Social norms and the gender gap in labor force participation: Evidence from Turkey

ABSTRACT

We use a novel two-step empirical strategy to examine the relationship between social norms and the gender gap in labour force participation (LFP) across provinces of Turkey. In the first step, we identify the unexplained part of the gender gap in LFP that remains after accounting for differences in observed characteristics between women and men for each province by implementing a decomposition method. In the second step, we investigate the role of social norms in explaining cross-province variation in the unexplained part. The results reveal that more egalitarian gender role attitudes, smaller gender gap in tertiary education, lower fertility and consanguineous marriage rates, and lower level of religiosity significantly predict a smaller unexplained part of the gender gap in LFP favouring males.

KEYWORDS:

• Gender gap in labour force participation
• social norms
• decomposition analysis
• Turkey

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/13504851.2022.2094315.

Disclosure statement

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

Notes

¹ See Ilkkaracan (2012) and Aldan (2021) for a review of studies that explore the drivers of female labour force participation in Turkey. To the best of our knowledge, there is no study that empirically examines the determinants of the gender gap in labour force participation in Turkey using a decomposition method.

² Although Turkish Statistical Institute has conducted the TLSS every year since 2003, the 2013 TLSS is the only survey representative at provincial level, allowing us to examine the gender gap in labour force participation across 81 provinces of Turkey.

³ $L{P}_m$ ($L{P}_w$) is a dummy variable that takes the value of 1 if the man (woman) participated in the labour force and 0 otherwise.

⁴ Table A.1 of Appendix presents the definition of variables used in the individual-level analysis and descriptive statistics based on the whole sample that includes all the 81 provinces by gender.

⁵ Since the logit model includes a constant term, the average value of the outcome of interest for men (women) must be equal to the average value of the predicted probabilities in the men (women) sample (i.e. ${\overline{LP}}_m=\overline{G(X_m{\stackrel{ˆ}{\beta }}_m)}$ and ${\overline{LP}}_w=\overline{G(X_w{\stackrel{ˆ}{\beta }}_w)}$).

⁶ We use data from the 2013 Turkish Demographic and Health Survey to create the gender role attitudes index. In this survey, women are asked whether they agree with several statements that capture their attitudes towards gender equality in various life domains. First, using answers to those seven survey questions, we assign a value of 1 to agreement (disagreement) with each gender-egalitarian (gender-nonegalitarian) statement and 0 otherwise. Second, the index is created by summing up the seven indicator variables. The higher the value of the index, the more egalitarian gender role attitudes the woman has. Third, to obtain region-specific gender role attitudes, we take the average value of the gender role attitudes index at the NUTS 2 level in which 81 provinces of Turkey aggregated into 26 regions. Unfortunately, the data are not available at the provincial level. See Table A.2 in the Appendix for the detailed information about the creation of the index.

⁷ Qur’an courses, offered by the Presidency of Religious Affairs, aim to provide individuals with Islamic religious knowledge, teach how to read Qur’anic verses in Arabic, and help them memorize parts of the Qur’an that are recited in religious prayers. The US Department of State’s International Religious Freedom Report (United States Department of State 2014) states that according to the Turkish government, 99% of the population is Muslim, the vast majority of which is Sunni Muslim. Representatives of other religious affiliations, on the other hand, claim that the actual percentage of Muslims is slightly lower than 99% and members of other religious affiliations, most of whom live in Istanbul and other large cities, constitute less than 1% of the population.

⁸ Table A.2 in the Appendix presents a detailed description of all the regional-level explanatory variables used in Equation 4(4) ${\mathrm{\Delta }}_p^u={\beta }_0+{\beta }_1{G}_r+E_p\delta +F_p\gamma +{\beta }_2{R}_p+X_p\alpha +{\lambda }_f+{\epsilon }_p$(4) , their data sources and descriptive statistics.

⁹ A statistically significant positive(negative) explained part indicates the expected decrease(increase) in the gender gap in LFP if men had the same endowments as women.

¹⁰ Since the correlation coefficient between fertility rate and consanguineous marriage rate is 0.92, we include these two dimensions of family structure separately.

¹¹ The discouraged worker hypothesis predicts a negative relationship between unemployment rate and female labour force participation. Tansel (2002) presents evidence in favour of the discouraged worker hypothesis using data for 67 provinces of Turkey for the years 1980, 1985 and 1990.