A logistic Regression Analysis of Poverty Status among Cassava Processors and Marketers in Benue State, Nigeria. Joseph Fefa
Yüklə 168,42 Kb.
|
- Bu səhifədəki naviqasiya:
- 3.1.5 Method of Data Analysis
- 3.1.6 Model Specification
- A Priori Expectation
- 4.1 Results and Discussions 4.1.1 Assessment of Income Generation from Cassava Processing and Marketing
- 4.1.2 Determination of Poverty Status of the Sampled Respondents
- Table 2: Distribution of sampled respondents by their poverty indices before and after joining cassava processing and marketing
- (i) Total Average Annual Income
- (iii) Poverty Gap Index (FGT 1 )
- (iv) Severity of Poverty (FGT 2 )
- Table 3: Results of the Estimation of the Logistic Regression Model
3.1.4 Method of Data Collection The data required for this study were basically primary and were collected through an open-ended and structured questionnaire, oral interview, personal observations and Focused Group Discussions (FGDs). These instruments helped in obtaining information for the study. 3.1.5 Method of Data Analysis Data were analyzed using descriptive statistics, budgetary and logit regression analyses. Descriptive statistics, including frequency counts, tables, charts, percentages and means were used to analyze the socio-economic characteristics of the respondents. A Multivariate Logit regression model was used to test the hypothesis stated, using maximum likelihood estimation procedure; the Hosmer-Lemeshow test was used to test goodness-of-fit of the model and the correlation matrix was used to test for multicollinearity, correlation between pairs of variables and how the synergies affected the dichotomous variable- poverty status. Also, the Headcount Index and Poverty Gap Index were used to measure the poverty status of the respondents. 3.1.6 Model Specification Oluwashola (2010), Apata, Apata, Igbalajobi and Awoniyi (2010), Yusuf, Adesanoye and Awotide (2008), Adeyemo, Oke and Akinola (2010), Chaudhry (2009) and Fefa (2012) provided a more flexible framework for analyzing cassava processing and marketing and poverty reduction in Benue State. This research adopted methods used by these researchers to analyze the relationship that exists between cassava processing and marketing and poverty reduction in Benue State. The parameters were estimated by maximum likelihood, with the likelihood function formed by assuming independence over the observations. Thus given
P(Y) measures poverty status, where, Y might be poor (1) or non-poor (0). By taking the natural logs of (1) and simplifying, the log likelihood transformed the structural equation to:  … (2)
Where:
LnYi = Natural log of Y(poverty status) Xki = A set of household socio-economic characteristics k = Parameters μi = Random disturbance term. 0 = Intercept From Equation (2), the model for this study was implicitly specified as: PTY= f(x1, x2, x3, x4, x5, x6, x7, x8, x9, x10) … (3) Where, PTY = dependent variable (poverty status), calculated as: 
If it was less than 1.5 US dollar, it meant the household was poor and it was assigned (1). If it was up to 1.5 US dollar, the household was non-poor and it was assigned (0). X1 = Annual income from cassava processing and marketing X2 = Quantity of cassava processed and marketed in bags (100kg). X3 = Number of square meals taken per day (1 if three square meals a day, 0 if otherwise). X4 = House type (1 if zinc roof and cemented completely, 0 if otherwise) X5 = Access to ‘improved’ medical services (1 if respondent visited dispensary, specialist and general hospitals, 0 if otherwise) X6 = Access to clothing (1 if at least 1 new cloth is purchased in a year as a result of cassava processing and marketing, 0 if otherwise). X7 = Level of education of the respondents (1 if the respondent attained secondary education and above, 0 if otherwise). X8 = House size X9 = Distance to the markets (local)(Kilometres) X10 = Ownership of processing machine Thus, the explicit form of the model became: PTY = β0+ β1X1+β2X2+β3X3+β4X4+β5X5+β6X6+β7X7+β8X8+β9X9+β10X10+ µi (4) β0 = Intercept of the model β1-β10 = Parameters µi = A random disturbance term.
