3.1 Determinants of financial inclusion among rice farmers
Descriptive analysis shows that 65.7 percent of the rice farmers were financially included against 34.3 percent who were financially excluded, mainly driven by mobile money technology. Women accounted for 38.4 percent of all financially included rice farmers in our sample.
Among the financially included rice farmers, 50.2 percent affirmed to own bank account in a microfinance or commercial bank, whereas 91.8 percent owned a mobile money account, suggesting an association between owning a bank account and owning a mobile money account.
Results of the probit model are presented in Table II. The Wald test assessed the validity, and the results indicated that the model fitted the data very well, with a Wald Chi-squared of 58.63, P = 0.000. Thus, the hypothesis that all the coefficients of the independent variables are jointly equal to zero was rejected. Our results reveal that out of the seven hypothesised covariates, only five variables significantly influence the financial inclusion level. The variables were gender, age, marital status, household headship, respondent’s education level, the education level of the respondent’s partner, and household size.
Table II: Probit model results for the determinants of financial inclusion among rice farmers
Financial inclusion Status
|
Coeff.
|
Std. Err.
|
z
|
P>|z|
|
Marg. Eff
|
Std. Err.
|
z
|
P>|z|
|
Respondent sex
|
-1.15923
|
0.354692
|
-3.27
|
0.001***
|
-0.39329
|
0.11252
|
-3.5
|
0.000***
|
Age
|
-0.01203
|
0.006116
|
-1.97
|
0.049**
|
-0.00416
|
0.00212
|
-1.97
|
0.049**
|
Marital status
|
0.029478
|
0.100808
|
0.29
|
0.77
|
0.010199
|
0.03487
|
0.29
|
0.77
|
Household headship status
|
0.442624
|
0.349827
|
1.27
|
0.206
|
0.149857
|
0.11529
|
1.3
|
0.194
|
Respondent education level
|
0.226074
|
0.09452
|
2.39
|
0.017**
|
0.078216
|
0.03256
|
2.4
|
0.016**
|
Partner education level
|
0.221035
|
0.096061
|
2.3
|
0.021**
|
0.076472
|
0.03313
|
2.31
|
0.021**
|
Household size
|
0.075943
|
0.031874
|
2.38
|
0.017**
|
0.026274
|
0.01101
|
2.39
|
0.017**
|
Where ***, **,* denote significance level at 1%, 5% and 10% respectively.
Y = PR (Financial Inclusion Status) (predict) = 0.70324881 and area under the ROC curve = 0.738
The log-likelihood estimates of the regression model indicated that age (P > 0.049), sex (P > 0.001), education level of the respondent (P > 0.017), education level of the respondent’s partner (P > 0.021) and household size (P > 0.017) are key factors driving financial inclusion among Gihanga district rice farmers. The age and sex of rice farmers are significantly and negatively associated with financial inclusion status. As the rice farmer in our sample is old, it is less likely for the farmer to use formal financial services. The dominance of digital finance can explain the negative relationship that drives financial inclusion in our sample. It is a new technology, whose adoption declines with age, as acknowledged by Wilson (2021).
The results on the influence of age contrast with those reported by Pena et al.(2014) and Amendola et al.(2016), but converge with the findings of Wardhono et al. (2016).
Two earlier studies found that age had a positive and significant influence on financial inclusion in Mexico and Mauritania, while the results of this study and those of Wardhono et al. (2016) show a negative correlation between age and financial inclusion. These findings might be due to differences in the targeted population when assessing the determinants of financial inclusion. In this study, we focused on uneducated rice farmers in rural areas and derived their livelihood from farming. This might reflect problems with technology adoption, given that the major part of financial inclusion is driven by mobile money technology.
On the other hand, our results on the sex variable contrast with Wardhono et al. (2016), who reported a better position in women’s financial inclusion status, while in our research, being women negatively influences financial inclusion status.
This might be due to limitations in asset ownership such as phones or any collateral required for women's loan-seeking purposes. Our results converge with those of Lotto (2018), who finds that being a woman in Tanzania reduces the likelihood of being financially included. Women’s exclusion from the financial system is not a new phenomenon and may be explained by various reasons. According to Demirguc-Kunt et al. (2018), cited by Lotto (2018), women are unable to pledge collaterals. This is associated with women’s lack of asset ownership in various societies, insufficient financial education, and inadequate business experience.
