The effect of education on household incomes using the Mincerian approach: a comparison between MENA and the rest of the world

ABSTRACT

This paper uses the Mincerian approach in an experimental way to examine the impact of education on household incomes (not labor earnings) of all workers (not just employees) across 162 countries. Our results are broadly similar to the conventionally estimated rates of return to education after allowing for the fact that earnings are only a part of total household incomes. We then apply the results to the case of Arab countries and find that our experimental approach can promisingly be used for the study of the impact of education not just on individual earnings of employees but also on total household incomes of all workers and more broadly on the economy and the labor market.

KEYWORDS:

• Rates of return to education
• household incomes
• labor markets
• Arab countries
• MENA

JEL CODES:

• C13
• I20
• I26
• J01
• J31

Acknowledgement

Earlier versions of the paper were presented at the ERF 20th Annual Conference (Cairo), the ERF/SILATECH/Development of Euro-Mediterranean Economic Research (DREEM) meeting “The Pulse of the Arab Street” (Paris), and the 5th Conference of the French National Center for Scientific Research (CNRS) (Dauphine University). The authors would like to thank Francois Bourguignon, Jad Chaaban, Chris Dougherty, Mourad Ezzine, Ahmad Galal, Hafez Ghanem, Nader Kabbani, Mohamed Ali Marouani, El Mouhoub Mouhoud, Jeff Nuggent, Aysit Tansel, as well as the editors and two anonymous referees for their comments and suggestions, and especially Harry Anthony Patrinos who made available the World Bank data used in the analysis.

Disclosure statement

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

Notes

1 Mincer (1974). For global reviews, see Psacharopoulos (1973; 1985; 1994), Psacharopoulos and Patrinos (2004; 2018) and Montenegro and Patrinos (2014).

2 Recent examples include Jean, Montenegro, and Orazem (2007) that covered 49 countries (4 Arab countries), King, Montenegro, and Orazem (2010) that covered 86 countries (7 Arab countries), and Montenegro and Patrinos (2013; 2014) that covered 139 countries (8 Arab countries). Regional or global reports of the World Bank, and other international and regional organizations, typically include comparative tables on RORE – see for example the report on MENA by the World Bank (2008). The aforementioned reviews reported initially RORE for Egypt, Jordan, Morocco Tunisia and Yemen, then they included Djibouti and Iraq and more recently added Lebanon, Mauritania, Syria, West Bank and Gaza all of which amount for half of the 22 Arab countries.

3 Montenegro and Patrinos (2014).

4 ILO (2018).

5 Chiswick (1977) quoted in Chiswick (2023).

6 Chiswick (2023, 27).

7 As noted in the next section, the MENA region includes Iran (not an Arab country) and only a subset of the Arab countries.

8 See Tzannatos, Diwan, and Ahad (2016).

9 Fuller (2003), Gatti et al. (2014), Daffron (2016), Assaad et al. (2018), Tzannatos (2021), OECD (2022).

10 The surveys were conducted between 2005 and 2013 with most been closer to the more recent years.

11 South Sudan became independent from Sudan in mid-2011, after the surveys used in this paper were conducted.

12 Iran has a population of nearly 90 million while Lebanon has 5.3 million, Jordan 11 million, Iraq is 45 million, Palestine (West Bank and Gaza) 4.5 million and Syria is 23 million. On the GCC side, the total population is less than 50 million of whom nearly half are expatriates.

13 However, in another classification (Little Data Book) the World Bank includes the GCC countries in the high-income group but not in the MENA group which is the classification in The MENA Data Book.

14 ILO/UNDP (2011).

15 Chiswick (1977) quoted in Chiswick (2023).

16 Chiswick (2023, 27).

17 The surveys include information on total annual household income before taxes as reported by the respondent who can be of any age more than or equal to 15 years and may or may not be working. Household income includes wages and salaries, remittances from family members living elsewhere, labor earnings and all other sources. It is converted to international dollars in 2009 purchasing power parity. See Tzannatos, Diwan, and Ahad (2016) for a full description of the sample and variables.

18 This assuming that there is no heteroscedasticity in the measurement error. This is a reasonable assumption since we have converted the dependent variable into logs. The measurement error should have a positive mean, but this does not matter for the present analysis. See Dougherty (2011, 309–310).

19 Sloane et al. (2003).

20 Chiswick (2023).

21 For a review see Bairagya (2020).

22 Psacharopoulos, Arriagada, and Velez (1992).

23 Idrus and Cameron (2000).

24 Demery and Chesher (1993).

25 Fairlie and Meyer (1996).

26 Clain (2000).

27 Williams (2003).

28 Garcia-Mainar and Montuenga-Gomez (2005).

29 Psacharopoulos (1994).

30 It should also be noted that even when using years of schooling, it does not necessarily mean that more years imply more learning, for example, in the presence of automatic progression and/or mandatory school leaving age.

