International Journal of Social Economics
Nonfarm enterprise participation and healthcare expenditure among farm
households in rural Ghana
Samuel Ampaw, Edward Nketiah-Amponsah, Nkechi Srodah Owoo, Bernardin Senadza,
Article information:
To cite this document:
Samuel Ampaw, Edward Nketiah-Amponsah, Nkechi Srodah Owoo, Bernardin Senadza, (2019)
"Nonfarm enterprise participation and healthcare expenditure among farm households in rural
Ghana", International Journal of Social Economics, Vol. 46 Issue: 1, pp.18-30, https://doi.org/10.1108/
IJSE-06-2017-0248
Permanent link to this document:
https://doi.org/10.1108/IJSE-06-2017-0248
Downloaded on: 04 June 2019, At: 04:42 (PT)
References: this document contains references to 45 other documents.
To copy this document: permissions@emeraldinsight.com
The fulltext of this document has been downloaded 118 times since 2019*
Users who downloaded this article also downloaded:
(2018),"Gender perspective on life insurance demand in Ghana", International Journal
of Social Economics, Vol. 45 Iss 12 pp. 1631-1646 https://doi.org/10.1108/IJSE-03-2017-0120
(2019),"Dynamics of unidimensional and multidimensional inequality in Pakistan: Evidence
from regional and provincial level study", International Journal of Social Economics, Vol. 46 Iss
2 pp. 170-181 https://doi.org/10.1108/
IJSE-07-2017-0278
Access to this document was granted through an Emerald subscription provided by emerald-
srm:534301 []
For Authors
If you would like to write for this, or any other Emerald publication, then please use our Emerald
for Authors service information about how to choose which publication to write for and submission
guidelines are available for all. Please visit www.emeraldinsight.com/authors for more information.
About Emerald www.emeraldinsight.com
Emerald is a global publisher linking research and practice to the benefit of society. The company
manages a portfolio of more than 290 journals and over 2,350 books and book series volumes, as
well as providing an extensive range of online products and additional customer resources and
services.
Emerald is both COUNTER 4 and TRANSFER compliant. The organization is a partner of the
Committee on Publication Ethics (COPE) and also works with Portico and the LOCKSS initiative for
digital archive preservation.
*Related content and download information correct at time of download.
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
The current issue and full text archive of this journal is available on Emerald Insight at:
www.emeraldinsight.com/0306-8293.htm
IJSE
46,1 Nonfarm enterprise participation
and healthcare expenditure
among farm households in
18 rural Ghana
Received 16 June 2017 Samuel Ampaw, Edward Nketiah-Amponsah, Nkechi Srodah Owoo
Accepted 4 May 2018
and Bernardin Senadza
Department of Economics, University of Ghana, Accra, Ghana
Abstract
Purpose – Rural poverty remains high in many developing countries, Ghana inclusive. This has implications
for healthcare affordability and utilization, and thus the attainment of universal health coverage. Nonfarm
diversification is seen as a means by which rural farm households can increase incomes and smooth
consumption including healthcare. The purpose of this paper is to investigate the impact of nonfarm
enterprise participation on healthcare expenditure among farm households in rural Ghana.
Design/methodology/approach – Using nationally representative household data from the sixth round of
the Ghana Living Standards Survey (GLSS 6), the paper employs endogenous switching regression and
propensity score matching techniques to account for potential selectivity bias.
Findings – Results indicate that households that participate in nonfarm enterprises earn higher incomes
and expend more on healthcare. Total household income and region of residence are significant
determinants of healthcare expenditure among farm households in rural Ghana. In addition, while in
nonfarm enterprise nonparticipating households the marital status of the head of household is important,
for participating households the head having at least secondary education significantly influences
healthcare expenditure.
Practical implications – Promoting nonfarm activities and hence raising the incomes of households in
rural areas of Ghana has the potential of increasing health capital through increased investments in health.
It will also positively impact access to and utilization of healthcare and ultimately contribute towards
increased farm and non-farm productivity.
Originality/value – Previous studies have only examined the determinants of nonfarm enterprise
participation or its impact on household welfare, poverty, inequality, food security and agricultural
investments. While evidence abounds on the positive impact of rural nonfarm enterprise participation on
household income, which in turn has implications for household health expenditure, the potential positive link
between rural nonfarm enterprise participation and household healthcare expenditure remains unexamined.
Keywords Poverty, Endogenous switching regression, Propensity score matching,
Household healthcare expenditure, Nonfarm enterprise participation, Rural Ghana
Paper type Research paper
1. Introduction
The wealth status of an individual plays an important role in his or her access to and
utilization of quality healthcare (Pritchett and Summers, 1996; Murray, 2006; Fan and
Habibov, 2009). However, poverty remains pervasive in many developing countries and
threatens the ability of these countries to achieve universal health coverage (UHC). According
to O’Donnell (2007) and Peters et al. (2008), the poor in developing countries suffer most from
ill-health due to impoverishment. While poverty could lead to ill-health, ill-health perpetuates
poverty (Peters et al., 2008). The ultimate goal of healthcare systems is universal access and
financial protection. The World Health Assembly resolution 58.33 seeks to promote equal
International Journal of Social
Economics access to quality healthcare for everyone who needs it (WHO, 2010). The importance of
Vol. 46 No. 1, 2019
pp. 18-30
© Emerald Publishing Limited The authors of this article have not made their research dataset openly available. Any enquiries
0306-8293
DOI 10.1108/IJSE-06-2017-0248 regarding the dataset can be directed to the corresponding author.
