Do orphaned girls spend more time on water collection? Evidence from rural Zambia

ABSTRACT

Orphanhood has long-term adverse effects on human capital formation but its short-run effects are mixed. We examine the short-run effect of improved access to safe water on orphans’ outcomes in rural Zambia. We utilize a unique dataset collected in a quasi-experimental setting from a groundwater development project and employ a difference-in-differences approach. We do not detect any immediate effect from new water access on health and educational outcomes. However, we observe a larger increase in time spent on water collection and water-related household chores for orphaned girls than for non-orphans living with their biological mothers. We conclude that orphaned girls are disadvantaged in that they are more responsible for collecting water and related household chores under improved access to safe water, although the availability of safe water per se benefits the whole population.

KEYWORDS:

• Orphans
• fetching water
• time use
• borehole
• groundwater development
• Zambia

1. Introduction

After the outbreak of the AIDS pandemic in the 1980s, the growing number of orphans attracted many researchers to explore the disparity between orphans and non-orphans caused by a parent’s death in Sub-Saharan African countries. Previous research has revealed that orphanhood, especially maternal orphanhood, has long-term adverse effects on human capital accumulation (Beegle et al., 2006, 2010; Case & Ardington, 2006; Evans & Miguel, 2007) but the evidence relating to short-run effects on educational outcomes in particular has been mixed (Ainsworth et al., 2005; Ainsworth & Filmer, 2006; Bennell, 2005; Case et al., 2004; Yamano et al., 2006; Zimmerman, 2003). This is partly because the adverse effect is ameliorated by strong family support networks (Foster, 2000; Lloyd & Blanc, 1996).

This study examines the short-run effects of an intervention to improve access to safe water on the wellbeing of orphaned and non-orphaned girls in Zambia who have suffered from limited access to this resource in the past. The adverse effect of lacking access to safe water is disproportionally concentrated on children in terms of a higher incidence of waterborne diseases and a heavy water collection burden, the latter being particularly the case for girls. To our knowledge, there has been little research to uncover the underlying processes of any possible daily disadvantage for orphans that will materialize in the long run. We utilize a dataset collected in a quasi-experimental setting to detect any disparities between orphaned and non-orphaned girls in terms of health, education, and time allocation, the unique feature of our survey. We thus reveal the heterogeneous impact of a groundwater development project from orphanhood status and from those living with their biological mother, and this is the contribution of the paper.

2. Materials and methods

2.1 The project and data description

The target project is the second phase of a JICA financed grant aid project for groundwater development in Luapula Province, Zambia. 32% of households in the country cared for foster and/or orphaned children in 2018 and 10% of children under age 18 not living with a biological parent were orphans (Government of Zambia, 2020). Luapula Province is located in the Northern territory and 16.3% of children under the age of 18 there did not live with a biological parent. The province suffers from the lowest access to safe water in 2010 in the country (Central Statistical Office, Republic of Zambia, 2011). The incidence of diarrhea was slightly below the national average and school attendance was lower than the average in other rural areas in 2012. (Ministry of Health, Republic of Zambia, 2014).

The groundwater development project constructed borehole-based water supply facilities with hand pumps at 216 sites between 2012 and 2013. The new boreholes ensure that water was not contaminated at source because of the reasonable depth from ground level and was tested to satisfy the national standards before hand-over to the residents (Japan International Cooperation Agency, 2014). The survey was conducted in 94 sites in 3 districts (Milenge, Mwense, and Nchelenge) in the province. The target sites of the survey were randomly selected from the villages going through project implementation and the control sites were from the villages without project implementation but with similar demographic and socio-economic conditions to the target sites. If the project expectedly fails to obtain water from new boreholes, the target site was converted to a control site and a new site converted from other control sites into target sites if drilling was successful in getting water from the new boreholes (Japan International Cooperation Agency, 2014). In the end, the number of ‘project sites’ with water access was 64 and that of the control sites was 30.

Our data consist of a baseline survey carried out June to July 2012 and an end-line survey from June to August 2013, both of which were conducted in the dry season. Borehole facilities were not available at the baseline and were constructed between February 2012 and April 2013. For our survey, at the baseline, eight households were randomly selected from each site for a total of 752 households. One hundred and seventeen households could not be revisited during the end-line survey period. Both surveys used the same questionnaires that contained a wide variety of variables of individuals, households, and communities. The distinctive feature is to conduct a detailed time use survey for the whole day. To our knowledge, we provide the first evidence on the differences in use of time between orphans and non-orphans.

We confine the sample to households interviewed in both surveys (434 in the treatment group and 201 in the control group). We focus on girls aged 7 to 17 at the baseline (512 girls) and 440 girls living in 286 households were also interviewed at the end-line. Table 1 reports that 81.4% of non-orphaned girls lived with their biological mother in 2012 (82.0% in 2013). 7.5% of girls were non-orphans but living separately from their mother in 2012 (4.5% in 2013). The proportion of female orphans was 11.1% in 2012 (13.4% in 2013). Among the orphans, about half were paternal orphans living with their surviving mother and half were orphans living separately from their mothers including those who were maternal and double orphans.

