Gender Norms, Labor Productivity, and the
Replacement Effect
Li-Ling Liang
12018-08-15
Research Master Thesis
Faculty of Economics and Business
University of Groningen
Abstract
This thesis studies the effect of gender norms on labor productivity. I hypothesize that more balanced gender norms encourages more higher-productivity women to join the labor market, which raises the average level of labor productivity. Using cross-section data at the regional level for 28 European countries, I provide evidence that gender norms is positively related to labor productivity. This effect is particularly present in male-dominated sectors, where more balanced gender norms encourage higher-productivity women to enter the sector, and replace poorer-performance men at the lower end. In neutral sectors there is no clear evidence of such replacement effect. In female-dominated sectors the effect is negative. These results suggest that unequal gender norms can cause misallocation of human resources.
Keywords: gender norms, labor productivity, misallocation JEL D02, D79, J16, Z13
2
I.
Introduction
Why do people in some countries produce much more than others? In the influential paper of Hall and Jones (1999), the authors distinguish between proximate factors and fundamental causes of economic development. Using a development accounting approach, they consider physical and human capital inputs as merely the proximate factors that affect production. There are deeper, more fundamental underlying structures which exert far more influence on labor productivity.
3
labor productivity against the five culture dimensions one by one. Next, I include all five cultural dimensions in one regression as explanatory variables to see which variable(s) remain significant. Through this process, I find that gender norms are the most relevant cultural dimension affecting my dependent variable, which is labor productivity
After demonstrating the importance of gender norms, I focus on the question: what mechanism makes gender norms play such an important role in labor productivity? I hypothesize that more balanced gender norms raise labor productivity because more higher-productivity women are encouraged to work in the labor market, which increases the level of average labor productivity. This conclusion of the thesis is in line with results of recent work such as Fortin (2005), which shows that gender-specific factors are an important determinant of female labor force participation, and Cooray and Potrafke (2011), which shows that culture is one of the key determinants of gender inequality. Other works such as Fernandez (2007), Fernandez and Fogli (2009) also give out similar message as the main findings of this thesis: that gender norms is an important factor that determines economic outcomes.
4
to encounter a female-unfriendly working environment, perhaps also doubts about her ability and discrimination. Because she must be able to overcome all these obstacles before entering the industry, she must be much more productive than the average men in the industry. Consequently, if entry of women causes some men to leave the industry, it is more likely that the higher productivity minority female drives out the males at the lower end.
To test my hypothesis of the replacement effect, I use cross-section data at the regional level of 28 EU countries. I find that although the effect of gender norms on labor productivity is generally positive, the presence of the replacement effect depends on the share of female employment in the sector. In other words, only in sectors with low percentage of female employees does the replacement effect appear to be strong, which is intuitively straightforward. Because female share of employment in each sector is different, all sectors can roughly be categorized into three types: male-dominated sectors, in which female share of employment is significantly below 50 percent, neutral sectors, in which male and female share of employment are roughly equal, and female-dominated sectors, in which female share of employment is larger than 50 percent. The replacement effect is observed in most of the male-dominated sectors and the result is robust for models applying alternative fixed effects. Yet in the neutral sectors, the replacement effect is not obvious and in the female-dominated sectors, the effect is negative. The findings show that cultural dimension indeed have real economic consequences.
5
and interprets the outcomes; Section VII concludes.
II.
Gender norms and Economic Development
In this section an introduction on the five cultural dimensions used for empirical analysis in Section III will be given. Following the study of Beugelsdijk et al. (2017b), the five cultural dimensions chosen are trust, work norms, market attitudes, attitudes towards democracy, and gender norms. The reasons why these cultural norms are important and their relation with economic development will be also be explained in this section. However, most part of this section will be dedicated to gender norms since gender norms are the focus of analysis of the whole thesis.
Gender norms
Gender norms are commonly agreed codes of behaviors for both men and women in a society. These expectations about what men and women should be like and behave usually regulate people’s behaviors and affect their economic decisions even though rational agents may not be fully aware of it. For example, different expectations about men and women’s career would affect families’ education decisions for their sons and daughters. Different gender roles that men and women are expected to play (men as bread earners; women take care of the households) would affect females labor force participation decisions. Discrimination against women may cause males and females to be offered with different pay for the same work, etc.
6
land. In areas where plough cultivation was employed, people use ploughs and sometimes animals to pull the ploughs to clean the land before planting crops. Because plough cultivation is more capital intensive and require stronger upper body strength and gripping power to control the plough, men have advantage over women in the agricultural work. Alesina et al. (2013) has found evidence that the distribution of hoe and plough cultivation areas around the world in the past roughly corresponds to the distribution of gender norms attitudes in the modern days. While plough cultivation gives males advantage over females in agricultural works, which in turns limited women to domestic works, hoe cultivation areas do not have such distinction, and gender norms in these areas are more balanced.
Apart from its persistency, another property which distinguish gender norms from other cultural norms is that it affects people’s identity by shaping how economic agents regard themselves as they choose what social role(s) to play (Akerlof and Kranton, 2000). The idea of behaving in a way that is consistent with one’s own social role(s) usually matters for important decisions in life such as education decisions (including how much and what kind of education to take) and career choosing. In other words, in regions where gender norms are more balanced, women on one hand are less likely to be confined in the domestic sphere and more likely to take part in the non-domestic production activity, on the other hand are less likely to be restricted to the social label attached to them to behave like women. Consequently, in these areas women with higher labor productivity are more likely to enter to the labor market and increase the average labor productivity.