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
universal access to and equity in healthcare for economic and social development is reflected Nonfarm
in several global efforts, including the millennium development goals (MDGs) and its sequel enterprise
the sustainable development goals (SDGs). In spite of these efforts, not much success has been participation
achieved (WHO, 2010). The WHO (2010) identifies several factors, including, unavailability of
resources, overdependence on out-of-pocket (OOP) payments, and inefficiency and inequity in
health resources utilization, as hindering the achievement of universal healthcare.
Healthcare financing in Ghana has gone through several phases. Akazili et al. (2014) 19
disclose that Ghana’s healthcare financing policies since the pre-independence era
have been modeled along the political ideologies of ruling governments. Thus, unlike the
British colonial government, who were capitalists and so OOP payments for
healthcare was practiced, the immediate post-independence socialist-oriented
government focused on providing free healthcare for its citizens (Akazili et al., 2014).
From the late 1960s onwards, tax-financed healthcare was provided in public health
facilities (WHO, 2016). However, this policy encountered quality and sustainability
challenges and in the 1980s, the user fee system, which is based on OOP payments was
introduced as part of the structural adjustment program (WHO, 2016). The user fee system
had adverse effects on access to healthcare (Akazili et al., 2014). Consequently,
community-based insurance schemes emerged in the 1990s to pool risk and to improve
access to healthcare. These community-based schemes gave impetus to the introduction of
the National Health Insurance Scheme (NHIS) in 2003. Poverty adversely impacts
Ghana’s NHIS enrollment as the poor are less likely than the rich to enroll on the scheme
(Akazili et al., 2014; Jehu-Appiah et al. 2011; Dixon et al., 2011). Although Ghana’s
NHIS is designed to be pro-poor, its sustainability is threatened by inadequate funds
(Alhassan et al. 2016). OOP payments have therefore re-emerged at some health facilities
(Owusu-Sekyere and Bagah, 2014).
Spatially, 78 percent of the poor in Ghana live in rural areas (Ghana Statistical Service,
2014), where healthcare facilities are relatively scarce and often provide limited services.
Therefore, in addition to facing higher user fees, rural households are likely to encounter
higher transportation costs than urban households while seeking and utilizing healthcare.
Even though urban households on average expend more on healthcare in absolute terms
compared to rural households, the reverse is the case in relative terms (Ghana Statistical
Service, 2014). The high costs associated with seeking and utilizing healthcare could
discourage poor rural households from increasing investments in health and thus,
undermine existing rural poverty alleviation efforts.
In their quest to improve household welfare, rural farm households in developing
countries diversify into nonfarm enterprise activities (Reardon et al., 2006; Barrett and
Reardon, 2000). According to Haggblade et al. (2010) and Rijkers and Costa (2012), rural
nonfarm participation accounts for between 40 and 50 percent of total household income in
Africa. Empirical evidence shows that households in developing countries participating in
nonfarm enterprise activities are more food secure (Owusu et al., 2010; Jabo et al., 2014;
Dedehouanou et al., 2015; Osarfo et al., 2016), have higher agriculture investments
(Dedehouanou et al., 2015), expend more on food ( Jabo et al., 2014) and have reduced poverty
and improved welfare ( Jabo et al. 2014; Kousar and Abdulai, 2013; Senadza, 2011).
Yet, household health expenditure is found to be positively affected by household income
(Parker and Wong, 1997; Rous and Hotchkiss, 2003; You and Kobayashi, 2011; Yildirim
et al., 2011; and da Silva et al., 2015). That a potential positive link exists between rural
nonfarm enterprise participation and household health expenditure, however, has not been
inadequately explored. The novelty of our paper lies in its attempt to contribute to the
extant literature by investigating the impact of nonfarm enterprise participation on
household healthcare expenditure in rural Ghana. We apply the endogenous switching
regression (ESR) and propensity score matching (PSM) techniques to account for potential
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
IJSE selectivity bias on nationally representative data consisting of 9,318 rural households.
46,1 Our findings indicate that nonfarm enterprise participating households in Ghana earn
higher incomes and expend more on healthcare, in affirmation of our hypothesis. The rest of
the paper is organized as follows. Section 2 presents a review of related empirical literature.
The empirical strategy is discussed in Section 3. Section 4 presents and discusses the
results. Section 5 concludes with policy implications.
20
2. Related empirical literature
The determinants of nonfarm enterprise participation and its impact on several household
level variables, such as, welfare, food security and agricultural investments, have been
extensively studied in recent times. On the determinants of nonfarm enterprise
participation, Nagler and Naudé (2014) emphasize household characteristics, individual
capabilities and institutional factors as important in a household’s decision to operate a
nonfarm enterprise. Factors such as age and gender of the head of household, level of
education of household head or members, household size, share of adults in the household,
farm income, access to credit, farming experience and membership of cooperative
societies, food shortage, rainfall, production technology adopted, distance to a main road,
among others, have featured prominently in many studies (see for instance, Zahonogo,
2011; Rijkers and Costa, 2012; Nagler and Naudé, 2014; Osondu et al., 2014). Participation
in rural nonfarm enterprise activities by households in developing countries is shown to
be crucial in addressing rural poverty through assessment of its impacts on household
welfare, agriculture, food security, food expenditure, risk management, among others.