Table 1. Child status and outcome variables

	Baseline (2012)		End-line (2013)
Orphan status and living arrangements	Obs.	%	Obs.	%
All	440	(100.0)	440	(100.0)
Non-orphans living with mother	358	(81.4)	361	(82.0)
Non-orphans living separately from mother	33	(7.5)	20	(4.5)
Orphans living with mother	22	(5.0)	29	(6.6)
Orphans living separately from mother	27	(6.1)	30	(6.8)
Outcome variables	Mean	S.D.	Mean	S.D.
Incidence of diarrhea symptoms	0.016	(0.125)	0.014	(0.116)
School attendence	0.800	(0.400)	0.709	(0.455)
Water collection (hours)	0.949	(1.105)	0.475	(0.867)
Water-related household chores (hours)	1.697	(1.543)	1.070	(1.261)

The sample is confined to girls aged 7 to 17 at the time of baseline survey and interviewed in both rounds.

Water-related household chores comprise water collection, washing, cleaning, and cooking.

Table 1 reports that the incidence rate of symptoms of diarrhea over the two weeks prior to the survey was 1.6% in 2012 (1.4% in 2013). School attendance rate was 80.0% in 2012 but had declined to 70.9% in 2013 as some girls graduated or dropped out of school. The average time spent for water collection per day was 0.95 hours in 2012, which decreased to 0.48 hours in 2013. Finally, the average time for water-related household chores comprising water collection, washing, cleaning, and cooking was 1.70 hours per day in 2012 and this had decreased to 1.07 hours in 2013.

2.2 Empirical strategy

This study employs a difference-in-differences (DID) approach. The central assumption for the methodology to be valid is the parallel trend between the project and control sites. The risk of an unexpected failure to find underground water and the replacement policy helped to assure the validity of the parallel trend assumptions because they contributed to the random assignment of the project. The balance test confirms no significant difference in all variables used in the estimation between project and control sites except a slight difference in the years of schooling for females.

We extend the simple DID specification to explore the heterogenous effects of the groundwater development project depending on child status and the living arrangements of girls. We employ the following specification by adding the interaction terms with dummy variables representing child status:

(1) $Y_{ijt}={\beta }_0+{\sum }_c{\beta }_1^c\cdot \left(S_j\times childstatu{s}_i^c\times t\right)+{\sum }_c{\beta }_2^c\cdot \left(S_j\times childstatu{s}_i^c\right)+{\sum }_c{\beta }_3^c\text{cdotchild}statu{s}_i^c+{\beta }_4\text{cdott}+X_{ijt}{\gamma }_1+X_{jt}{\gamma }_2+{\epsilon }_{ijt},$(1)

i refers to a child and j points to a site (or village) and t is time (t = 0 for baseline and t= 1 for end-line). Y_ijt is the dependent variable and takes four forms: (1) a binary variable that takes the value of 1 if a child had diarrhea in the past two weeks and 0 otherwise, (2) an indicator that takes the value of 1 if a child attends school currently and 0 otherwise, (3) a continuous variable of time in hours spent on water collection and (4) a continuous variable of time in water-related household chores comprising of water collection, washing, cleaning, and cooking. S_j is a binary variable that takes the value of 1 if the site has a successful borehole and 0 otherwise. $Childstatu{s}_i^c$ denotes a set of dummy variables that captures orphan status and living arrangements with mother shown in Table 1. $X_{ijt}$ is a vector that includes a set of household (individual) characteristics, and $X_{jt}$ is a vector containing a set of site j’s characteristics other than S_j including dummy variables to capture district-level fixed effects and the survey months. ${\beta }_0$ to ${\beta }_4$ and ${\gamma }_1$ and ${\gamma }_2$ are the parameters to be estimated. ${\epsilon }_{ijt}$ is assumed to be an i.i.d. error term. We employ the ordinary least squared estimation to obtain the coefficients. All standard errors are clustered at the village level. Table 2 reports on the summary statistics of the explanatory variables.

Table 2. Summary statistics

Baseline (2012)	Mean	S.D.	Min.	Max.
	(A)	(B)	(C)	(D)
Individual characteristics	n = 440
Age	11.5	(3.05)	7	17
Household characteristics	n = 286
Project site at the time of end-line survey (=1)	0.676	(0.469)	0	1
Female headed household (=1)	0.206	(0.405)	0	1
Age of the head	44.0	(11.58)	18	79
Highest education (years) among females	5.048	(3.041)	0	12
Highest education (years) among males	6.729	(2.996)	0	12
Household size	6.437	(2.157)	2	15
Ratio of dependents to household size	0.531	(0.189)	0	1
Monthly consumption per capita (thousand ZMK)	111.2	(80.1)	16.01	648
Value of durable assets (million ZMK)	1.527	(2.194)	0.013	29.40
Surveyed in June (=1)	0.664	(0.473)	0	1
Surveyed in July (=1)	0.336	(0.473)	0	1
Village characteristics	n = 94
Population	467.3	(524.9)	48	3360
Average asset per household (million ZMK)	6.313	(5.214)	1.448	29.30

The sample is confined to girls aged 7 to 17 at the time of baseline survey and interviewed in both rounds.