7
employment, and pay gaps caused by gender norms usually have negative effect on economic growth and development. The reason why education and employment gaps are bad for economic growth is quite straightforward: more education increase human capital, while higher employment of female may raise total productivity.
Knowles et al. (2002) builds a neo-classical model which allows male and female education decisions to vary. With education gaps between male and females entering the model, their empirical study shows that female education has a positively significant effect on labor productivity. Similarly, the study of Klassen (2002) and Klassen and Lamana (2009) also find that gender gaps in education and employment has a negative effect on economic growth because the gaps significantly lower the average level of human capital. In addition, gender pay gaps caused by unequal gender norms may also result in negative effect on economic growth.According to Galor and Weil (1996), large gender pay gap can reduce female employment and increase fertility, which lowers economic growth from a long-term perspective.If gender norms become more balanced, the gaps mention above would shrink. The cross-country study of Liang (2017) shows that as gender norms becomes more balanced, females’ average years of education relative to males’ increases, and the results hold both in Europe and around the world. With respect to pay gaps, when gender norms become more balanced, people would also be less tolerant to different payment for the same work between men and women.
8
usually have lower productivity and are less likely to participate in commercial market than men due to difference access in inputs, resources, and services. The authors also point out that social norms is one of the key factors that cause these differences. Similar study of Ali et al. (2016) using Uganda data shows that the responsibility of child care is the main driver of the gap.
Apart from the labor market, there is gap of study performance between male and female student, in particular in fields such as mathematics and natural science. Halpem et al. (2005) argue that it has to do with social conditions and gender norms, while Guiso et al. (2008a) find a negative relationship between gender equality and gender gap in mathematic performance. The above discussion shows that both productivity gap in the labor market (either due to motherhood or unequal access to input and resources) and study performance can be drastically changed if gender norms are shifted to a more balanced side, and the source of productivity loss will be eliminated.
9
author shows that although women has done pretty well in accessing to higher education over time, they are still disadvantaged relative to men in terms of post-college outcomes. The study of Correll (2014) shows that gender norms have huge impact on what women think they can and should do at early stage and therefore determine abilities of individuals. Consequently, it affects self-assessment of women and can distort their decisions to pursue careers in math or natural science
The above discussion has shown how tremendous negative impact can unequal gender norms have on economic growth and development. In fact, some studies have attributed to the development stagnation to gender inequality (Diebolt and Faustine, 2013). Even though in developed countries much efforts to achieve gender equality have been spent and the situation has improved a lot, the misallocation of talent and abilities caused by unequal gender norms still result in huge negative effect. This thesis therefore fills the gap in the literature on how allocation efficiencies would be improved if gender norms become more balanced and women have wider choices on the profession they can take. While most studies about the impact of gender (in)equality on productivity focus on the second channel and the misallocation problem of the third channels, this thesis providesempirical evidence about the gain of efficiency that may come from any of the three channels if the misallocation problem is eased.
Trust, work norms, market attitudes, and attitudes towards democracy
10
The study of the relation between work norms and economic performance starts from Max Weber’s work, which emphasizes the effect of Protestant work ethic on economic development. In recent years, a number of empirical works has find channels through which work norms affect economic behaviors, such as Hamilton et al. (2003) and Stutzer and Lalive (2004).
The freedom of speech, thinking and publishing protected by democratic regimes encourage innovation and entrepreneurship, which is essential to economic growth and technology progress (Gorodnichenko and Roland, 2017). Finally, market attitudes reflect the degree people accept outcomes of competition under free market mechanism. People with pro-market attitudes consider the outcomes of free competition fair and just and are less likely to demand intervention from government for redistribution. Therefore, less resources are devoted to perform redistribution and less efficiency loss would occur, which is positive for economic development (Alesina and Angeletos, 2005).
III. Empirical Strategy
1. Identify the most important cultural dimension(s) for labor productivity
11
Consequently, in the first part of analysis my strategy is to regress log of labor productivity of sector s in region r of country c on the control variable and the five culture dimensions one by one, together with the country and the sector fixed effects, as Equation (1) shows:
𝑙𝑛(𝑙𝑎𝑏𝑜𝑟𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦)𝑐𝑟𝑠
= 𝑐 + 𝛽 ∗ 𝑐𝑢𝑙𝑡𝑢𝑟𝑎𝑙𝑛𝑜𝑟𝑚𝑠𝑖 + 𝐹𝐸𝑐 + 𝐹𝐸𝑠+ 𝜀𝑐𝑟𝑠 (1)
Given the relevance of the five cultural dimensions as discussed in Section II, I expect these five cultural norms to be statistically significant when each of them are included in the regression separately. In the next step, I will pool the five cultural dimensions together in one regression. The cultural dimension(s) which is/are the most important to labor productivity is defined as the cultural dimension(s) whose coefficient remain statistically significant when all five cultural dimensions are included in the regression.
2. Exploring the Mechanism behind the identified link
12
labor market, including women with high labor productivity, which increases the average level of labor productivity.