Studies that have found positive impact of nonfarm enterprise participation on household
welfare include Owusu et al. (2010), Kousar and Abdulai (2013), Jabo et al. (2014),
Dedehouanou et al. (2015) and Osarfo et al. (2016). Kousar and Abdulai (2013), for instance,
used ESR to investigate the impact of rural nonfarm employment on household welfare
among a sample of 341 households in Pakistan. They employed the PSM technique to
affirm the robustness of the results from the ESR approach. Their results showed that
participation in nonfarm enterprise activities while it reduced poverty, also increased
average household expenditure.
Examining the effect of off-farm self-employment on agricultural expenditure and food
security in Niger, based on a sample of 1,942 rural households and applying the ESR
technique, Dedehouanou et al. (2015), found that nonfarm participating households
expended more on agricultural inputs and food than nonparticipating households. Similar
findings on the impact of nonfarm participation on household food security was obtained by
Owusu et al. (2010) and Osarfo et al. (2016) for rural northern Ghana, as did Jabo et al. (2014)
for rural households in Nigeria. Korir et al. (2011) investigated the effectiveness of
investments in nonfarm enterprise activities as risk management strategies in Kenya. Using
a randomly selected sample of 100 farm households from Uasin Gishu County, they
established that, unlike farm income, income from nonfarm work stabilized total household
incomes. However, the OLS technique employed for the estimation suffers from potential
selection bias.
Few studies have explored the determinants of household healthcare expenditure in
developing countries. You and Kobayashi (2011) studied the determinants of individual
OOP expenditure in China. Controlling for selection bias with the use of the Heckman
selection model, their results indicate that as people age, they tend to spend more on
healthcare. In addition, persons suffering ill-health, earning higher incomes, living
with educated household heads, and residing in urban areas and the middle or eastern
regions of China, were found to invest more in their health. Furthermore, the study
disaggregated the effect of insurance programmes and found a positive effect of
insurance ownership on healthcare expenditure. Malik and Syed (2012) investigated the
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
determinants of OOP payments in Pakistan based on the Pakistan Standard of Living Nonfarm
Measurement (PSLM) Survey in 2004/05. The study amongst other factors, including enterprise
household characteristics, found non-food household expenditure as a positive predictor participation
of household health expenditure in Pakistan. While the literature provides evidence of the
favorable impact of participation in rural nonfarm activities on rural poverty and
household welfare, which in turn has implications for household health expenditure, the
potential positive link between rural nonfarm enterprise participation and household 21
healthcare expenditure remains unexamined. Our paper is novel, in that it seeks to
ascertain the impact of nonfarm participation on health expenditure among farm
households in rural Ghana.
3. Methodology and data
3.1 Empirical approach
Empirical literature affirm the self-selection of households into nonfarm activities (see inter
alia, Kousar and Abdulai, 2013; Jabo et al., 2014; Dedehouanou et al., 2015). As such, nonfarm
enterprise participating households may systematically differ from nonparticipating
households. Using observational dataset where there is nonrandom selection into positions
would therefore be characterized by selectivity bias. The OLS estimates of the impact of
nonfarm enterprise participation on OOP health expenditure would then be inconsistent and
inefficient. Following Mare and Winship (1987), we specify the analysis of covariance model
as follows:
X
Yi ¼ a0Z i þ ak kXkiþei; (1)
where household is denoted by i (i¼ 1,…, I ). Yi represents outcome for the ith household, αk
measures parameters to be estimated, Xki is the value on the kth measured independent
variable and εi is stochastic disturbance term. The participation status, Zi, enters the model
as an independent dummy variable for the entire pooled sample of nonfarm participants
and nonparticipants.
The decision to participate or not to participate in nonfarm enterprise is endogenous in
the household health expenditure function. Thus, the paper adopts the ESR model and PSM
technique to account for selection bias. Notwithstanding, while ESR model accounts for
selection bias due to observed and unobserved characteristics of households (Mare and
Winship, 1987), the PSM approach controls only for selection bias resulting from observed
characteristics (Rosenbaum and Rubin, 1983).
The ESR model is represented as:
X
Y 0i ¼ b0kXkiþe0i if Z i ¼ 0 ðnonparticipantsÞ; (2)k
X
Y 1i ¼ b1kXkiþe1i if Z i ¼ 1 ðparticipantsÞ; (3)k
X Y
Zni ¼ X þe ðparticipation decision functionÞ; (4)k k ki 3i
where Y0 and Y1 represent household health expenditure of nonparticipating and
participating households, respectively; Xk is a vector of explanatory variables; the β0k and
β1k are vectors of estimated parameters, Z
n
i is a latent dependent variable of participation,
which takeQs the value 1, if the household participates in nonfarm enterprise and zero
otherwise; k represent vector of estimated parameters from the participation equation.
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
IJSE The (stochastic) error terms e0, e1 and e3 are assumed to follow a trivariate normal
46,1 distribution with a zero mean and non-singular covariance matrix given by:2 2 3s0 s01 s03
Covðe e 6 70; 21; e3Þ ¼ 4s01 s1 s13 5:
s03 s13 1
22
Given the normality assumption and the normalization constraints, the endogenous
switching model can be estimated via maximum-likelihood (ML) method (Maddala, 1983;
Mare and Winship, 1987). Lokshin and Sajaia (2004), observes that the Full Information
Maximum Likelihood (FIML) estimation is an efficient way of estimating the ESR model,
whereby the probit selection equation and the health expenditure equation are
simultaneously estimated with consistent standard errors. The FIML log-linear function,
as proposed by Lokshin and Sajaia (2004) is:
XN
ln Li ¼
e
Z i ln f
1i
– ln s1þ ln ^ ðys 1iÞ
i¼1 1
þð e1–Z iÞ ln f 2i – ln ss 0þ lnð1^ ðy0iÞÞ ; (5) qffiffiffiffiffiffiffiffiffiffiffi 0
where θji equals Xigþ
r 2jeji3:=sj = 1rj , j is 0 or1, ρj represents the correlation
coefficient between the error term e3 of the participation decision function and the error terms
e0 ande1 of the health expenditure equations. Mare and Winship (1987) reveal that the
estimated correlation coefficient ( ρj) indicates the presence or otherwise of self-selection. There
is positive selection out of nonfarm activities when ρ0W0 and there is positive selection into
nonfarm activities when ρ1o0.