In this study, 1, 2, 3, 4, 5, 6, 7, 10 were expected be negatively signed, while 8, 9, were expected to be positively signed. 3.1.7 Poverty Indices Poverty status was measured using the headcount ratio and poverty gap measures. Headcount ratio was expressed as:
Where: H = headcount ratio with values ranging from 0 to 1. The Poverty Gap was measured using the Foster-Greer-Thorbecke (FGT) metric, which is a generalized measure of poverty within an economy. It combines information on the extent of poverty (as measured by the headcount ratio), the intensity of poverty (as measured by the Total Poverty Gap) and inequality among the poor (as measured by the Gini and coefficient of variation for the poor). The FGT measure was developed by Professors Eric Thorbecke, Joel Greer and James Foster in 1984. (See Foster, Greer and Thorbecke, 1984) The formula for the FGT was given by:  ... (6)
Where Z = an agreed upon poverty line ($1.5 in this case) N = number of people in an economy H = the number of poor (those with incomes at or below Z) Yi = individual incomes α = “sensitivity” parameter (FGT Index and takes on values of 0, 1, and 2) The FGT measure corresponds to other measures of poverty for particular values of α. For α = 0, the formula reduces to:
which is Headcount ratio, or fraction of the population which lives below the poverty line. If α = 1 then the formula is:  ... (8)
Which is the average poverty gap (APG) or amount of income necessary to bring everyone in poverty right up to the poverty line, divided by total population. This refers to amount an average person would have to contribute in order for poverty to be just eliminated. A good deal of technical literature on poverty uses α = 2 version of the metric.
In this form, the index combines information on both poverty and income inequality among the poor i.e. the severity of poverty. Specifically in this instance the FGT can be rewritten as:  ... (10)
Where Cv = coefficient of variation among those with incomes less than Z, H is the total of the poor as above, and µ is given by  ... (11)
The α = 2 version is a standard used by World Bank and other international agencies for measuring poverty. The Gini-coefficient can be calculated using the formula below:
Where
μ = mean income of the population Pi = income rank of P of individual i, with income X, such that the richest person receives a rank of 1 and the poorest a rank of N. 4.1 Results and Discussions 4.1.1 Assessment of Income Generation from Cassava Processing and Marketing Data on respondents by income generated before and after joining cassava processing and marketing are presented in Table 1. Table 1: Distribution of sampled respondents by average annual incomes before and during cassava processing and marketing activities
Source: Fefa, 2012. Table 1 shows that 61.3% of the respondents earned an average annual income of less than Generally, cassava processing and marketing has increased the proportion of respondents earning up to A poverty line of To determine by how much cassava processing and marketing have actually increased the income of the sampled respondents, the ratio of the aggregate income of the respondents before they joined cassava processing and marketing to their aggregate income when they joined cassava processing and marketing was computed. Data obtained indicate that aggregate annual income before cassava processing and marketing was The ratio (R) = R = = 2 This ratio indicates that getting involved in cassava processing and marketing has doubled the respondents’ income. This increase in income undoubtedly has improved the quality of life of the respondents and hence has reduced poverty. This finding of 100% increase in income is consistent with Akighir (2011), who reported that aggregate income of respondents increased by 104% when they were involved in rice processing and marketing. 4.1.2 Determination of Poverty Status of the Sampled Respondents In order to determine the poverty status of sampled respondents, the poverty line of US$1.5 was used to estimate the respondents’ status before and when they were involved in cassava processing and marketing. These estimates were further used to classify the respondents into a category of either being poor or non-poor. These criteria were used alongside the Foster-Greer-Thorbecke (FGT) index and the different dimensions of poverty and incidence, FGT0, FGT1, FGT2 and Gini coefficient were calculated. The results obtained are presented in Table 2. Table 2: Distribution of sampled respondents by their poverty indices before and after joining cassava processing and marketing
Source: Authors’ Computation Table 2 shows data on poverty lines of the respondents before and after joining cassava processing and marketing. The table further shows that a respondent with an average annual income greater or equal to Similarly, Table 2 shows that after joining cassava processing and marketing, the respondents’ upper poverty line was It can be observed from Table 1 that a respondent who is considered to be moderately poor joining cassava processing and marketing, that is, with an average annual income of below Table 2 shows the Foster-Greer-Thorbecke (FGT) indices of the incidence of poverty – FGT0, FGT1, and FGT2 based on the classification of the respondents as non-poor, moderately poor and core poor before and after joining cassava processing and marketing. The table shows that the proportion of core poor fell from 38% before respondents joined cassava processing and marketing to 21% after engaging in the business. The proportion of moderately poor remained fairly stable at 23% and 24% respectively. However, the proportion of non-poor rose significantly from 39% before joining cassava processing and marketing to 55% after taking to the business. A further confirmation that cassava processing and marketing have improved the quality of life of the respondents