To test for this hypothesis, I develop the following model as shown by Equation (2):
𝑙𝑛(𝑙𝑎𝑏𝑜𝑟𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦)𝑐𝑟𝑠
= 𝑐 + 𝛼 ∗ 𝑔𝑒𝑛𝑑𝑒𝑟𝑛𝑜𝑟𝑚𝑠𝑐+ 𝛽 ∗ 𝑠𝑒𝑐𝑡𝑜𝑟𝑑𝑢𝑚𝑚𝑦𝑠
+ 𝛾 ∗ 𝑔𝑒𝑛𝑑𝑒𝑟𝑛𝑜𝑟𝑚𝑠𝑐 ∗ 𝑠𝑒𝑐𝑡𝑜𝑟𝑑𝑢𝑚𝑚𝑦𝑠 + 𝐹𝐸𝑐 + 𝐹𝐸𝑠+ 𝜀𝑐𝑟𝑠(2)
The model includes a constant term, the gender norms variable, a sector dummy, the interaction term between the gender norms variable and the sector dummy, the country-level fixed effects, and the sector-level fixed effects. I expect 𝛼 to be positive because according to the above discussion, more balance gender norms should encourage more competent women to join the labor market. But the focus of my analysis is the sign of 𝛾. If the hypothesis is correct, that more balanced gender norms increase labor productivity because the replacement effect is present, then 𝛾 should be positive. This is because for a given sector, if more balanced gender norms cause the entry of more competent women into the sector, thus increase labor productivity, then both equal gender norms and the sector dummy are positively related labor productivity. Consequently, 𝛾 should be positive.
IV. Data Sources
Dependent Variable
13
persons worked. The GVA data is the gross value added for a given sector in a given region. That is, for each given region, the GVA data for a complete set of ISIC economic activities are collected. The data of GVA, the implicit deflator, as well as numbers of persons employed for each sector in each NUTS2 region in the 28 European countries are collected from EUROSTAT. The sector specific data at the regional level allows me to apply sectoral fixed-effects model to exclude the influence of omitted sector factors on dependent variable. To capture the effect of cultural dimensions on long-term labor productivity, I use GVA and labor data of year 2007 to get rid of the shock of the financial crisis starting from year 2008. Table I provides information about sector codes and the corresponding sector each code represents in the following analysis.
Cultural Variables
Data for the cultural variables are taken from EVS. EVS is a cross-country research project that dedicates to collecting people’s values and attitudes by means of interviewing random respondents in European countries. The questions asked covers a variety of topics, including people’s attitudes towards politics, environment, gender, market, government, and religion, to name just a few. The surveys start as early as 1981. In the later years three more waves of interviews had been conducted (in year 1990, 1999, and 2008, respectively) and involved more countries. Following the approach of Beugelsdijk et al. (2017b), I use the data of wave 2008 for my analysis, which is the largest scale of survey conducted so far.
14
norms. Factor analysis and Cronbach alpha test are then performed to ensure that the chosen questions precisely capture the culture dimensions they are supposed to measure. Next, the responses of the interview at the individual level are recoded so that all responses take the value of either one or zero, and the index is construct in the way that the higher scores represent the higher degree in favor of economic development. Finally, the individual data are aggregated to the regional and then to the country level.
Control Variable
The control variable is average years of schooling at the regional level taken from Beugelsdijk et al. (2017a). Intuitively, it is rather straightforward to explain why degree of education should be included as the control variable. First of all, higher level of education by itself increases labor productivity. In addition, level of education is also closely related to other cultural dimension variables. For example, people with higher education have higher level of trust to others because education teaches people to cooperate. (Alesina and Giuliano, 2015) Higher education also make people to be more open-minded to question the status quo and shifts the society toward a secondary- or tertiary-sector based economy, which fosters more balanced gender norms. Moreover, education also has direct link to professionalism, which plays a key role shaping work norms, and make people hold a more liberal attitude towards market opening because higher educated people usually have higher level of skills and are less afraid of competition. Last but not least, it is widely known that education is crucial in maintaining the well-functioning of the democracy system (Glaeser et al., 2007).
15
Industrial Classification (ISIC Rev.4) system. However, both databases do not provide detailed data for each of the ISIC sector so the data obtained from the two sources needed to be further aggregated. Therefore, in this thesis the 21 sectors are further re-grouped into 8 sector groups for the following data-collecting reasons: Both the sectoral data of GVA from EUROSTAT and the female share employment share data from ILO do not provide data for each of the 21 sectors separately. Therefore, to make all data collected from different sources compatible, I have to combine some of the neighboring sectors to the extent that all sectoral data from various sources are categorized consistently such that running regression is possible. In the remaining part of the thesis, the 21 ISIC sectors will be re-grouped and presented as the following combination: Sector A (Agriculture Sector) as a group; Sector B-E (Mining, Manufacturing, Utilities, and Water Supply Sectors) as a group; Sector F (Construction Sector) as a group, Sector G-J (Wholesale and Retail Trade, Transportation and Storage, Accommodation and Food Services, and Information and Communication Sectors) as a group; Sector K (financial and Insurance Activities Sector) as a group; Sector L-N (Real Estate Activities, Professional, Scientific, and Technical Activities, and the Administrative and Support Services Sectors) as a group; Sector O-Q (Public Administration and Compulsory Social Security, Education, and Human Health and Social Work Sectors) as a group; Sector R-U (Arts, Entertainment, and Recreation, Other Services, Household Activities, and Extraterritorial Activities Sectors) as a group.
re-16
grouped for the use in regression, and what type each sector-group belongs to, which will be further illustrated in Figure 1 and the paragraph below.