The propensity score, defined as the probability of participating in nonfarm enterprise
conditional on pre-participation characteristics, such as age and education, is expressed as
(Rosenbaum and Rubin, 1983):
pð XiÞ prob Z i ¼ 19Xi ¼ E Z i9Xi : (6)
Equation (4) of the ESR model (probit regression estimates) is used to compute the
propensity score values. Following Kousar and Abdulai (2013), we match participants with
non-participants of similar propensity scores, using the nearest neighbor matching (NNM)
and radius matching (RM) algorithms.
The average treatment effect on treated (ATT) is used to estimate the causal impact of
nonfarm enterprise participation on OOP health expenditure. This computes the difference
between the observed health expenditure for nonfarm participating households and its
counterfactual health expenditure had the participating households chosen not to
participate. Again, following Kousar and Abdulai (2013), the average treatment effect
on treated (ATT) of nonfarm enterprise participation on health expenditure is computed
as follows: ¼ ¼ ð Þ
ATT E E Y Y Z 1; p X ¼
E E Y Z ¼ 1; pðX Þ
1 i 0i i
i 1i i i
E Y 0i Z i ¼ 0; pðXiÞ jZ i ¼ 1 ; (7)
where Zi{0, 1} is the indicator of exposure to treatment (nonfarm enterprise participation), Xi
denotes the vector of pre-participation characteristics and Y1i and Y0i are the potential
outcomes in observed and counterfactual scenarios, respectively.
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
3.2 Data Nonfarm
The study uses the sixth round of the Ghana Living Standards Survey (GLSS 6). The data were enterprise
collected over a period of 12 months (from October 18, 2012 to October 17, 2013) by the Ghana participation
Statistical Service, with technical support from the World Bank. The GLSS is modeled along
the lines of the World Bank’s Living Standards Measurement Survey (LSMS). The data are a
nationally representative sample of 16,772 households, made up of 9,327 rural and 7,445 urban
households, in 1,200 enumeration areas. The GLSS 6 data set provides detailed information on 23
key socio-economic and demographic variables such as household characteristics, education,
migration, tourism, household agriculture income and expenditures, household nonfarm
enterprise, among others. Additionally, it provides information on living conditions and
well-being of households in Ghana. In all, 9,318 rural farm households was used for the
analysis. Table I presents description and measurement of the variables.
4. Results
4.1 Characteristics of nonfarm participating and non-participating households
Table II presents the summary statistics of the variables by participation status. The means
test indicates that except for age and the locational dummy (north), rural nonfarm
participating households differ significantly in characteristics from non-participating
households. This systematic difference affirms the existence of sample heterogeneity
suggested in the literature.
4.2 Determinants of rural nonfarm enterprise participation
The results from the participation model of the ESR and the probit regression are reported in
columns 4 and 5 of Table III, respectively. The two results are qualitatively similar.
The variables that significantly influence participation are total household income, age of
household head, age of household head squared, male household head, education (basic) of
household head, household size, ownership of livestock and farm lands, electricity consumption
and receipt of remittance. The coefficient for age is positive while that of age-squared is
negative, indicating that participation in nonfarm activities increase with the age of the
household head up to 50 years and thereafter diminishes with further increases in age[1].
This finding lends support to Abdulai and Delgado (1999), Abdulai and CroleRees (2001) but
not Dary and Kuunibe (2012) and Osondu et al. (2014). Male-headed households are less likely
to participate in nonfarm enterprise activities than female-headed households in consonance
with the findings of Dary and Kuunibe (2012) for rural Ghana and Nagler and Naudé (2014) for
Nigeria. The attainment of basic education by the household head is positively correlated with
nonfarm enterprise participation in rural Ghana. Wealthier rural households have a higher
predicted probability of participating in nonfarm enterprise activities, suggesting the
importance of “pull factors”, while operating rural nonfarm enterprise also increases with
household size. Abdulai and CroleRees (2001), Nagler and Naudé (2014) and Tran (2015) made
similar observations. Ownership of agricultural assets (livestock and farm land), consumption
of electricity and receipt of remittance are reported to have significant and positive effects on
nonfarm participation among farm households.
The Wald test of independence of both the participation and health expenditure
equations is statistically significant at the 1 percent level. This proves that the equations are
jointly dependent, thus affirming the endogeneity problem. The Wald χ2 test shows that the
correlation coefficients (ρ0 and ρ1) are jointly statistically different from zero (even though
ρ1 is insignificant). Being negatively signed then reveals that nonfarm enterprise
participating households benefit more from participation than the average population.
Conversely, since the correlation between the health expenditure equation for nonfarm
nonparticipating households and the participation equation (ρ0) is negative and significant,
households which do not participate in nonfarm activities are revealed to be worse off than
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
IJSE
46,1
24
Table I.