is provided by the severity of poverty index (FGT2). Although this index does not indicate a serious severity of poverty among respondents before joining cassava processing and marketing, it was further reduced (from 0.203 to 0.137) when they joined cassava processing and marketing. The Gini Coefficient also shows that before cassava processing and marketing an income inequality of 0.25 existed among the respondents. This Gini Coefficient from economic theory is tolerated and could be considered that income was almost equitably distributed among respondents. But equity in income distribution improved when the respondents joined cassava processing and marketing as the ratio fell to 0.09. This is in line with Ali and Thorbecke (2000), who reported that reducing inequality has a larger positive impact on poverty than does growth; Akighir (2011), who reported that rice processing and marketing activity have reduced poverty, augmented the quality of life and reduced inequality in the divergence among respondents’ income in Kwande Local Government Area. Table 3: Results of the Estimation of the Logistic Regression Model
Source: Authors’ Computations from SSPS 17.0 Nagelkerke R- Square=0.580 Chi-Squared=520.516 -2LL=0.001 *** Significant at 1%, ** Significant at 5%, and * Significant at 10%. Table 3 indicates that the coefficient of AVINC variable (i.e. average annual income of respondents from cassava processing and marketing) is negative (-0.007), correctly signed and is statistically significant at the 5% level of significance. This implies that average annual income has influence on the probability of a respondent being non-poor. The Exp(B) or odds ratio of 0.583 indicates that a unit increase in average annual income of the sampled respondents from cassava processing and marketing would reduce their likelihood of being poor by 58.3%. Table 3 further shows that the coefficient of the QTYPROC (i.e. quantity of processed cassava) is also negative (-0.018), correctly signed but not statistically significant at 5% level of significance. This implies that the quantity of cassava processed only shows a weak influence on the probability of a respondent being non-poor. The parameter estimate for the number of times a cassava processing and marketing household feeds in a day (NOSQM) with income generated from the enterprise is negative (-0.455), correctly signed and statistically significant at the 5% level of significance. This implies that the number of times a household feeds in a day has an influence on the odds of a respondent being non-poor. The Exp(B) is 0.670. The coefficient of HOUTYP (i.e. the type of house) a sampled respondent sleeps in financed from income generated from cassava processing and marketing, is negative (-0.907), correctly signed and statistically significant at the 5% level of significance. This implies that the type of household a respondent sleeps in has influence on the probability of a sampled respondent being non-poor. The parameter estimate of ACCHLTH (i.e. access to ‘improved’ health facilities) of the respondent is positively – incorrectly signed (13.369), but it is statistically significant at 10% level of significance. This implies that access to ‘improved’ health by a sampled respondent would tend to increase his poverty status. This may be due to the fact that ‘improved’ health facilities in Benue State, the study area, are in short supply and very expensive, and hence access to them would rather impoverish those patronizing them. The coefficient of ACCLTH (i.e. access to clothing) of a sampled respondent is negative – correctly signed (-0.518) but it is not statistically significant. This implies that even though the parameter estimate agrees with economic theory, the variable is not significant in explaining the poverty status of sampled respondents in the study area. The level of education (LEDU) of a sampled respondent has a negative (-0.782) relationship with poverty status, and is statistically significant at the 5% level of significance. This implies that a respondent’s level of education influences on the probability of him or her being non-poor. The Exp(B) is 0.345. The coefficient of household size (HOUSIZE) of a sampled respondent has a positive – incorrect sign (0.925) and it is not statistically significant. This may be due to the fact that in the study area, a large number of household sizes belong to the group of dependants. A high level of dependency is more likely to throw a sampled respondent into poverty than otherwise. The Exp(B) of 0.509 indicates that a sampled respondent is 49.1% (i.e. 100-50.9)* probable to be poor. *Exp (B) values assume that all estimates meet their a priori expectations. The coefficient of distance to a local market (DIST) is positively – correctly signed and statistically significant at 5% level of significance. This implies that DIST has influence on the probability of a sampled respondent being poor. The Exp (B) is 0.630. Lastly, the coefficient of ownership of processing machine (OPMACH) is negatively-correctly signed and statistically not significant at the 5% level of significance. This implies that ownership of processing machine has influence on the probability of a sampled respondent being non-poor. The Exp (B) of 0.330 indicates that the odds of a sampled respondent being non-poor are explained 33% by his personal machine. Thus, the Chi-Squared value of 520.516 which is significant beyond 0.001 per cent shows that the model has performed well. The Nagelkerke R-Square of 0.580 shows that the explanatory variables influence 58% of the log likelihood of cassava processors and marketers being non-poor i.e. cassava processing and marketing activities of the respondents tend to influence their poverty status by 58%. Yüklə 168,42 Kb. Dostları ilə paylaş:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
… (1)
… (5)
... (7)
... (9)
…. (12)