17
Table 1 List of Sector Codes, the Corresponding Sector Names, the Re-Grouping Combination, and the types they belong to
Sector Code
Full Sector Name Re-Grouping used in Regression
Male or Female Dominated Sector
Sector Code
Full Sector Name Re-Grouping used in Regression
Male or Female Dominated Sector A Agriculture;
forestry and fishing
A Male-Dominated L Real estate activities L-N
Female-Dominated B Mining and
quarrying
B-E Male-Dominated M Professional, scientific and
technical activities
C Manufacturing N Administrative and support service activities
D Electricity, gas, steam and air conditioning supply
O Public administration and defence; compulsory social security O-Q Female-Dominated E Water supply; sewerage, waste management and remediation activities P Education
F Construction F Male-Dominated Q Human health and social work activities
G Wholesale and retail trade; repair
G-J Neutral R Arts, entertainment and
recreation
R-U
19 of motor vehicles
and motorcycles
H Transportation and storage
S Other service activities
I Accommodation and food service activities
T Activities of households as employers; undifferentiated goods- and services-producing activities of households for own use
J Information and communication
U Activities of extraterritorial organizations and bodies
K Financial and insurance activities
K Neutral
Figure 1 Average Female Employment Share by Sector of the 28 European Countries
Sources: International Labor Organization and Author’s own Calculation
V.
Baseline Results
Baseline Results 1
Column (1) to Column (5) of Table 2 show the results of equation (1) and Column (6) shows the result of pooling all five cultural norms together in one regression (Equation (2)). All regressions include the country- and the sector-fixed effects, and the standard errors are heteroscedasticity robust. From Column (1) to Colum (5) it is clear that except for gender norms, coefficients of the other four cultural dimensions are not significant even when each dimension is included as the sole cultural variable in the regression, which indicates that gender norms are perhaps the most relevant cultural dimension to labor productivity. Consequently, when pooling all cultural variables together in Column (6) we need to pay attention to see if the coefficient of the gender
32.4 33.1 8.1 45.2 44.7 64.7 57 31.2 31.1 7.9 43.3 56.3 46 65.5 60 A B - E F G - J K L - N O - Q R - U SH A R E O F F EMA LE E MPL O YME N T SECTOR CODE
21
norms variable remains significant.
Column (6) shows the result of pooling all cultural variables in one regression. As
Table 2 Baseline Result- Identify the Major Cultural Norm(s)
(1) (2) (3) (4) (5) (6) Avg. Yr. of Schooling 0.1255*** ( 0.0292) 0.1319*** ( 0.0317) 0.1325*** ( 0.0316) 0.1402*** ( 0.0271) 0.1319*** ( 0.1655) 0.1286*** ( 0.0255) Gender Norms 0.2159* ( 0.1190) 0.1996* ( 0.1038) Trust 0.1150 (0 .1385) .0819 ( 0.1171) Work Norms 0.1187 ( 0.0848) 0.1500* ( 0.0875) Market Attitudes 0.3208 ( 0.1959) 0.3365* ( 0.1842) Democracy Attitudes 0.1395 ( 0.1563) 0.0278 ( 0.1155) Constant 8.3511*** ( 0.3947) 8.3497*** ( 0.3815) 8.3338*** ( 0.4152) 8.1261*** ( 0.3392) 8.3102*** ( 0.3797) 8.0050 ( 0.3676) Observations 1699 1699 1699 1691 1699 1691 R-squared 0.2852 0.2739 0.2530 0.2675 0.2722 0.2873
Country FE YES YES YES YES YES YES
22
it shows, gender norms are still significant when all cultural variables are included. One stand deviation increase of the gender norms scores increase about 19.9% of labor productivity, which is quite a lot of difference.
If we only focused on Column (6), one may argue that the coefficients of work norms and market attitudes are also significant, so work norms and market attitudes should be considered factors at least as important as gender norms to labor productivity. Nevertheless, one should also note that the work norms and market attitudes are not significant in Column (3) and Column (4), when they are added to the regression as the sole cultural variables. This sends out two messages. Firstly, work norms and market attitudes are not significant by itself, but instead is caused by the addition of the other variables (trust and attitudes towards democracy). Secondly, this means that the order by which the five cultural dimension variables are added to the regressions can affect significance of their coefficients. However, to be considered as an important factor that affect labor productivity, a cultural dimension variable must not change its significance with the order by which it is added to the regression. That’s why it is necessary to combine the results from Column (1) to Column (6) to pin down the cultural dimension that is the most relevant to labor productivity.
Baseline Results 2
23
is supposed to be the focus of analysis in this part, we find that there is no clear and consistent pattern, either about the signs or about the significance of the coefficients. The significantly positive sign of the coefficient of the interaction term only appear in the Agriculture Sector, Mining Sector, Manufacturing Sector, Utilities Sector, and Water Supply Sector (Column (1) and (2)). In the Construction Sector (Column (3)), the Wholesale and Retail Trades Sector, the Transportation and Storage Sector, the Accommodation and Food Services Sector, the Information and Communication Sector (Column (4)), and the Financial and Insurance Activities Sector (Column (5), the estimated coefficient is statistically insignificant regardless of the signs, and in the Real Estate Sector, the Professional, Scientific and Technical Activities Sector, and the Administration and Support Activities Sector (Column (6), the Public Administration Sector, the Education Sector, the Human Health and Social Security Sectors (Column (7)), and the Arts, Entertainment and Recreation, the Other Services Sector, the Household Activities Sector, and the Extra Territorial Activities Sector (Column (8)) they are most of the time negative and significant. How do we explain these results?