31 See Miles (1997, 7) who states ‘we also assume (following Mincer 1958 and numerous outer studies since) that the log-household income is a quadratic function of age’.

32 Miles (1997) in his study of the determinants of household incomes in the UK uses the same independent variables as we do.

33 We have no explanation for this though a possible one is that interviews conducted during weekends took place in the presence of more household members (compared to an individual respondent) who collectively had more information about all sources of income in their households. Another explanation is that respondents felt they should state higher levels of income that the actual ones in the presence of other household members.

34 See Appendix Table A1b in Tzannatos, Diwan, and Ahad (2016).

35 According to Patrinos (2016) ‘age is commonly used to approximate experience’ though the author rightly adds that the number of years spent in school should be taken into account in principle. In practice information is on experience may not be readily available and researchers commonly use ‘potential’ experience (age minus years of schooling minus school entry age or age minus the time at which the workers left schooling). Potential experience is problematic as it assumes that individuals have been at work continuously since they left school thereby resulting in a serious measurement error especially in the case of women and male unemployment. In addition, children may not start school at a specific age due to, for example, illness, cost, or distance, or when there is a high incidence of serial repeaters leading to measurement error in using years of potential experience. This error can be especially prevalent in developing countries which form the majority of our sample (ibid.) Finally, from an econometric perspective, potential experience includes schooling that is correlated with the error term leading to biased and inconsistent estimates (Card 1999).

36 Miles (1997).

37 The inclusion of employment sector variables can be contested both theoretically (endogeneity) and econometrically (multicollinearity). The reason is that education, earnings and type of employment are simultaneously considered in making human capital investments and more generally labor supply decisions. Ideally, one would like to have more adequate data than we do for jointly estimating education, earnings and sector of employment and, especially in the case of women, decisions regarding whether to participate or not in the labor force.

38 Tzannatos, Diwan, and Ahad (2016).

39 Like Montenegro and Patrinos, we adopt a parsimonious specification of the regression equation being well aware that there are differing functions of education in countries at different levels of development (Vandenbussche, Aghion, and Meghir 2006). Also, this specification assumes that an additional year of educational attainment increases the human capital stock by an equal amount without distinguishing whether that additional year corresponds to an additional schooling attainment at the elementary level or at the university level (Wößmann (2003). In addition, there can be cross-country issues of measurement errors, omitted variables and heterogeneous responses that call for the use of more sophisticated specifications than OLS, like instrumental variables or two-stage least squares (Angrist and Krueger 1991; 2001).

40 ILO (2018).

41 In EAP there is information on earnings for East Timor, Palau, Papua New Guinea, Solomon Islands and Tuvalu but not for Myanmar and Vietnam that are included in the Gallup surveys.

42 The two countries are Belize and Guyana.

43 In fact, the public sector is the only sector showcasing a balanced distribution of men and women … with women accounting for half of the workforce in 2023. See World Economic Forum (2023). The report further notes the concentration of women in healthcare, care services, education and consumer services.

44 As the RORE are found to be higher for women than men, it is generally claimed that investment in girls’ education can contribute more to economic growth than investment in boys’ education. This is not necessarily the case as RORE in the case of women reflect differences between more educated and less educated women, while the earnings of women are lower than those of men. According to the ILO (2019), women earn on average about 20 per cent less than men across the world. Applying the RORE reported by Montenegro and Patrinos in the table, an extra year of education increases the earnings of men and women by exactly the same amount i.e. 9.1 percent (100*9.1 and 80*11.4). This assumes that earnings reflect productivity as do indeed the RORE. Of course there is a myriad of reasons justifying to invest more in girls’ education (World Bank 2012).

45 Naguib and Aref (2024).

46 Tzannatos (2008), table 4A.8.

47 ILO (2018).

48 The measurement of labor shares and their movement over time is a debated economic issue. The inclusion of China can change their values significantly (Bai and Qian 2010). For some regions, save Africa, the labor share is on average above 50 percent rising to more than 55 percent in the Americas, Europe and Central Asia. These estimates are challenged as too low by Gollin (2002) who argues that international cross-section data fail to take into account the labor income in small firms and concludes that the shares for most countries lie between 65 and 80 percent. Guerriero (2019) following the theoretical conceptualization of Atkinson (2000), constructs six different indices for measuring labor shares and applies them to more than 150 economies and finds that most labor shares lie between 60 and 70 percent.

49 Mincer (1974), Card (1999).

50 Angrist and Krueger (1991). In that study, the authors also note that in the specific US case they ’find a remarkable similarity between OLS and TSLS estimates [and] differences between OLS and TSLS estimates are typically not statistically significant’ (27–28).