Our findings on the education of both the respondents and their partners revealed a positive correlation with the financial inclusion status of rice farmers in the district. This is similar to the findings of Pena et al. (2014) and Kouadio and Gakpa (2020), where education is one of the most important determinants of financial inclusion with respect to individual characteristics. Household size has a positive influence on the likelihood of a rice farmer being financially included. Our results support the findings of Gitaharie et al. (2017) and Ackah and Acquah (2012) in Indonesia and Ghana, respectively in which household size is positively associated with financial inclusion. This might imply that most of the time, a large family is subjected to the diversification of livelihood activities, increasing the probability of having an activity requiring mobile money or formal financial services.
The literature also suggests that larger households may have more income recipients, although this is not a rule given that wealthy households revealed minor differences compared to poor households in terms of their size(Rutstein and Kiersten, 2004).
3. 2 Effect of Financial Inclusion on Rice Farmers Wealth Assets and Food Security
A descriptive analysis of the wealth assets index showed interesting patterns in the repartitioning of rice farmers into different quintiles. It reveals an almost equal distribution of 20 and 19.5 percent of financially included and excluded rice farmers, respectively in the first quintile, while it is almost the same trend for the second quintile with 21 and 19.6 percent for financially excluded and included rice farmers, respectively. However, an interesting difference is observed in the fourth and fifth quintiles, with a large percentage skewed toward financially included rice farmers.
Regarding the food consumption score, descriptive analysis revealed that almost 84 percent of financially included rice farmers fell in the acceptable tercile against 81.2 percent of financially excluded.
Such differences are not impressive, but a noticeable difference is observed when we examine the proportion of the two groups based on the first tercile of poor food diversity. A high proportion of financially excluded (6.8 percent) falls into the poor tercile of food groups compared to financially included (3.9 percent) people.
3.2.1. Prediction of rice farmers’ probability associated with their financial inclusion Status
After predicting the probabilities associated with rice farmers' financial inclusion status (Table III), we explored the patterns of the propensity scores and tested the main validity requirements for PSM implementation.
Table III: Prediction of rice farmers’ financial inclusion status probabilities
Variable name
|
Coef.
|
Std. Err.
|
z
|
P>|z|
|
[95% of Confid. Interval]
|
Respondent sex
|
-1.1592
|
0.35469
|
-3.27
|
0.001***
|
-1.85441
|
-0.46404
|
Respondent age
|
-0.012
|
0.00612
|
-1.97
|
0.049**
|
-0.02401
|
-4.1E-05
|
Marital status
|
0.02948
|
0.10081
|
0.29
|
0.77
|
-0.1681
|
0.227058
|
H.H. Headship status
|
0.44262
|
0.34983
|
1.27
|
0.206
|
-0.24303
|
1.128272
|
Respondent education level
|
0.22607
|
0.09452
|
2.39
|
0.017**
|
0.040818
|
0.411329
|
Partner’s education
|
0.22104
|
0.09606
|
2.3
|
0.021**
|
0.032759
|
0.409312
|
level
|
|
|
|
|
|
|
Household size
|
0.07594
|
0.03187
|
2.38
|
0.017**
|
0.013471
|
0.138416
|
The overall mean propensity score was 0.67954 with a standard deviation of 0.17914. The mean propensity score for the treatment group was 0.7239665, with a standard deviation of 0.1675508, while the mean and standard deviation were respectively 0.5896703 and 0.1683216 for the control group. Figure 4 shows the propensity distributions of the treatment and control groups.
Following the range of propensity scores accounting for common support between treatment and control, visual inspection demonstrated strong common support based on the covariates between the two groups. The graphical view shows a symmetrical distribution between the treatment and control groups, with a high proportion of people under common support, whereas the off-support proportion is insignificant. The region of common support was defined with a propensity range of 0.1883995–0. 929398. The following sections focus on the impact of financial inclusion on rice farmers' wealth assets and food security in Gihanga district.
3.2.2 Impact of financial inclusion on rice farmers’ wealth assets rice farmers in Gihanga distrit.
Using the Kernel matching method, results for wealth assets and food consumption scores are reported in Table IV and V respectively. A two-tailed t-test with 95% level of confidence and seven degrees of freedom was retained. The corresponding t-statistic was 2.365. Table IV presents the average treatment effect of financial inclusion on the wealth assets of Gihanga District rice farmers.