Description and
measurements
of variables
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
Variable Description Mean SE
Health expenditure Log of total household health expenditure (Ghana cedis) 4.001 0.017
Participate Nonfarm participation status of farm household: 1¼ participates; 0.304 0.005
0¼ otherwise
Age Age of household head (years) 47.406 0.17
Age square Age of household head square
Gender Gender of household head: male¼ 1; 0¼ female 0.761 0.004
No education Level of education of household head: 1¼ no education; 0¼ otherwise 0.549 0.005
Basic education Level of education of household head: 1¼ basic; 0¼ otherwise 0.360 0.004
Secondary education Level of education of household head: 1¼ secondary; 0¼ otherwise 0.091 0.003
Married Marital status of household head: 1¼married; 0¼ otherwise 0.642 0.005
Income Log of total household income (Ghana cedis) 8.035 0.014
North Location dummy: 1¼ residing in northern Ghana 0.377 0.005
(Northern, Upper East and West regions); 0¼ otherwise
Household size Number of household members 4.755 0.031
Livestock Ownership of livestock: 1¼ livestock owned; 0¼ otherwise 0.459 0.005
Land Ownership of agricultural land: 1¼ land owned; 0¼ otherwise 0.139 0.004
Electricity Household consumes electricity: 1¼ consumes electricity; 0.363 0.004
0¼ otherwise
Remittance Household receives remittances: 1¼ receives; 0¼ otherwise 0.331 0.005
Nonfarm
Nonparticipating subsample Participating subsample
Variable (n¼ 6,486) mean SE (n¼ 2,832) mean SE Mean difference enterprise
participation
Health expenditurea 3.945 0.021 4.112 0.031 −0.167***
Age 47.438 0.214 47.334 0.270 0.104
Gender 0.750 0.005 0.787 0.008 0.037***
No education 0.570 0.006 0.502 0.009 0.067***
Basic education 0.335 0.006 0.419 0.009 −0.084*** 25
Secondary education 0.096 0.004 0.079 0.003 0.017***
Income 7.833 0.018 8.497 0.022 −0.664***
Married 0.612 0.006 0.713 0.009 −0.101***
North 0.380 0.006 0.368 0.009 −0.012
Household size 4.361 0.036 5.658 0.057 −1.297***
Livestock 0.417 0.006 0.553 0.009 −0.136***
Land 0.126 0.004 0.169 0.007 −0.043***
Electricity 0.341 0.006 0.412 0.009 −0.071*** Table II.
Remittance 0.313 0.006 0.373 0.009 −0.061*** Summary statistics
Note: aSample size for health expenditure estimation is 5,475 of variables by
Source: Computed by authors from GLSS 6, 2012/2013 participation status
the average population. Following Mare and Winship (1987) and Asfaw et al. (2012), and
since ρ0o0 and ρ1o0, selectivity does not result from benefits derivable from the decision
to participate or otherwise in nonfarm enterprise activities and thus, selection bias is less
serious. We conclude, therefore, that selectivity is based on the systematic differences in the
characteristics of participating and nonparticipating households.
4.3 Determinants of household healthcare expenditure
Column 1 of Table III reports the OLS estimates of the determinants of household health
expenditure in rural Ghana. However, given the evidence of selection bias, the OLS estimation
is unreliable. The full maximum likelihood estimation results from the ESR model of
the determinants of health expenditure among nonfarm enterprise nonparticipating and
participating households are reported in Columns 2 and 3 of Table III, respectively.
Total household income is a significant determinant of healthcare expenditure among farm
households in rural Ghana. The magnitude of the coefficient on income for nonfarm
participating households (0.124) being greater than that for nonparticipating households
(0.078) is read as an indication that the effect of income on household health expenditure is
larger for participating households compared to nonparticipating households. By inference,
any activity which increases the total income of rural farm households will likely make them
expend more on health. This positive relationship between household income and health
expenditure has been confirmed by Parker and Wong (1997), Rous and Hotchkiss (2003), You
and Kobayashi (2011), Yildirim et al. (2011) and da Silva et al. (2015). Rural households in the
three northern regions of Ghana (upper east, upper west and northern) expend less on health
compared to their counterparts in southern Ghana, irrespective of participation status.
The household head having secondary education positively affects healthcare expenditure in
participating households only, albeit significant at 10 percent level. Malik and Syed (2012) and
Brinda et al. (2014), obtained similar results. Households with a married household head
expend more on health expenditure only in nonparticipating households.
4.4 Impact of nonfarm enterprise participation on household healthcare expenditure
The OLS results reported in Column 1 of Table III posts a positive coefficient for the
participation dummy variable, even though it is insignificant. However, the evidence of
selection bias renders the OLS estimates unreliable. The ESR and PSM techniques are
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
IJSE
46,1
26
Table III.