51 Ibid. In a subsequent paper, Angrist and Krueger (2001) explore the usefulness of using instrumental variables estimation (IV) and two-stage least squares and how these two approaches can adress issues of measurement errors, omitted variables and heterogenous repsonses in OLS estimation. Though these more elaborate methods are not void of pitfalls (ibid. pp. 79–80) they can be used more readily in country case studies than in the case of global comparisons where the interest is more on cross-country patterns than levels in specific countries.

52 Psacharopoulos (1994).

53 The average year of the surveys in Table 3 can be calculated to be 2002.

54 ILO (2019).

55 Schwartz (2010).

56 World Bank (2016).

57 For Kuwait using data for 1983 Psacharopoulos (1994) reports a RORE of 4.5 percent, while for UAE the reported rate was 5.5 percent for men and 2.6 percent for women in 2009 (Vazquez-Alvarez 2010).

58 As primary education is nearly universal, there is no much difference across countries. Using secondary education as an indicator, the enrolment rate in Sub-Saharan Africa was 44 percent in 2022 compared to the world average of 77 and 73 percent among Arab states (http://data.uis.unesco.org/index.aspx?queryid=3813).

59 There are several other observations that can be made with respect to individual countries included in Table 4. For example, the RORE tend to be higher where the business climate is better and the private sector more active, with Morocco’s highest values being a case in point.

60 Montenegro and Patrinos (2022).

61 Tzannatos (1999).

62 World Bank (2003).

63 World Bank (2013); Assaad et al. (2018).

64 Pritchett (1999).

65 Assaad (2014) also confirms that importance of low labor demand in the MENA region.

66 Cammett et al. (2015).

67 See Schultz (1975) and for an empirical investigation confirming his hypothesis that human capital is more valuable in countries with greater economic freedom, see King, Montenegro, and Orazem (2010).

68 Nabli et al. (2006).

69 Baduel, Geginat, and Pierre (2019).

70 Galal (2017).

71 Diwan, Malik, and Atiyas (2019).

72 Devarajan (2016).

73 Ummuhan, Huitfeldt, and Wahba (2006).

74 Diwan and Akin (2015).

75 ILO/UNDP (2011).

76 Dessus (2001) notes that, in addition to the effect from reduced public spending on education, the average quality of education may have been reduced further by the education expansion as less academically motivated or able students enrolled in schools over time.

77 For example, the proportion of public sector employment in the first jobs of women with secondary education and above, the prime group for seeking government employment, was more than 75 percent in Tunisia and Egypt and nearly 60 percent in Jordan in the 1970s but has declined to around 30 percent in all three countries (Assaad 2014).

78 ILO/UNDP (2011).

79 Razzaz (2013).

80 Nabli et al. (2006).

81 Nabli (2007).

82 Arab countries have the lowest scores in the index of government accountability something that is compatible with the outcome of the privatization. This provides some explanation about the causes of the Arab Spring as does also the diverging fortunes between ordinary citizens and the establishment’s insiders. This is now however, the focus of this paper and more can be found in ILO and UNDP (2012) and Tzannatos (2021).

83 See collection of paper in Diwan, Malik, and Atiyas (2019).

84 For example, see the case of Egypt in Diwan, Philip Keefer, and Schiffbauer (2020); Schiffbauer (2014).

85 Tzannatos (2022).

86 Diwan and Haidar (2021).

87 Baduel, Geginat, and Pierre (2019).

88 Galal and Kanaan (2010).

89 World Bank ICA surveys as quoted in ILO/UNDP (2011).

90 Fisch (2007) presented at Sony annual 2009 shareholder meeting-Innovation America

91 Effectively the skills shortages as stated by GCC countries refer to their own citizens, the majority of whom are given jobs in the public sector based more on nationality and entitlement than merit and productivity.

92 Almeida and Aterido (2010).

93 Gatti et al. (2014, 61).

94 World Bank (2008).

95 ILO and UNDP (2012).

96 Tzannatos (2021).

97 World Bank (2018).

98 Salehi-Isfahani and Murphy (2006).

99 Diwan (2016). Associated with this approach to education is the fact that social dialogue is subdued in the Arab region (Tzannatos 2014).

100 Diwan, Tzannatos, and Akin (2018).

101 World Bank (2003).

102 Laszlo (2005).

103 Devarajan (2016).

104 ‘Quality’ applies to all levels of education. ‘Low relevance’ to the labor market needs applies mainly to postsecondary education.

105 See, for example the collection of papers in ERF (2015).

106 The legal differences in the treatment of women and men include the right to apply for passport, ability to be head of household, choice where to live, conferring citizenship to children, getting a job without permission, freedom to travel and so on. See World Bank (2016).

107 Kawar and Tzannatos (2013).

108 Kawar and Tzannatos (2017).

109 Tinbergen (1975).

110 See papers in Diwan, Malik and Atiyas (2019).

Additional information

Funding

This research was financially supported by the Economic Research Forum (ERF, Cairo), Silatech (Doha) and UNDP.