24
Table 3 Baseline Result- Country and Sector Fixed Effects
Dependent Variable: Log Labor Productivity Measured by Persons
(1) (2) (3) (4) (5) (6) (7) (8)
Sector code A B-E F G-J K L-N O-Q R-U
Avg. yr. of Schooling 0.1270*** (0.02917) 0.1260*** (0.0292) 0.1253*** (0.0292) 0.1256*** (0.0292) 0.1256*** (0.0292) 0.1261*** (0.0292) 0.1258*** (0.0292) 0.1256*** (0.0292) Gender Norms -0.1400 (0.2111) 0.0938 (0.1372) 0.2615** (0.1114) 0.2415* (0.1285) 0.2365* (0.1319) 0.3572** (0.1442) 0.2767** (0.1342) 0.2422* (0.1186) Sector Dummy -2.5115*** (0.3580) 0.5169** (0.2334) 0.7115*** (0.2118) 0.7399*** (0.2020) 1.4030*** (0.2875) 2.2182*** (0.3784) 0.8019*** (0.1545) 0.3776 (0.3107) GenderNorms* SectorDummy 2.0329*** (0.5289) 0.6977* (0.3750) -0.2606 (0.2909) -0.2742 (0.2534) -0.2207 (0.4884) -1.5135** (0.7270) -0.5995*** (0.2141) -0.2812 (0.4756) Constant 9.9054*** (0.3456) 8.4143*** (0.3908) 8.3275*** (0.3980) 8.3355*** (0.4011) 8.3385*** (0.3954) 8.2649*** (0.4052) 8.3141*** (0.3981) 8.3351*** (0.3946) Observation 1699 1699 1699 1699 1699 1699 1699 1699 R-squared 0.2852 0.2830 0.2866 0.2845 0.2846 0.2851 0.2847 0.2844
Country FE YES YES YES YES YES YES YES YES
Sector FE YES YES YES YES YES YES YES YES
27
quotas imposed by the Swedish Social Democratic Party at the municipal level and find that municipalities where the quotas have the largest impact becomes more likely to select female leaders.
For this explanation to make sense, however, the only puzzle is Sector F, which is the most male-dominated of all sectors but we still don’t observe the replacement effect which is supposed to be indicated by a positive and significant coefficient of the interaction term. A possible explanation is that there are too few female employees in the Construction Sector (only less than 10%) for the replacement effect to be strong and clear. In addition, the Construction Sector is more physically demanding for women, which may explain why the share of female employment is so low. Nevertheless, increasing female employment should still increase labor productivity of the sector since the coefficient of gender norms is positive and significant.
28
across countries or regions, a finding in line with the study of Hsieh and Klenow (2009). To sum up, the findings in the male-dominated sectors and female-dominated sectors together show that people’s gender norms attitudes can have economic consequences by causing misallocation of human resources.
VI. Robustness Checks
In this section results of several alternative settings are presented to ensure that the baseline results are robust and consistent across different settings. First of all, I consider alternative models with the country×sector fixed effects and the region fixed effects. Next, I also test whether the baseline findings of the presence (and the lack of) the replacement effect correspond to the renowned development gaps between the north-versus-the-south and the-new-versus-the-old EU member states. Finally, I replace sector dummy with female employment share, a direct measure of females’ involvement in the job market, to see if the same conclusion still holds.
Alternative fixed effects models
The baseline model only controls for the country fixed effects and the sector fixed effects. However, this setting may fail to control for some omitted characteristics which are country-sector specific. In addition, since data at the regional level is used for analysis in this study, it is possible that some unobserved factors at the regional level also account for results observed in the baseline model.
