Written by Gracie Rosenbach and Emily Schmidt
In June 2019, IFPRI researchers Emily Schmidt, Gracie Rosenbach, and Rachel Gilbert partnered with the Australia National University and the University of Papua New Guinea (UPNG) to teach a 3-day evening seminar for the Master of Economics and Public Policy Program in the UPNG School of Business and Public Policy titled Describing, Transforming, and Analyzing Data Using Stata. Approximately 40 students participated in the seminar, the majority of whom are also full-time government analysts in various departments, including the Department of the Prime Minister and the National Executive Council; the National Planning and Monitoring; Agriculture and Livestock; the Central Bank; and Higher Education; among others. The goals of the training were to: give the participants an overview of Stata so that they could use this tool when preparing their masters’ theses; disseminate and encourage the use of the data from the IFPRI Papua New Guinea (PNG) Household Survey on Food Systems (collected in May-July 2018); and share survey findings with the participants that may be informative for their own policy planning processes within their roles in various government departments.
The training covered topics related to describing, transforming, and analyzing data using Stata, and participants practiced with the data from the IFPRI Papua New Guinea Household Survey on Food Systems. By the end of the training, participants were able to use Stata to both analyze and visualize their results by using tabulations, summary statistics, and by creating tables and figures. Participants learned additional specific skills, such as how to identify data inconsistency issues of the raw data, and how to use various commands to target their analyses based on their research questions.
Interspersed throughout the training were presentations on initial results by IFPRI researchers from the survey on the topics of poverty, anthropometry, and non-farm enterprises. These results, as well as research and insights from other organizations, were also presented at a half-day workshop that same week in order to engage a larger audience of government officials and development partners. The purpose of these presentations during the training was two-fold: first to show how we used Stata to analyze and present the results related to our research questions, and second to highlight key findings from the survey that the participants and their government departments can incorporate into their evidence-base. Following the presentations, the participants were given the code to reproduce a variety of figures from the analyses as a way to highlight these functions of Stata, and for them to be able to create similar figures of their own using survey data.
On the first day of the training, Emily Schmidt presented an overview of the objectives and sampling strategy of the IFPRI Papua New Guinea Household Survey on Food Systems. She explained that the survey targeted rural households in only a few districts within the provinces of East Sepik, Madang, West Sepik, and the Autonomous Region of Bougainville, underlining that the results from the survey are not representative at the national or provincial level, but still provide a holistic picture of rural households in these specific areas. This was followed by a classroom discussion about what key topics the survey data can inform. Given that the survey included detailed consumption and expenditure modules, we were able to calculate a spatially-adjusted poverty line for each province in the dataset in order to evaluate poverty prevalence by survey area (Table 1). These data can then be paired with a variety of other data that was collected in the survey including information on agricultural production, labor profiles, migration activity, perceptions of poverty, and anthropometry measurements.
Table 1: Share of the survey sample that is poor by survey area
| Poverty headcount (percent)
Poverty line |
Poverty line
+10% |
Poverty line
-10% |
|
| ARoB | 55.9 | 61.7 | 49.2 |
| East Sepik | 47.4 | 52.1 | 42.6 |
| Madang | 44.6 | 52.9 | 37.7 |
| West Sepik | 46.6 | 52.8 | 41.3 |
| Total* | 48.1 | 54.5 | 42.2 |
Source: Authors’ calculations from the IFPRI Papua New Guinea Household Survey on Food Systems, 2018. ARoB is the Autonomous Region of Bougainville.
Rachel Gilbert presented the second day of the training. She discussed the survey data related to child growth and development outcomes analyzed from the anthropometric data. The data analysis suggests that 32% of children under 5 years of age in the survey households are stunted in their growth and 9% of the children in the sample are wasted. This stunting rate is slightly above the World Health Organization threshold of 30% which is considered “a very high prevalence”. However, these stunting rates are applicable only to the very specific sample in the survey, and so they are not comparable to previous studies. Child growth was significantly associated with the mother’s height and body mass index (BMI) measurements (often correlated with health outcomes), and family planning practices (including birth interval/spacing). Child growth was measured as a child’s height-for-age z-score (HAZ), which tells the deviation of a child’s height from the global average for other children of that child’s gender and age. The HAZ is also used to determine whether or not a child is stunted. The survey finds that taller mothers and mothers with higher body mass indices were more likely to have children with higher HAZs (indicating better child growth). Additionally, older siblings (children who have a lower birth order, e.g. were born first) were more likely to have higher HAZ scores, as were children who were born after a longer birth interval (e.g. the mother waited for more than 24 months in-between pregnancies). These results suggest that women’s health is crucial for ensuring children’s health and therefore improving the overall health of the country over time.
Figure 1. Stunting and wasting prevalence by survey area
Source: Authors’ calculations from the IFPRI Papua New Guinea Household Survey on Food Systems, 2018. ARoB is the Autonomous Region of Bougainville.
On the third day of training, Gracie Rosenbach presented recent findings from the survey on the importance of non-farm enterprises, and the differences between enterprises owned by men and women. More than one-third of households in the sample have a non-farm enterprise (that is, earned off-farm income aside from wage employment, such as selling prepared food or petrol, or sewing clothes) (Figure 2). There were similar numbers of male, female, and joint-owned enterprises in the sample, however, female-owned enterprises had considerably lower earnings than the other ownership types. Regression analysis suggests that there may be additional differences between these ownership types and their reasons for starting their enterprise. For example, additional household labor appears important for all enterprise types, however having additional income is associated with a household having a male-owned enterprise, while having experienced a drought in the last 5 years is associated with a household having a female-owned enterprise. These results suggest that female-owned enterprises are important for a household’s food security since they may be started as a risk diversification strategy.
Figure 2. Household income types by survey area and poverty status
Source: Authors’ calculations from the IFPRI Papua New Guinea Household Survey on Food Systems, 2018. ARoB is the Autonomous Region of Bougainville, HHS is households, and NFE is non-farm enterprise.
The three-day course aimed to increase capacity among key government analysts, as well as present and discuss survey results in a setting that afforded greater discussion of overall findings and interaction among participants. Throughout the course, the participants challenged research hypotheses and analysis decisions. The most frequent feedback received from the course evaluation was that the course should be longer in order to have more time to practice the Stata skills that were being taught. The participants were incredibly enthusiastic about the evidence-based policy possibilities of utilizing the data from the IFPRI Papua New Guinea Household Survey on Food Systems. A follow-up course is planned for mid-October 2019 to continue to explore and analyze PNG data to best inform national and local policies to improve food security.