Estimation results
from OLS,
endogenous switching
regression, and probit
regression models
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
Explanatory OLS (1) Endogenous switching regression model (2) (3) (4) Probit model (5)
Variables Hlthexpa Hlthexp0b Hlthexp1c Participate Participate
Participate 0.0711 (0.0536)
Income 0.1097*** (0.0226) 0.0778*** (0.0246) 0.1237* (0.0667) 0.1912*** (0.0218) 0.1873*** (0.0167)
Age 0.0060 (0.0069) −0.0042 (0.0071) −0.0107 (0.0155) 0.0352*** (0.0069) 0.0405*** (0.0071)
Age2 −0.0000 (0.0001) 0.0000 (0.0001) 0.0001 (0.0002) −0.0004*** (0.0001) −0.0004*** (0.0001)
Gender −0.1123 (0.0704) −0.0524 (0.0656) −0.0755 (0.1036) −0.1670*** (0.0587) −0.1535*** (0.0525)
Married 0.1275** (0.0624) 0.1256** (0.0608) 0.0851 (0.0866) 0.0365 (0.0527) 0.0401 (0.0470)
Basic education 0.0767 (0.0499) −0.0176 (0.0555) 0.1277 (0.0917) 0.2773*** (0.0462) 0.2504*** (0.0425)
Secondary education 0.1513* (0.0838) 0.1039 (0.0810) 0.2270* (0.1284) 0.0140 (0.0761) 0.0401 (0.0700)
North −0.2877*** (0.0833) −0.2545*** (0.0819) −0.4941*** (0.1124) 0.0099 (0.0768) 0.0349 (0.0473)
Household size 0.0507*** (0.0081) 0.0568*** (0.0073)
Electricity 0.1881*** (0.0593) 0.158*** (0.0390)
Livestock 0.1638*** (0.0478) 0.257*** (0.0407)
Land 0.1118* (0.0677) 0.143*** (0.0510)
Remittance 0.1841*** (0.0504) 0.162*** (0.0392)
Constant 2.9881*** (0.2197) 3.2423*** (0.2406) 3.5672*** (1.1397) −3.3029*** (0.2396) −3.549*** (0.207)
Observations 5,475 5,475 5,475 5,475 9,318
R2 0.0329
ρ0 −0.3397*** (0.0832)
ρ1 −0.1593 (0.2513)
ESRM
Wald χ2 (8)¼ 34.62 Prob W χ2¼ 0.0000 Log pseudolikelihood¼ −12,166.8
Wald test of indep. eqns.: χ2(2)¼ 14.16 Prob W χ2¼ 0.0008
Probit regression
Wald χ2 (13)¼ 576.09 Prob W χ2¼ 0.0000 Log pseudolikelihood¼ −5,162.48 Pseudo R2¼ 0.0890
Notes: aHealth expenditure of all households. bHealth expenditure of nonfarm enterprise nonparticipating households. cHealth expenditure of nonfarm enterprise
participating households
Source: Computed by authors from GLSS 6; Standard errors in parentheses. *po0.1; **po0.05; ***po0.01
therefore employed to correct for the selection bias. Table IV presents the expected actual and Nonfarm
counterfactual health expenditures (i.e. conditional expectations) of participating households. enterprise
The average treatment effect on treated (ATT) is computed from the conditional expectations. participation
It measures the difference between the average health expenditure of nonfarm participating
households and their average counterfactual health expenditure had they not participated in
nonfarm activities. The estimated ATT suggests that nonfarm enterprise participation
significantly enables rural households in Ghana to increase expenditure on healthcare. 27
Based on the propensity scores, two matching algorithms – NNM and RM – were used
to pair treated units with control units. The two matching methods were used to check for
the robustness of the ATTs to the matching methods. The ATTs were computed using the
propensity scores technique proposed by Becker and Ichino (2002). Generated from a
probit model, propensity scores which satisfied the balancing property were employed.
The results of the ATTs are presented in Table V. The results show that the choice of
matching algorithm is relevant. This is because the two matching methods yield different
results. Corroborating the conclusions drawn based on the ESR estimate, the RM method
reveals that households which participate in nonfarm enterprise expend more on
healthcare. However, the ATTs from the PSM are lower, suggesting that the ATTs from
the PSM approach are underestimated. As sensitivity analysis Table V presents the
results of RM method with radius of 0.1 and 0.5. Increasing the radius from 0.1 to 0.5
increases the ATT.
5. Conclusion
This paper used the most recent and nationally representative data set to investigate the
impact of nonfarm enterprise participation on household health expenditure among farm
households in rural Ghana. It also shed light on the predictors of nonfarm enterprise
participation and household healthcare expenditure in rural Ghana. The results indicate that
participation in nonfarm enterprise enables rural households in Ghana to expend more on
healthcare. Total household income and region of residence are significant determinants of
healthcare expenditure among farm households in rural Ghana. In addition, whereas the
education of household head significantly affects the health expenditure of nonfarm
enterprise participating households, the health expenditure of nonparticipating households
is influenced by the household head’s marital status. Age, gender and education of
household head, total household income, the number of household members, ownership of
Nonfarm enterprise participation decision Table IV.
To participatea Not to participateb ATT Conditional
expectations and
Participating household 4.1175 3.4211 0.6964*** average treatment
Notes: aObserved; bCounterfactual. *po0.1; **po0.05; ***po0.01 effects on treated
Source: Computed by author from GLSS 6, 2012/2013 (ATTs), 2012/13
Matching algorithms Number of treated units Number of control units ATT SE t-Statistics Table V.
Nearest neighbor 3,318 1,307 0.071 0.046 1.559 ATT Estimation ofimpact of nonfarm
Radius (0.1) 1,258 1,609 0.200*** 0.050 3.970 enterprise
Radius (0.5) 1,258 1,609 0.219*** 0.049 4.516 participation
Notes: *po0.1; **po0.05; ***po0.01 on household
Source: Computed by authors from GLSS 6, 2012/2013 health expenditure
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
IJSE livestock and farm land, electricity consumption and receipt of remittances are significant
46,1 predictors of nonfarm enterprise participation among farm households in rural Ghana.