male-29
Table 4 Robustness Check- Country×Sector Fixed Effects
Dependent Variable: Log Labor Productivity Measured by Persons
(1) (2) (3) (4) (5) (6) (7) (8)
Sector A B-E F G-J K L-N O-Q R-U
Avg. yr. of Schooling 0.1261*** (0.0180) 0.1251*** (0.0179) 0.1253*** (0.0179) 0.1252*** (0.0180) 0.1253*** (0.0179) 0.1259*** (0.0178) 0.1256*** (0.0179) 0.1255*** (0.0179) Gender Norms 0.0619 (0.1147) 0.3033** (0.1274) 0.2681* (0.1384) 0.1711 (0.1393) 0.1652 (0.1372) 0.2963** (0.1451) 0.2412* (0.1435) 0.2188 (0.1408) GenderNorms *SectorDummy 0.8795*** (0.3258) -0.4990 (0.3534) -0.2983 (0.3429) 0.4794 (0.2997) 0.5436 (0.4006) -0.8616*** (0.2562) -0.2494 (0.2158) -0.0311 (0.2958) Constant 9.0124*** (0.2224) 9.0004*** (0.2232) 9.0022*** (0.2240) 9.0025*** (0.2229) 9.002*** (0.2230) 9.009*** (0.2226) 9.0064*** (0.2241) 9.0050*** (0.2247) Observation 1699 1699 1699 1699 1699 1699 1699 1699 R-squared 0.0411 0.0020 0.004 0.0005 0.0376 0.0315 0.0059 0.0037 Country* Sector FE
YES YES YES YES YES YES YES YES
31
Table 5 Robustness Check- Region Fixed Effects
Dependent Variable: Log Labor Productivity Measured by Persons
(1) (2) (3) (4) (5) (6) (7) (8)
Sector A B-E F G-J K L-N O-Q. R-U
Sector Dummy -2.5616*** (0.1667) 0.4802*** (0.1110) 0.6845*** (0.0838) 0.7634*** (0.0834) 1.4262*** (0.1368) 2.2474*** (0.1486) 0.8187*** (0.0744) 0.4011*** (0.1102) Gender Norms *Sector Dummy 2.1222*** (0.2698) 0.7630*** (0.1785) -0.2124* (0.1205) -0.3162** (0.1241) -0.2623 (0.2266) -1.5656*** (0.2466) -0.6297*** (0.1118) -0.3233* (0.1759) Constant 11.2407*** (0.0262) 9.8698*** (0.0338) 9.8692*** (0.0335) 9.8694*** (0.0334) 9.8694*** (0.0334) 9.8694*** (0.0330) 9.8695*** (0.0334) 9.8694*** (0.0334) Observation 1699 1699 1699 1699 1699 1699 1699 1699 R-squared 0.3317 0.3035 0.2880 0.2866 0.2874 0.2729 0.2829 0.2865
Sector FE YES YES YES YES YES YES YES YES
Region FE YES YES YES YES YES YES YES YES
In the next test, I control for another kind of fixed effects-- regional fixed effects and the results is shown in Table 5. Because the regional fixed-effects model absorbs most of the omitted variables at the regional level, the results on the average years of education and gender norms are not available (and not reported) here because all regional-level variation are absorbed. Consequently, the effect of the non-regional level variables becomes more prominent, including the interaction term. From Table 5 it can be seen that the interaction terms in almost all sectors yield results that is consistent with the groups they belong to, which shows that the baseline finding is also robust in the regional fixed-effects setting.
Gender norms Attitudes and the North-South/ East-West Difference
33
and b) the EU15 countries (which include Austria, Belgium, Denmark, Greece,
Germany, Finland, France, Ireland, Italy, Luxemburg, the Netherlands, Portugal, Spain, Sweden, and United Kingdom) versus the new member states (which include Czech Republic, Croatia, Cyprus, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovakia, Slovenia, Poland, and Romania). I then create dummies for both groups and interact the dummies with the gender norms. Table 6 reports the results of the regression. As the table shows, while the interaction term of both the northern and southern countries are insignificant (Column (2), and (4)), the interaction term of the EU15 nations are significantly negative (Column (1) and (3)), which is the typical pattern of the female-dominated sectors that we already observed from Table 3 to Table 5. Taken together, this means that the north-south divide does not differ a lot in terms of cross-sector distribution of female employment share; instead, the adjustment of female employment share is a matter of old-and-new EU states (east-west) difference.
Direct measure of Female Share of Employment
34
with the concern that results so far discussed may be caused by a particular variable (namely sector dummies only). At least the results in Column (3) and Column (4) of table 7 confirm that the mechanism of the replacement effect is robust even when sector dummy is replaced with female labor share.
Dependent Variable: Log Labor Productivity Measured by Persons
(1) (2) (3) (4)
Region EU15 Northern EU15 Northern Avg. yr. of Schooling 0.1191*** (0.0281) 0.1271*** ( 0.0303) 0.1191*** (0.0176) 0.1271*** ( 0.0183) Gender Norms 1.3015** (0.5684) 0.1141 ( 0.1713) 1.3015*** (0.3177) 0.1141 ( 0.1372) Gender Norms *Sector Dummy -1.2240** (0.5751) 0.1710 ( 0.3213) -1.2240*** (0.3543) 0.1710 ( 0.2733) Constant 8.3697*** (0.3427) 8.3227*** ( 0.4132) 9.0234*** (0.2127) 8.9764*** ( 0.2420) Observation 1699 1699 1699 1699 R-squared 0.0281 0.2922 0.2424 0.0087
Country FE YES YES NO NO
Sector FE YES YES NO NO
Country× Sector FE
NO NO YES YES
Robust standard errors in parentheses *** p<0.01, ** p<0.05, *p<0.1
Table 7 Robustness Check- Female Employment Share by Sector
(1) (2)* (3) (4)* (5) (6)*
Dependent Variable: Log Labor Productivity Measured by Persons Avg. yr. of Schooling 0.1259*** (0.0290) 0.1262*** (0.0292) N/A (omitted) N/A (omitted) 0.1251*** (0.0192) 0.1252*** (0.0179) Gender Norms 0.7294* (0.3880) 0.5459 (0.3280) N/A (omitted) N/A (omitted) 0.2834 (0.3480) 0.0889 (0.3216) Female share of employment -0.3962 (0.7147) -1.9828*** (0.6765) -0.2607 (0.2841) -1.8520*** (0.2694) N/A (omitted) N/A (omitted) GenderNorms *FemaleShare -1.4268 (0.9396) -0.8917 (0.7756) -1.653*** (0.3500) -1.1222*** (0.2929) -0.1778 (0.6935) 0.3434 (0.6504) Constant 8.4462*** (0.4421) 8.9391*** (0.4612) 10.2537*** (0.0730) 10.6480*** (0.0687) 9.0695*** (0.2403) 9.0024*** (0.2235) Observation 1505 1699 1505 1699 1505 1699 R-squared 0.2963 0.3596 0.2350 0.3450 0.0003 0.0006
Country FE YES YES NO NO NO NO
Sector FE YES YES YES YES NO NO
Region FE NO NO YES YES NO NO
Country×Sector FE
NO NO NO NO YES YES
VII. Conclusion
In this thesis I examine the relationship between gender norms and labor
productivity, and try to figure out the mechanism behind the positive relationship between more balanced gender norms and increased labor productivity level. The empirical strategy for my analysis consists of two parts. In the first part, I regress labor productivity on several cultural dimensions’ variables and find that gender norms are indeed the most important factor that contributes to the increase of labor productivity. In the second part, I seek to explain why this is the case. I argue that more balanced gender norms encourage higher-productivity women to join the labor market. Therefore, the increase of higher-productivity women working as employees boosts the total labor productivity.