Table IV: Effects of financial inclusion on rice farmers’ household wealth in Gihanga district
Variable
|
Sample
|
Treated
|
Controls
|
Difference
|
S.E.
|
T-Stat
|
Wealth Assets
|
Unmatched
|
-0.0507477
|
0.01649425
|
-0.067242
|
0.10994024
|
0.61
|
|
ATT
|
-0.0507477
|
0.06288741
|
-0.1136351
|
0.12673157
|
0.9
|
Based on the findings in Table IV, the difference in ATT between the treated and control groups was negative, although not significant (|0.9 |< 2.365). This result implies that there is no difference between financially included and excluded rice farmers in the Gihanga district in terms of wealth assets based on the number and quality of their possessions. Such results would be a temptation to support the argument that financial inclusion driven by credit makes some people poorer and does not improve their living conditions (Stewart et al., 2010). However, reality is more complex and requires cautious conclusions. These results enrich the contrast in the literature on the impact of financial inclusion interventions, which are small and variable. According to Duvendack and Mader (2019), on average, financial services may not have a meaningful net positive effect on poor or low-income users, although some services have positive effects on some people. Gopalaswamy et al. (2016) found positive but small effects on income, female empowerment, asset creation, and consumption expenditure.
Our findings fall into this mixed background on the impact of financial inclusion. Financial inclusion status was assessed in a community of rice farmers, mainly relying on farming as a major part of their livelihoods. While predicting the likelihood of financial inclusion, sex and age negatively influenced the likelihood of financial inclusion. This is because financial services are mainly delivered via digital finance technology using mobile money accounts, whereas old rice farmers are not as technology as young rice farmers.
The wealth asset index employed to measure rice farmers' poverty reduction better reflects long-term welfare (Rutstein and Johnson, 2004) resulting in the accumulation of possessions over the years. Therefore, it is logical that old adult rice farmers are better off in terms of wealth assets because they have had a sufficient time to accumulate assets and improve their living conditions.
Age has been in fact a subject of interest to various researchers in the fields of welfare and household economics. In the United States of America, Vandernbroucke and Zhu (2017) found that 25% of the population is under 35years and holds only 5 percent of total wealth, while 16% of the population is between 54 and 64years and holds 34% of the total wealth.
The same trend was observed by Taylor et al. (2011), where households headed by adults aged 65 years and above were better off in terms of net worth compared to households headed by adults aged 35 years and below. In this regard, it is logical that the financial inclusion mainly observed in younger generations can not lead to substantial changes in wealth accumulation based on asset ownership in the context of this study.
3.2.3 Impact of financial inclusion on food security among rice farmers in Gihanga distrit
Using the kernel matching method, the findings in table V show the positive influence of financial inclusion on food security proxied by food consumption among Gihanga district rice farmers.
Table V: Impacts of financial inclusion on Food Consumption among rice farmers
Variable
|
Sample
|
Treated
|
Controls
|
Difference
|
S.E.
|
T-test
|
FCS
|
Unmatched
|
71.9428571
|
56.6982143
|
15.2446429
|
2.30694174
|
6.61
|
|
ATT
|
71.9428571
|
57.2809524
|
14.6619048
|
3.81342875
|
3.84
|
Based on the findings in Table V, the difference between the treated and control groups is significantly positive (T-computed 3.84 > t-theoretic 2.65). That is, 14.6619048 weighted food consumption scores, implying that rice farmers financially included are better off in terms of food security. These findings align in part with those of Stewart et al. (2010) in terms of the impact of financial inclusion on food security.
Although he found that microcredit has mixed impacts and micro-savings have no impact on income, all related services positively impacted health, food security, and nutrition. The savings promotion effect was relatively small but positively significant on intermediate outcomes, such as savings amount and enterprise propensity, but also on broader poverty measures, such as household expenditure, income, and food security (Steinert et al., 2018).
Peters et al. (2016) found that financially included individuals in South Asia appreciated the ability to stabilise their families' consumption patterns as a short-term benefit of financial services. However, these positive impacts contradict the findings of Doocy et al. (2005) in Ethiopia and Nanor Michael (2008) in Ghana based on randomised control studies. Their results showed a significant difference in household diet and food security due to micro-credit.