The evidence of a positive impact of nonfarm enterprise participation on healthcare
expenditure has implications for policy. Promoting nonfarm activities and hence raising the
incomes of households in rural areas of Ghana has the potential of increasing health capital
through increased investments in health. It will also positively impact access to and utilization
28 of healthcare, and enhance the achievement of UHCS. Therefore, the policy of integrated rural
development should be given greater attention as it will not only result in the provision of
services, including infrastructure and social amenities to rural areas but also aid in the
development and promotion of nonfarm livelihood activities in the rural areas of Ghana.
Note
1. From the participation equation of the endogenous switching regression (Model 4), ∂Participatei/
∂Agei¼ 0.0351754−0.0007086Agei¼ 0, and hence Agei¼ 50. Similar, from the probit regression
(Model 5), ∂Participatei/∂Agei¼ 0.0404751−0.0008076 Agei¼ 0 and hence Age¼ 50. In either case,
the second order condition for a maximum is satisfied.
References
Abdulai, A. and Delgado, C.L. (1999), “Determinants of nonfarm earnings of farm-based husbands and
wives in Northern Ghana”, American Journal of Agricultural Economics, Vol. 81 No. 1, pp. 117-130.
Abdulai, A. and CroleRees, A. (2001), “Determinants of income diversification amongst rural
households in Southern Mali”, Food Policy, Vol. 26 No. 4, pp. 437-452.
Akazili, J., Welaga, P., Bawah, A., Achana, F.S., Oduro, A., Awoonor-Williams, J.K., Williams, J.E.,
Aikins, M. and Phillips, J.F. (2014), “Is Ghana’s pro-poor health insurance scheme really for the
poor? Evidence from Northern Ghana”, BMC Health Services Research, Vol. 14, p. 637.
Alhassan, R., Nketiah-Amponsah, E. and Arhinful, D. (2016), “A review of the National Health
Insurance Scheme in Ghana: what are the sustainability threats and prospects?”, PLoS ONE,
Vol. 11, p. 11.
Asfaw, S., Shiferaw, B., Simtowe, F. and Lipper, L. (2012), “Impact of modern agricultural technologies
on smallholder welfare: evidence from Tanzania and Ethiopia”, Food Policy, Vol. 37, pp. 283-295.
Barrett, C.B. and Reardon, T. (2000), Asset, Activity and Income Diversification Among African
Agriculturist: Some Practical Issues, Cornell University, Ithaca, pp. 1-40.
Becker, S. and Ichino, A. (2002), “Estimation of average treatment effects based on propensity scores”,
Stata Journal, Vol. 2 No. 4, pp. 358-377.
Brinda, E.M., Andrés, R.A. and Enemark, U. (2014), “Correlates of out-of-pocket and catastrophic health
expenditures in Tanzania: results from a national household survey”, BMC International Health
and Human Rights, Vol. 14, p. 18.
da Silva, M.T., Barros, A.J., Bertoldi, A.D., Jacinto, P.D., Matijasevich, A., Santos, I.S. and Tejada, C.A.
(2015), “Determinants of out-of-pocket health expenditure on children: an analysis of the 2004
pelotas Birth Cohort”, International Journal for Equity in Health, Vol. 14, p. 53.
Dary, S.K. and Kuunibe, N. (2012), “Participation in rural non-farm economic activities in Ghana”,
American International Journal of Contemporary Research, Vol. 2 No. 8, pp. 21-28.
Dedehouanou, S.F., Ousseini, A., Harouna, A.L. and Maimounata, J. (2015), “Spillovers from off-farm
self-employment opportunities in rural niger”, PEP Policy Brief No. 128, December, available at:
https://portal.pep-net.org/document/download/25654
Dixon, J., Tenkorang, E. and Luginaah, I. (2011), “Ghana’s national health insurance scheme: helping
the poor or leaving them behind?”, Environment and Planning C. Government and Policy, Vol. 29
No. 6, pp. 1102-1115.
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
Fan, L. and Habibov, N. (2009), “Determinants of accessibility and affordability of health care in post- Nonfarm
socialist Tajikistan: evidence and policy options”, Global Public Health, Vol. 4 No. 6, pp. 561-574. enterprise
Ghana Statistical Service (2014), “Ghana living standards survey”, Report of Sixth Round (GLSS6). participation
Accra: Ghana Statistical Service.
Haggblade, S., Hazell, P. and Reardon, T. (2010), “The rural non-farm economy: prospecrs for growth
and poverty reduction”, World Development, Vol. 38 No. 10, pp. 1429-1441.
Jabo, M., Ismail, M., Abdullah, A. and Shamsudin, M. (2014), “The impact of non-farm income 29
generating activities on the food security status of rural households in Nigeria”, International
Journal of Agricultural Science and Veterinary Medicine, Vol. 2 No. 4, pp. 121-131.
Jehu-Appiah, C., Aryeetey, G., Spaan, E., de Hoop, T., Agyepong, I. and Baltussen, R. (2011), “Equity
aspects of the national health insurance scheme in Ghana: who is enrolling, who is not and
why?”, Social Science Medicine, Vol. 72 No. 2, pp. 157-165.
Korir, L., Legat, K. and Njehia, B. (2011), “The role of non-farm investments in agricultural risk
management in Kenya, current research”, Journal of Economic Theory, Vol. 3 No. 2, pp. 62-68.
Kousar, R. and Abdulai, A. (2013), “Impacts of rural non-farm employment on household welfare in
Pakistan”, 2nd AIEAA Conference – Between Crisis and Development: which Role for the
Bio-Economy, Parma.
Lokshin, M. and Sajaia, Z. (2004), “Maximum likelihood estimation of endogenous switching regression
models”, The Stata Journal, Vol. 4 No. 3, pp. 282-289.
