Abstract

In most countries, rates of female labour force participation have continually increased over the last 50 years. Yet, there is a marked gender gap in labour force participation rates. This stylised fact has important implications for economies, as there is a positive relationship between development and female participation rates (Dufló, 2012). Accordingly, governments have developed a battery of strategies based on classic human capital theory in order to enhance women's participation in the labour market and promote economic development (Mincer 1962, Behrman and Wolfe 1984, Killingsworth and Heckman 1986, Blundell et al. 1987). In particular, the expansion of higher education, access to contraceptive methods, childcare provision, training programmes and flexible working-hours are among the factors that have been implemented to increase female labour force participation. More recently, a scarce but growing body of the economic literature has evidenced the existence of social factors that are not captured by observed human capital measures but can increase rates of women's participation in the labour market even further. Effectively, social interactions among women may create strategic spillovers and amplify the effect of changes in individual incentives through a social multiplier (Manski 1993, Durlauf 2004). These studies recognise that women are socially situated within a social network that could modify the individual labour market behaviour (Mota et al. 2016, Maurin and Moschion 2009, Nicoletti et al. 2016). This direct interdependency among individuals' decisions, which is not mediated by markets or contracts, is called social interactions (Brock and Durlauf, 2001b). This paper contributes to this line of research and provides empirical evidence of the magnitude of social effects at neighbourhood level on female labour force participation (FLP) in Chile. In particular, the aim of this paper is to analyse how social interactions among low-skilled women affect their labour market behaviour, contemporaneously and in subsequent periods. I study neighbourhoods, as they constitute a natural form of social interactions. Peer effects at neighbourhood level can be explained by three main factors. First, psychological factors, such as inspiration or identity, would induce women to participate in work like other women (Akerlof and Kranton, 2000). Second, the existence of interdependences in the constraints that women face may affect the costs of choosing to participate in the labour market when other women are working. And third, an information transmission mechanism could affect the perceptions that women have regarding the opportunities in the labour market and advantages of work (Durlauf, 2004). This research analyses the case of Chile, a middle-income country that exhibits low rates of female labour force participation, despite the good indexes of education and growth. According to the Chilean Longitudinal Social Study (ELSOC), in 2016 only 47.2% of women with less than tertiary education were participating in the labour force. This figure is 49.8% in 2019. Despite this increment, the rate remains low compared to the average rate in the OECD (67%) or Latin American countries (52%). I follow the literature of discrete choice with social interactions (Brock and Durlauf, 2001a) and extend a standard model of female labour force participation (Blundell and Powell, 2004) by including in the utility function of a given woman the choices of other women within the neighbourhood. I focus on the extensive margin of female labour supply, as this is where observed adjustments may be more important due to the rigidity of the Chilean labour market and the prevalence of fixed working-hours contracts. Social interactions are modelled as proportional spillovers. Positive endogenous social effects indicate strategic complementarity. This implies that women, when deciding whether to participate or not, are influenced positively by women who participate and live in the same neighbourhood. In order to estimate the model, I construct a unique dataset by merging the average socioeconomic characteristics of women and labour outcomes at the census zone level from the Chilean Census 2017 to the surveyed women in the four waves of ELSOC, from 2016 to 2019. I study census zones, which are a very good proxy of the individuals’ neighbourhood, to quantify the causal effect of female neighbours' behaviour on low-skilled women’s individual decisions. The challenge inherent in identifying endogenous social effects is conveyed by Manski (1993). First, a reflection problem arises due to simultaneous movements in outcomes among neighbours. Second, unobserved shocks and the institutional environment that affects the entire neighbourhood could lead to correlations in unobserved attributes (Topa, 2011). Third, individuals may sort into different neighbourhoods on the basis of their neighbours' characteristics or because they have similar preferences (Topa, 2001). As a consequence, serious bias in the estimation of endogenous social effects can arise (Weinberg et al. 2004, Ludwig et al. 2001, Bayer et al. 2008). In order to address the reflection problem, the richness of the dataset allows me to implement a novel approach that is at the frontier of empirical studies of social interactions. Effectively, in order to identify the causal effect, I run two exercises. Firstly, I exploit the geographical location of households and use two instruments for the FLP rate of closest neighbours: i) the ratio between children younger than 6 years old and women living in the census zone. I argue that neighbours’ young children are strongly correlated with neighbours’ participation, but not with the determinants of individual participation (Maurin and Moshion, 2009); ii) the FLP rate of distant neighbours, who live in adjacent census zones, not the own zone. The key identification assumption is that women who reside in these distant zones do not affect the outcome of a given woman directly; they only affect her choices through her closest neighbours (Bramoullé et al. 2009). Secondly, I exploit the longitudinal dimension of the data in order to analyse the impact of the rate of FLP in on period on the individual decision of low-skilled woman in subsequent periods. This exercise allows me to control for correlated unobservable variables by adding individual and neighbourhood fixed effects. The results indicate that once I correct by endogeneity, there is no contemporaneous social effect of neighbours’ participation on the individual decision of participation in the labour market for low-skilled woman. Nevertheless, when I exploit the longitudinal dimension of the data, results suggest that the previous rate of FLP in the neighbourhood affect the individual decision of low-skilled women in subsequent periods. Effectively, a 10 percentage point increase in labour market participation among neighbours generates an increase of about 9 percentage points in a low-skilled woman’s probability of participation in the labour market. This is the first paper that quantifies the elasticity of own labour market participation to neighbours’ participation of women in a Latin American country, and it is relevant as the existence of this effect implies a social multiplier that can be used for policy to expand the effect of small changes in incentives or resources.