Given the above empirical findings, these results join the cap of the dominant narrative of financial inclusion impacts on human well-being, especially that of the poor. The positive effects observed in this study are not surprising since food is at the first level of basic needs in Maslow’s hierarchy of needs (Hopper, 2016). This might be explained by the fact that financially included people are better informed about dietary behaviour, given that the food security measure retained for the current study is based on food diversity groups.
3.2.4 Role of gender patterns of financial inclusion on wealth assets and food security
The role of gender patterns of financial inclusion on rice farmers' poverty reduction in terms of wealth assets and food consumption scores was assessed using propensity score matching and the Z test for means comparison between the two groups based on the sex criterion. A within-sex comparison was also done by the Z-test. The financially included and excluded men were compared, and the same comparison was applied to women. A 95% significance level with 6 degrees of freedom corresponding to a Z score of 1.960 was adopted.
When financially included rice farmers are compared by sex, we find that there is a significant difference in terms of the wealth assets index skewed to women, with a mean difference of 0.224248 [Pr (Z < z) = 0.0411]. These results imply that financially included women are better off than are men in our sample as presented in Table VI.
Table VI: Mean comparison of wealth assets between financially included men and women
However, there was a significant difference in food consumption scores skewed toward men, with a mean difference of 0. 5636174 [Pr (|Z| > |z|) = 0.0000] as detailed in table VII.
Table VII: Mean comparison of Food consumption score between financially included men and women.
Additionally, further analysis using PSM compared financially included and financially excluded men and women, as shown in the table VIII and IX, respectively. From the women’s group, being financially included leads to positive effects in terms of wealth assets, although it is not statistically significant with ATT = 0.28307728 (T = 1.34 < t = 1.960).
Table VIII: Impacts of financial inclusion on wealth asset among women rice farmers
Variable
|
Sample
|
Treated
|
Controls
|
Difference
|
S.E.
|
T-test
|
Wealth Quintiles
|
Unmatched
|
0.103696
|
0.005262
|
0.09843351
|
0.210067
|
0.47
|
|
ATT
|
0.119094
|
-0.16398
|
0.28307728
|
0.21177
|
1.34
|
However, a positive and significant effect was observed in terms of food consumption scores with an ATT = 11.9538462 and (T = 2.26 > t = 1.960).
Table IX: Impacts of financial inclusion on Food Consumption among women rice farmers
of Gihanga
Variable
|
Sample
|
Treated
|
Controls
|
Difference
|
S.E.
|
T-Stat
|
Food Consump Sc.
|
Unmatched
|
70.33333
|
58.70861
|
11.62472
|
4.069357
|
2.86
|
|
ATT
|
69.61538
|
57.66154
|
11.95385
|
5.289696
|
2.26
|
Owing to the limitations of the PSM technique, a comparison between financially included and financially excluded men was implemented using the Z test. The results show a negative difference, but not significant, in terms of wealth assets between financially included and excluded men, with a mean difference of -0.1757102 [Pr (|Z| > |z|) =! 0.275].
Table X: Mean comparison of wealth assets between financially excluded and included men.
Group
|
Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
|
Financially Included
|
158 − .0548755 .0795557 1 − .2108019 .1010508
|
Financially excluded
|
51 .1208346 .140028 1 − .1536152 .3952845
|
diff
|
− .1757102 .1610495 − .4913615 .1399411
|
Pr (Z < z) = 0.1376 Pr (|Z| > |z|) = 0.2753 Pr (Z > z) = 0.862 |
However, a positive and significant difference was observed in food consumption scores. The results show a mean difference of 11.18106 [Pr (|Z| > |z|) = 0.0000] between the two groups in our sample as presented in Table XI.
Table XI: Mean comparison of food consumption scores between financially excluded and
included men.
Group
|
Obs Mean Std. Err. Std. Dev. [95% Conf. Interval]
|
Financially Incl
|
158 63.75949 .0795557 1 63.60357 63.91542
|
Financially Ex.
|
51 52.57843 .140028 1 52.30398 52.85288
|
diff
|
11.18106 .1610495 10.86541 11.49671
|
Diff = mean (Financially Included) - mean (Financially excluded) Z = 69.4262 |
Ha: diff < 0 Ha: diff! = 0 Ha: diff > 0
Pr (Z < z) = 1.0000 Pr (|Z| > |z|) = 0.0000 Pr (Z > z) = 0.0000