Maddala, G.S. (1983), Limited-Dependent and Qualitative Variables in Economics, Cambridge University
Press, New York, NY.
Malik, A.M. and Syed, S.I. (2012), “Socio-economic determinants of household out-of-pocket payments
on healthcare in Pakistan”, International Journal for Equity in Health, Vol. 11, p. 51.
Mare, R.D. andWinship, C. (1987), “Endogenous switching regression models for the causes and effects
of discrete variables”, CDE Working Paper No. 87-32, Wisconson.
Murray, S. (2006), “Poverty and health”, Canadian Medical Association Journal, Vol. 174 No. 7, pp. 923-923.
Nagler, P. and Naudé, W. (2014), “Non-farm entrepreneurship in rural africa: patterns and
determinants”, IZA Discussion Paper No. 8008, Bonn: IZA, Institute for the Study of Labour.
O’Donnell, O. (2007), “Access to health care in developing countries: breaking down demand side
barriers”, Cadernos de Sau′de Pu′blica, Vol. 23, pp. 2820-2834.
Osarfo, D., Senadza, B. and Nketiah-Amponsah, E. (2016), “The impact of nonfarm activities on rural
farm household income and food security in the upper East and Upper West regions of Ghana”,
Theoretical Economics Letters, Vol. 6 No. 3, pp. 388-400.
Osondu, C.K., Obike, K.C. and Ogbonna, S. (2014), “Determinants of decision to non-farm
entrepreneurship by women farmers in Ikwuano Iga of Abia State”, European Journal of
Agriculture and Forestry Research, Vol. 2 No. 4, pp. 41-52.
Owusu-Sekyere, E. and Bagah, D.A. (2014), “Towards a sustainable health care financing in ghana: is
the national health insurance the solution?”, Public Health Research, Vol. 4 No. 5, pp. 185-194.
Owusu, V., Abdulai, A. and Abdul-Rahman, S. (2010), “Non-farm work and food security among farm
households in Northern Ghana”, Food Policy, Vol. 36 No. 2, pp. 108-118.
Parker, S. and Wong, R. (1997), “Household income and health care expenditures in Mexico”, Health
Policy, Vol. 40 No. 3, pp. 237-255.
Peters, D.H., Garg, A., Bloom, G., Walker, D.G., Brieger, W.R. and Rahman, M.H. (2008), “Poverty and
access to health care in developing countries”, Annals of the New York Academy of Sciences,
Vol. 1136, pp. 161-171.
Pritchett, L. and Summers, L.H. (1996), “Wealthier is healthier”, Journal of Human resources, Vol. 31
No. 4, pp. 841-868.
Reardon, T., Berdegu’e, J., Barrett, C. and Stamoulis, K. (2006), “Household income diversification into
rural nonfarm activities”, in Haggblade, S., Hazell, P.R. and Reardon, T. (Eds), Transforming the
Rural Nonfarm Economy, Johns Hopkins University Press, Baltimore, pp. 115-140.
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)
IJSE Rijkers, B. and Costa, R. (2012), “Gender and rural nonpfarm entrepreneurship”, World Development,
46,1 Vol. 40 No. 12, pp. 2411-2426.
Rosenbaum, P. and Rubin, D. (1983), “The central role of the propensity score in observational studies
for causal effects”, Biometrika, Vol. 70 No. 1, pp. 41-55.
Rous, J. and Hotchkiss, D. (2003), “Estimation of the determinants of household health care
expenditures in Nepal with controls for endogenous illness and provider choice”, Health
30 Economics, Vol. 12 No. 6, pp. 431-451.
Senadza, B. (2011), “Does non-farm income improve or worsen income inequality? Evidence from rural
Ghana”, African Review of Economics and Finance, Vol. 2 No. 2, pp. 233-244.
Tran, T.Q. (2015), “Nonfarm employment and household income among ethnic minorities in Vietnam”,
Economic Research-Ekonomska Istraživanja, Vol. 28 No. 1, pp. 703-716.
World Health Organization (2010), Health Systems Financing - a Path to Universal Coverage, World
Health Organization, Geneva.
World Health Organization (2016), Public Financing for Health in Africa: from Abuja to the SDGs,
World Health Organization, Geneva.
Yildirim, J., Yilmaz, E. and Korucu, N. (2011), “The determinants of out-of-pocket payments: evidence from
selected hospitals in Ankara, Turkey”, Applied Economics Letters, Vol. 18 No. 12, pp. 1159-1162.
You, X. and Kobayashi, Y. (2011), “Determinants of out-of-pocket health expenditure in China”, Applied
Health Economics and Health Policy, Vol. 9 No. 1, pp. 39-49.
Zahonogo, P. (2011), “Determinants of non-farm activities participation decisions of farm households in
Burkina Faso”, Journal of Development and Agricultural Economics, Vol. 3 No. 4, pp. 174-182.
Further reading
Habibov, N.N. and Fan, L. (2008), “Modeling prenatal health care utilization in tajikistan using a two-stage
approach: implication for policy and research”,Health Policy and Panning, Vol. 23 No. 6, pp. 443-451.
Corresponding author
Bernardin Senadza can be contacted at: bsenadza@ug.edu.gh
For instructions on how to order reprints of this article, please visit our website:
www.emeraldgrouppublishing.com/licensing/reprints.htm
Or contact us for further details: permissions@emeraldinsight.com
Downloaded by University of Ghana At 04:42 04 June 2019 (PT)