Monthly consumption per capita is adjusted by using adult equivalence scales and measured in the real term at the price level of 2012. Assets per household include the value of residence, residential and agricultural land, and durable assets. 1USD was worth approximately 5200ZMK as of June 2012.

3. Results

The coefficients of our interest are those in the interaction term among the project site dummy, the end-line dummy and child status with living arrangements. Table 3 reports that orphan status and living arrangements with their biological mother do not make any difference on the project’s impact on the incidence of diarrhea and school attendance. In contrast, the coefficient for water collection time is 0.528 (95% confidence interval [CI]: −0.063–1.120) for orphans living with their mother and 0.602 (95% CI: 0.002–1.203) for orphans living separately from their mother, suggesting that orphaned girls spend more time on water collection by 0.5–0.6 hours per day under the improved access to safe water. The coefficients are 0.252 (95% CI: −0.156–0.660) for non-orphans living with their mother and 0.383 (95% CI: −0.580–1.346) for non-orphans living separately from their mother.

Table 3. Estimation results

	Diarrheal incidence (=1)	School attendance (=1)	Water collection (hours)	Water-related chores (hours)
	(A)	(B)	(C)	(D)
Project, orphan status, living arrangements, and year dummy variables
Project site * Year 2013 * Non-orphan living with mother (=1)	−0.004	−0.042	0.252	0.544
	[−0.040, 0.032]	[−0.184, 0.100]	[−0.156, 0.660]	[−0.007, 1.095]
Project site * Year 2013 * Non-orphan living separately from mother (=1)	−0.051	−0.104	0.383	1.035
	[−0.141, 0.039]	[−0.331, 0.123]	[−0.580, 1.346]	[−0.263, 2.332]
Project site * Year 2013 * Orphan living with mother (=1)	−0.009	−0.015	0.528	0.580
	[−0.040, 0.021]	[−0.242, 0.213]	[−0.063, 1.120]	[−0.567, 1.726]
Project site * Year 2013 * Orphan living separately from mother (=1)	−0.012	−0.038	0.602	1.217
	[−0.038, 0.015]	[−0.308, 0.232]	[0.002, 1.203]	[0.337, 2.096]
Year 2013 (=1)	−0.004	−0.080	−0.738	−1.092
	[−0.033, 0.024]	[−0.197, 0.038]	[−1.103, −0.374]	[−1.583, −0.602]
Individual characteristics	Yes	Yes	Yes	Yes
Household characteristics	Yes	Yes	Yes	Yes
Village characteristics	Yes	Yes	Yes	Yes
R sq.	0.032	0.123	0.146	0.146
No. of children	440	440	419	419
No. of observations	880	880	838	838

95% confidence intervals in parentheses.

The sample is confined to girls aged 7 to 17 at the time of baseline survey and interviewed in both rounds.

Water-related household chores comprise water collection, washing, cleaning, and cooking.

When it turns to the project impact on the total time for household chores related to water, the coefficient is 0.544 (95% CI: −0.007–1.095) for non-orphaned girls living with their mother, whereas the coefficient for non-orphaned girls living separately from their mother is 1.035 (95% CI: −0.263–2.332). The coefficient for orphaned girls living separately from their mother is 1.217 (95% CI: 0.337–2.096), suggesting that they increased time spent on water-related chores by 1.217 hours per day. An additional analysis (not shown) that examines the difference in the project’s impact between orphaned and non-orphaned girls in consideration of living arrangements with their biological mother reveals that orphaned girls living separately from their mother spend more time on water-related household chores by 0.673 (95% CI: −0.128–1.474) hours per day, compared to non-orphaned girls living with their mother.

4. Discussion

We found that improved access to safe water did not have an immediate impact on the incidence of water-borne diseases or school attendance but increased time spent on water collection and household chores related to water for girls. We find a heterogenous effect from orphan status and living arrangements. While non-orphaned girls increased their burden on water-related household chores, orphaned girls increased their burden more than non-orphans and the lack of a mother further deteriorates their circumstances. Orphaned girls are disadvantaged in that they are more responsible for collecting water and related household chores under improved access to safe water, although the availability of safe water per se benefits the whole population.

5. Conclusion

We provide new evidence that orphaned girls are more responsible for collecting water and related household chores under improved access to safe water. We acknowledge some limitations of this study. Those limitations include the small number of orphans in our sample and the dated timing of data collection, though we do not see any evidence that circumstances surrounding orphans have changed in the subsequent time period. Further research should advance more evidence to fill the gap in understanding between short-run and long-run adverse effects on orphaned girls in considering the motives to foster children including orphans and the further possible daily discrimination such as that over nutrition intake.

Acknowledgments

This study was conducted at the JICA Ogata Sadako Research Institute for Peace and Development, which gave us formal permission to use the data. We thank Etsuko Masuko, Teruki Murakami and Taku Seo for their constructive comments. The views expressed in the paper are solely those of the authors.

Disclosure statement

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

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.