37
and Technical Activities Sector, and the Administrative and Support Services Sector (Sector L-N), the Public Administration and Compulsory Social Security Sector, the Education Sector, and the Human Health and Social Work Sector (Sector O-Q), and the Arts, Entertainment, and Recreation, the Other Services, the Household Activities, and the Extraterritorial Activities Sector (Sector R-U), whose distribution of female share of employment mostly fall in the range exceeding 50 percent. I call these sector female-dominated sectors.
In my second part of empirical analysis, I create a sector dummy for each sector and interact the sector dummy with the gender norms variable. If my hypothesis that more balanced gender norms encourage more higher-productivity women to the job market (thus increase labor productivity by replacing lower-performing men) is true, then a positive estimated coefficient of the interaction term should be observed for a given sector.
The baseline results show that the sign and statistical significance of the interaction term to a large extend depends on which type a sector belongs to. For the male-dominated sectors, the coefficients of the interaction term are usually positive and significant. For the neutral sectors, coefficients of the interaction term are usually not significant. For the female-dominated sectors, coefficients of the interaction term are usually negatively significant, while coefficients of the gender norms variable are usually positive and significant, which means that even though the replacement effect is negative in the female dominated sectors, more balanced gender norms still increases labor productivity by encouraging more productive women to work in the sectors.
male-38
dominated sectors. These higher-productivity women replace poorer-performing men and increase labor productivity of the sectors, which explain the positive and significant coefficients of the interaction term in these sectors. In the neutral sectors, such replacement effect is not obvious since the share of male and female employers is roughly equal, which is indicated by insignificant coefficients of the interaction term. Finally, in the female-dominated sectors, coefficients of the interaction term are negatively significant because the replacement effect is negative. The above findings not only show that people’s gender attitudes can have economic consequences, but also demonstrate that cultural norms can lead to misallocation and inefficiencies in resources use, which cannot be observed from human and physical capital inputs but may reflect in the difference of TFP. Values and beliefs can therefore explain TFP gaps across countries and regions.
39
effect, I expect a negative sign of the interaction term. The results confirm my hypothesis.
The findings of the thesis have important policy implications. First of all, it shows that cultural norms does affect economic outcomes by affecting allocation efficiency of (human) resources. Focusing on the study of gender norms, this thesis shows that more balanced gender norms can yield sizeable productivity. This is because in sectors where the females are minority, the entry of additional higher-productivity females may replace poor-performing males at the lower end and increase the total sector productivity. Consequently, promoting female employment in these male-dominated sectors can yield productivity dividends. Secondly, these efficiency gains can be achieved by a change of gender norms attitudes. When gender norms become more balanced, more female would be willing to work in male-dominated sector based on own productivity. Therefore, the promotion of more balanced gender norms can serve as the catalyst to achieve the goal of productivity gains by mean of human resources re-allocation.
References
Aghion, Philippe, Yann Algan, Pierre Cahuc, and Andrei Shleifer. 2010. “Regulation and Distrust.” The Quarterly Journal of Economics, 125 (3): 1015–1049.
Akerlof, George A., and Rachel E.Kranton. 2000. “Economics and Identity.” The
Quarterly Journal of Economics, 115(3): 715-753.
Alesina, Alberto, and George-Marios Angeletos. 2005. "Fairness and Redistribution."
American Economic Review, 95 (4): 960-980.
40
Economic Literature, 53 (4): 898-944.
Alesina, Alberto, Paola Giuliano, and Nathan Nunn. 2013. “On the Origins of Gender Roles: Women and the Plough.” Quarterly Journal of Economics, 128 (2): 469-530.
Ali, Daniel, Derick Bowen, Klaus Deininger, and Marguerite Duponchel. 2016. “Investigating the Gender Gap in Agricultural Productivity: Evidence from Uganda.”
World Development, 87: 152-170.
Baltrunaite, Audinga, Piera Bello, Alessandra Casarico, and Paola Profeta. 2014. “Gender quotas and the quality of politicians.” Journal of Public Economics, 118: 62-74.
Besley, Timothy, Olle Folke, Torsten Persson, and Johanna Rickne. 2017. “Gender Quotas and the Crisis of the Mediocre Man: Theory and Evidence from Sweden.”
American Economic Review, 107 (8): 2204-2242.
Beugelsdijk, Sjoerd, Mariko J. Klasing, and Petros Milionis. 2017a. “Regional economic development in Europe: the role of total factor productivity.” Regional
Studies, 52 (4): 461-476.
