Data literacy in economic development

Abstract

In economic development and other economics electives, students regularly encounter economic measures of absolute and relative deprivation, from poverty measures like the Foster-Greer-Thorbecke index to measures of distribution like the Gini index. By “doing economics,” students practice applying economic measurement to real-world data and develop more general data literacy. The author proposes a series of exercises starting with stylized 10-household economies, proceeding to nationally representative cross-sectional surveys using MS Excel or Google Spreadsheets, and culminating in students applying their acquired data literacy to a team project. The data sources are easily tailored to alternative household surveys in low- and middle-income countries that include the required variables. Students learn data literacy through recognizing the properties of rectangular data, visualizing data appropriately, and creating aggregate economic measures.

Keywords:

• Data literacy
• development
• economic education
• inequality
• poverty

JEL CODES:

• A22
• I32

Acknowledgment

The author thanks Vis Taraz, Madeline Wettach, attendees at CTREE and LAC-DEV, and two anonymous referees for helpful comments on the article.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Although values of α larger than 2 can be used, their value in terms of instruction is limited. Unless a student has taken political philosophy, e.g., she would not understand that as α tends to infinity the value of the FGT index tends towards a Rawlsian Rule by which society judges poverty only on the basis of the poorest of the poor.

2 Five axioms for inequality measurement are considered important in the literature: 1) The Pigou-Dalton transfer principle, 2) income scale independence, 3) Dalton’s principle of population, 4) anonymity or symmetry, and 5) decomposability (see Cowell [2000] and Litchfield [1999]). Measures like the Atkinson index and the Theil Index (or the class of generalized entropy measures) are beyond this course.

3 See Palma (2011) and Cobham and Sumner (2013a, b) for a fuller explanation as to why the Gini index is insensitive to changes in the tails of the income distribution: the basic intuition is that substantial weight is given to deciles 2 through 5 of the income distribution and the tails are not given substantial weight in the Gini as each part of the distribution is equally weighted. This characteristic of the Gini is likely to be axiomatically true (Shorrocks 1980).

4 To give an accurate Gini coefficient, the cumulative income total should be divided by 5, which is the equivalent of the area of a smooth triangle. Note, though, that this produces a different Gini if a student cumulatively sums the heights of 10 columns, which would give a total area for A + B of 5.5. I accepted both answers in the class exercise, although the cumulative sum of columns results in a lower Gini coefficient.

5 Taylor and Lybbert (2015) include the Mexico Rural Household Survey in their supporting material for their textbook along with some exercises similar to those I ask students to perform with these data. However, my exercises are grounded in initial exercises in-class as presented earlier, along with additional questions and ideas that Taylor and Lybbert do not include.

6 A variety of work has used the NIDS to understand South African poverty. See Leibbrandt et al. (2010), Leibbrandt and Levinsohn (2011), Finn and Leibbrandt (2013a), and Jansen et al. (2014).

7 Because each row corresponds to a household and we cannot easily manipulate this to create an individual-level dataset, we can compute only a household income-based Gini index. Although this is unfortunate from the perspective of consistency, in terms of pedagogy and learning outcomes the main concern is for students to think through what is required to find a Gini index and what the unit of observation implies about what gets computed.

8 I cannot provide the full NIDS data (for 2008 or 2014/15) as a spreadsheet, but this value is produced from the NIDS data in the spreadsheet that I can provide to those interested who request the data from the DataFirst at the University of Cape Town (www.datafirst.uct.ac.za). The Gini is found using the covariance formula as shown in the supplementary material. As with poverty, several authors have used the NIDS to examine inequality in South Africa and found results similar to those that I found, such as Leibbrandt, Finn, and Woolard (2012), Finn and Leibbrandt (2013b), and Finn, Leibbrandt, and Levinsohn (2014). Admittedly, the measures should be weighted by their survey weights, but such a discussion is beyond the ambit of the course I teach.

9 Showing students shortcuts to highlight many cells at once, e.g., CMD + SHIFT + DOWNARROW on a Mac or CTRL + SHIFT + DOWNARROW on a PC with Windows, results in them saving time. I also recommend demonstrating to students how to double-click on the bottom corner of a cell for a formula to be applied to a whole column.

Additional information

Funding

Support was received from the TIER Faculty Fellowship.