However, Respondents’ total income (p = 0.018) is statistically significant at 5% level of significance.

3.7.1.1. Parameter Estimates of the results of frequency of saving

From Table 8, for savings done on weekly basis relative to seasonal savings, respondents’ total income is positive and significant. This is the multinomial logit estimate for a unit increase in income, if savings is to be done on weekly basis relative to savings being done seasonally given other variables in the model are held constant.

4.7.2. Respondents’ total income

From the table, it can be seen that if a respondent was to increase his/her total income by one unit, the odds ratio (Exp(B)) of preferring to save weekly to saving seasonally would be expected to increase by a factor of 1.000 (with no difference between the magnitude of the effects). This means that one can confidently accept the hypothesis that the odds ratio in question is 1 (the value expected if there was no effect). More generally, this suggests that if a respondent was to increase his/her total income by one unit, it is expected that he/she would be more likely to be indifferent to save weekly or seasonally.

Similarly, from the table considering savings done on monthly basis relative to savings done seasonally, respondents’ total income is positive and significant. Moreover, from the results, it can be seen that if a respondent was to increase his/her total income by one unit, the odds ratio (Exp(B)) of preferring to save monthly to saving seasonally would be expected to increase by a factor of 1.000 (with no difference between the magnitude of the effects). This also means that one can confidently accept the hypothesis that the odds ratio in question is 1 (the value expected if there was no effect). Thus, suggesting that if a respondent was to increase his/her total income by one unit, it is expected that he/she would be more likely to be indifferent to save monthly or seasonally.

Arguably, the above findings are quite significant as they take the decision to save beyond the two-stage sequential process suggested by Rodriguez and Meyer (1988) to include a third stage. The first stage of the decision according to Rodriguez and Meyer (1988) involves a decision on the proportion of the income to be consumed after which a residual is left called saving; and a second stage, a decision on the allocation of the surplus funds, that is, the savings among alternative forms of saving, namely, formal or informal. The envisaged third stage of the savings decision takes into account the time horizon of savings or time preference to save which has been widely considered as an important determinant of wealth accumulation (Laajaj, 2015). The underlying premise of the time horizon for savings options is a reflection of two important scenarios. On one hand is the need for liquidity at short notice by rural households since such households are relatively more risk-averse and may have to fall on cash-in-hand to handle emergencies (National Council of Applied Economic Research, 2011). That is, such households exhibit present-biased preferences, meaning they may be tempted to spend any cash-in-hand (Dupas & Robinson, 2010; Laibson, 1997). This implies that such households may find it difficult to resist the urge to spend money if the money is readily available in the house but less so if the money is saved far away from them such as in a bank (Dupas & Robinson, 2010).

Since by the nature of their dominant income generating activities, the respondents’ income streams are subject to seasonal as well as price variations, they may have to shorten the time-frame of holding onto surplus funds after consumption by quickly saving such funds as a way of overcoming present-biased preferences in order to avoid any future poverty trap. Hence, the anticipation of a gloomy future may induce the rural householder to shorten his/her savings time horizon, that is, reduce the time-frame he/she may hold onto surplus funds in order to reduce the distress caused by the anticipation of poverty. In essence, a shorter time-frame of holding onto surplus funds causes the individual or household to be more future biased, that is, quick saving of surplus funds to provide a safety net for the future (Kempson & Finney, 2009).

On the other hand, the respondents may have to lengthen the savings time horizon by holding onto to surplus funds in the event of worsening future consumption especially with the dwindling fortunes of the tomato industry in Ghana. In other words, the choice of the longer time-frame is motivated by the wish to hold on to their financial resources so as to retain some capacity for their present consumption and other needs before thinking of saving (Zeller, 2000). Hence, this may explain the indifference in saving weekly, monthly or seasonally among the respondents.

3.7.3. Years of education

In terms of saving on weekly basis relative to saving seasonally, years of education is positive and significant. Similarly, years of education is significant and positive for saving monthly relative to saving seasonally.

From Table 8, if a respondent is to increase his/her number of years of education by one point, the odds ratio (Exp(B)) of such a respondent preferring to save weekly relative to saving seasonally would be expected to increase by a factor of 1.066 (6.6%). Similarly, if a respondent is to increase his/her number of years of education by one point, the odds ratio (Exp(B)) of such a respondent preferring to save monthly relative to saving seasonally would be expected to increase by a factor of 1.123 (12.3%). These suggest that if a respondent was to increase his/her education by one unit, it is expected that s/he would be more likely to prefer to save either weekly or monthly over saving seasonally.

Consistent with the fact that education is considered as one of the most important determinants of income (Fields, 1980), it has also been found to have links with the development of an individual’s skills and attitudes relating to money management matters (Niculescu-Aron, 2012) particularly on taking decisions relating to saving. In other words, education particularly financial education has a significant impact on decisions regarding saving methods and savings instruments (Niculescu-Aron, 2012) of which the choice of the period for one’s savings is crucial. Therefore, education is seen as a catalyst to create a desirable behaviour of saving “now” (weekly or monthly) rather than procrastinating savings to the “future” that is, saving seasonally.

3.7.4. Average amount saved per period

Results in Table 8 indicate that for savings done on weekly basis relative to seasonal savings, the average amount saved is significant but negative—this is the multinomial logit estimate for a unit increase in amount saved on weekly basis relative to amount saved seasonally given other variables in the model are held constant. Thus, if a respondent was to increase his/her amount saved by one unit (cedi), the multinomial log-odds of choosing to save weekly as compared to saving seasonally would be expected to decrease by .995 (99.5%) while holding all other variables in the model constant.