Beugelsdijk, Sjoerd, Mariko J. Klasing, and Petros Milionis. 2017b. “Value Diversity and Regional Economic Development.” The Scandinavian Journal of Economics, forthcoming.
Boschini, A.D. 2002. “Gender-Specific Educational Choices, Earning Gaps and Technological Change.” Dissertations in Economics 2002:3, Stockholm University.
Boschini, A.D. 2006. “The Impact of Gender Stereotypes on Economic Growth.” GEM-IWG Working Paper 06-16.
Cooray, Arusha and Niklas Potrafke. 2011. “Gender Inequality in Education: Political Institutions or Culture and Religion?” European Journal of Political Economy, 27 (2): 268–280.
41
Croppenstedt, Andre, Markus Goldstein, and Nina Rosas. 2013. “Gender and Agriculture: Inefficiencies, Segregation, and Low Productivity Traps.” The World
Bank Research Observer, 28 (1): 79–109.
Diebolt, Claude, and Faustine Perrin. 2013. “From Stagnation to Sustained Growth: The Role of Female Empowerment.” The American Economic Review, 103 (3), PAPERS AND PROCEEDINGS OF THE One Hundred Twenty-Fifth Annual Meeting OF THE AMERICAN ECONOMIC ASSOCIATION: 545-549.
Fernandez, Raquel. 2007. ‘‘Women, Work, and Culture.’’ Journal of the European
Economic Association, 5 (2-3): 305–332.
Fernandez, Raquel, and Alessandra Fogli. 2009. ‘‘Culture: An Empirical Investigation of Beliefs, Work, and Fertility.’’ American Economic Journal: Macroeconomics, 1 (1): 146–177.
Fortin, Nicole M. 2005. ‘‘Gender Role Attitudes and the Labour Market Outcomes of Women Across OECD Countries.’’ Oxford Review of Economic Policy, 21 (3): 416– 438.
Galor, Oded and David N. Weil. 1996. “The Gender Gap, Fertility, and Growth.” The
American Economic Review, 86 (3): 374-387.
Glaeser, Edward L., Giacomo A. M. Ponzetto, and Andrei Shleifer. 2007. “Why does democracy need education?” Journal of Economic Growth, 12: 77–99.
Gorodnichenko, Yuriy and Gerard Roland. 2017. “Culture, Institutions, and the Wealth of Nations.” Review of Economics and Statistics, 99 (3): 402-416.
Guiso, L., F. Monte, P. Sapienza, and L. Zingales. 2008. “Culture, Gender, and Math.”
Science, 320(5880): 1164-1165.
Guiso, L., P. Sapienza, and L. Zingales. 2008. “Social Capital as Good Culture,”
Journal of the European Economic Association, 6 (2–3): 295–320.
42
Halpem, D.; J. Wai, and A. Saw. 2005. In Gender Differences in Mathematics, edited by A. M. Gallagher and J. C. Kaufman, 48-72. New York: Cambridge University Press.
Hsieh, Chang-Tai and Peter J. Klenow. 2009. “Misallocation and Manufacturing TFP in China and India.” The Quarterly Journal of Economics, 124 (4): 1403–1448.
Jacobs, Jerry A. 1996. “Gender Inequality and Higher Education.” Annual Review of
Sociology, 22: 153-185.
Klasen, Stephan. 1999. “Does Gender Inequality Reduce Growth and Development? Evidence from Cross-Country Regressions.” World Bank Policy Research Report on Gender and Development Working Paper Series No. 7.
Klasen, Stephan. 2002. “Low Schooling for Girls, Slower Growth for All? Cross‐ Country Evidence on the Effect of Gender Inequality in Education on Economic Development.” The World Bank Economic Review, 16 (3): 345–373.
Klasen Stephan and Francesca Lamanna. 2009. “The Impact of Gender Inequality in Education and Employment on Economic Growth: New Evidence for a Panel of Countries.” Feminist Economics, 15 (3): 91-132.
Knack, Stephen and Philip Keefer. 1997. “Does Social Capital Have an Economic Payoff? A Cross-Country Investigation.” The Quarterly Journal of Economics, 112 (4): 1251–1288.
Knowles, Stephen, Paula K. Lorgelly, and Dorian P. Owen. 2002. “Are Educational Gender Gaps a Brake on Economic Development? Some Cross-Country Empirical Evidence.” Oxford Economic Papers, 54 (1): 118-149.
Liang, Li-Ling. 2017. “Gender Norms and Educational Attainment: A Comparative Study of Countries and Regions”, LPR paper, University of Groningen
O’Brien, Diana Z. and Johanna Rickne. 2016. “Gender Quotas and Women's Political Leadership.” American Political Science Review, 110 (1) : 112-126.
43
Jersey: Princeton University Press.
Stutzer, Alois and Rafael Lalive. 2004. “The Role of Social Work Norms in Job Searching and Subjective Well-Being.” Journal of the European Economic
Association, 2 (4): 696–719.
Appendix
Table 1 List of European Countries Included in the EVS Data
Country Name Country Name
Austria Italy
Belgium Latvia
Bulgaria Lithuania
Croatia Luxemburg
Cyprus Malta
Czech Republic Netherlands
Denmark Poland Estonia Portugal Finland Romania France Slovakia Germany Slovenia Greece Spain Hungary Sweden