The Wald test statistic for the predictor amount saved is 26.859 with an associated p-value of 0.000. If the alpha level is set to 0.05, then the null hypothesis is accepted and therefore it can be concluded that for savings on weekly basis relative to saving seasonally, the regression coefficient for amount saved has be to statistically different from zero given the other predictors in the model. If a respondent was to increase his/her amount saved by one unit, the odds ratio (Exp(B)) of preferring to save weekly to saving seasonally would be expected to decrease by a factor of 0.995 (99.5%). More generally, this suggests that if a respondent was to increase his/her amount saved by one unit, it is expected that s/he would be more likely to prefer to save seasonally over saving weekly.

Similarly, from Table 8 considering savings done on monthly basis relative to savings done seasonally, the amount saved per period is negative but significant. Moreover, from the table, it can be seen that if a respondent was to increase his/her amount saved per period by one unit, the odds ratio (Exp(B)) of preferring to save monthly to saving seasonally would be expected to decrease by a factor of 0.997 (99.7%). More generally, this suggests that if a respondent was to increase his/her amount saved by one unit, it is expected that he/she would be more likely to prefer to save seasonally over saving monthly.

The decision to save on seasonal basis relative to the other periods of savings among the respondents is underlined by the fact that the nature of their dominant economic activity which is tomato cultivation makes them earn their income on seasonal basis. The income earned by these is basically for consumption smoothing purposes throughout the year, for wealth accumulation, and for contingency purposes in case of bad harvest. Therefore, if these farmers are able to save at all, the regularity of such savings tends to suffer. This results in a situation where the frequency of their savings is more likely to be tied to the period when they earn their income from tomato cultivation unless they are engaged in other income generating activities. Thus, the choice of saving seasonally over the other periods for saving is in line with the longer time-frame of households holding onto their surplus funds in order to retain some capacity for their present consumption and other needs before thinking of saving (Zeller, 2000).

3.7.5. Engagement in non-farm income generating activities

This is a dummy variable, and it is the odds ratio of comparing engagement in non-farm income generating activities to non-engagement in non-farm income generating activities for saving weekly, monthly or seasonally given that the other variables in the model are held constant. From Table 8, in terms of saving on weekly basis relative to saving seasonally, engagement in non-farm income generating activities is positive and significant. Similarly, the variable is significant and positive for saving monthly relative to saving seasonally. This suggests that if a respondent was to engage in non-farm income generating activities in addition to his/her tomato production, it is expected that s/he would be more likely to prefer to save weekly and monthly over saving seasonally.

This suggests that by engaging in non-farm income generating activities in addition to their agricultural production, farm households make positive gains in income per capita (Seng, 2015) enabling them to save more on regular basis than when income generation is restricted to agricultural production. This is because in addition to income from agricultural production, engagement in non-farm activities creates more market openings for farm households to increase their level of income which has the potential of enabling them to stabilize household income thereby allowing them to save on more regular basis.

4. Conclusion

It has been recognized that rural households place great value on savings services which are tailored to match the frequency of their income generating activities. It is against this background that the study sought to determine the frequency of savings among tomato farmers in three regions of Ghana. The results of the study indicated that education can play a major role in the frequency of rural savings in that it can be applied to create a desirable behaviour of saving “now” among rural households rather than procrastinating savings to the “future” which is full of uncertainties. Moreover, the study indicates that engagement in non-farm activities can serve as a platform to create more market openings for farm households to increase their income levels which have the potential of enabling them to stabilize their household income thereby allowing them to save on more regular basis.

From the foregoing, the following policy recommendations are being suggested:

• Efforts should be made by the government for more funds allocation not only for formal education, primary, secondary and tertiary levels, but also for adult education programmes with the hope that such programmes will increase literacy levels among the adult population. Moreover, education of farmers by particularly extension agents should aim at expanding the horizon of farmers on the importance of financial savings. These considerations, it is envisaged, would ensure that adults particularly those in the rural areas take informed decisions on savings.

• Programmes should be drawn by stakeholders in rural development to help rural households diversify their income sources by engaging in non-farm income generating activities in addition to their on-farm activities. This can help to create more avenues to increase income levels to enable them save on more regular basis and help them to reduce their vulnerability to risk factors such as erratic rainfall frequency.

• Financial intermediation efforts by stakeholders in rural finance should be tailored to match the income generating activities of rural dwellers. This, it is hoped, will go a long way to cut down operational costs in extending financial services to rural households.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Daniel Aidoo-Mensah

Daniel Aidoo-Mensah is an experienced agricultural economist with agricultural extension background, and microfinance specialist with universal banking experience. He graduated Excellent cum Laude at the master’s level in Microfinance (Development Economics) from Università Degli Studi Di Bergamo in Italy. He stayed and did voluntary work in Sri Lanka as part of his dissertation work when schooling in Italy. He worked at the Methodist University Ghana, where he handled courses in agricultural economics and agribusiness from 2009 to 2021. Currently, he is a lecturer in the Department of Agribusiness Management and Consumer Studies School of Agriculture and Technology University of Energy and Natural Resources Dormaa-Campus. He holds a PhD in Agricultural Economics from the Kwame Nkrumah University of Science and Technology. He spent 12 months as a Visiting Scholar at Makerere University, Uganda as part of his PhD programme where he also had certificates in project planning and management and project monitoring and evaluation.