Exchange rate fluctuations and immigrants’ labour market outcomes: New evidence from Australian household panel data Ha Trong Nguyen *
Alan Duncan
Bankwest Curtin Economics Centre, Curtin University
Bankwest Curtin Economics Centre, Curtin University
Abstract: In this paper, we exploit plausibly exogenous changes in exchange rates across home countries over time and panel data to identify the causal impact of exchange rate fluctuations on Australian immigrants’ labour market outcomes. We present new and robust evidence that, unlike immigrants in the US, those in Australia as a whole do not reduce their yearly labour market outcomes when the local currency appreciates. While female immigrants don’t adjust their labour activities, male immigrants reduce their weekly labour supply and hence earn less when the Australian dollar appreciates. This work also highlights the importance of controlling for individual heterogeneity as well as the gender when modelling the labour market behaviour of immigrants.
Key words: exchange rate, labour supply, immigrants, Australia. JEL classification: F31, J22, J61.
*
Acknowledgements: We would like to acknowledge the comments made by Nina Pavcnik, the Co-Editor, and the anonymous reviewers for Journal of International Economics. We thank Hielke Buddelmeyer and participants at the 25th Australian Labour Market Workshop for their comments and suggestions. We gratefully acknowledge research assistance from Christian Duplock and Huong Le. This paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Project was initiated and is funded by the Australian Government Department of Social Services (DSS) and is managed by the Melbourne Institute of Applied Economic and Social Research (Melbourne Institute). The findings and views reported in this paper, however, are those of the authors and should not be attributed to either DSS or the Melbourne Institute. Corresponding author: Bankwest Curtin Economics Centre | Curtin Business School | Curtin University | Tel:+61 8 9266 5711 | Fax:+61 8 9266 2373 | Postal: Bankwest Curtin Economics Centre, GPO Box U1987, Perth WA 6845, Australia | Email:
[email protected].
1.
Introduction
International migration plays an important role for both home and host countries. It is therefore important to understand the factors influencing how immigrants fare in host countries. This paper contributes to the literature of international migration by examining the impact of exchange rate fluctuations on labour market outcomes of immigrants. The labour market outcomes of immigrants are of particular interest as they themselves are one of the major determinants of economic development of the host countries. Labour market outcomes of immigrants also have direct and important impacts on the development of their home countries, for example by international money transfer or by investment channels. Exchange rates may have some impact on the labour market outcomes of immigrants because they affect the purchasing power of income earned in the host country. In theory, a relative appreciation of the host country’s currency can have an ambiguous impact on the labour supply and subsequently the income of immigrants who want to make monetary transfers to their home countries or who at some point in the future may return to their home countries and bring with their labour income earned in the host country. As an example, in order to make the same amount of home country currency transfer, immigrants may need to work less because the host country’s currency appreciation could lower the price of transfer. Alternatively, income impact could cause the immigrants to work more while they are still in the host country. Therefore, the direction and magnitude of the impact of exchange rate on the labour supply is an empirical issue. Nekoei (2013) is the first study to provide evidence of the impact of exchange rate volatility on the labour supply of immigrants to the United States of America (US). His approach uses a repeated sample of cross-sections of immigrants drawn from the Current Population Survey (CPS) over sixteen years to capture exchange rate impacts, with identification achieved by treating home country exchange rate movements as exogenous price shocks that affect the average labour market outcomes of immigrants from that country. Nekoei (2013) finds that US immigrants worked fewer hours and hence earn less in response to an appreciation of the US dollar (USD). He also finds that responses to exchange rate volatility are largely similar for males and females. One problem with this approach is that immigrant groups are in continuous change. People enter and exit the host country. This alters the composition of immigrants over time in the host country, with the result that changes in migrants’ labour market outcomes that are presumed to be driven by exchange rate movements at the home
1
country level may instead be attributable to changes in the composition and unobservable attributes of immigrants from that home country group. This paper extends the approach of Nekoei (2013) to provide new empirical evidence from Australia of the relationship between exchange rate movements and immigrants’ labour market outcomes. Specifically, we employ a rich source of panel data that tracks the labour market behaviour of a sample of migrants to Australia over time. This provides us with an opportunity to control for individual time-invariant characteristics that may be correlated with both the probability of returning to the home country (due to exchange rate fluctuations) and to the work decisions of immigrants in the host country. This paper makes two important contributions to the literature. First, it is currently the only study to analyse the impact of exchange rate fluctuations on the labour market outcomes of immigrants to Australia. Like the US, Australia is a country with a high immigrant population. However, previous comparative studies that use both US and Australian data have shown that the composition and economic behaviour of immigrants from the two countries are not the same (Antecol et al., 2003; Antecol et al., 2006; Chiswick et al., 2008). It is therefore necessary to study whether the previous US findings about the labour market impact of exchange rate fluctuations still hold for Australia. Second, and more importantly, this paper uses panel data to control for unobserved individual heterogeneity when capturing exchange rate effects on immigrants’ labour market outcomes. Previous studies have demonstrated the importance of controlling for unobserved individual heterogeneity in modelling labour market outcomes of immigrants in Australia (Cobb-Clark et al., 2012; Breunig et al., 2013), Canada (Hum and Simpson, 2004), Germany (Fertig and Schurer, 2007) and the US (Borjas, 1985; Hu, 2000; Duleep and Dowhan, 2002; Dustmann and Fabbri, 2005; Lubotsky, 2007; Abramitzky et al., 2014). This paper thus provides an effective test of the robustness of previous study in which findings are based on cross-sectional data, and which therefore fails to account for unobserved individual fixed effects. Using twelve years of the Household Income and Labour Dynamics in Australia (HILDA) panel and exchange rates for 65 countries of origin, we find that immigrants do reduce their labour market activities in response to an appreciation of the Australian dollar (AUD). However, we also find that the relationships between exchange rate movements and immigrants’ labour market outcomes are often overstated when individual heterogeneity is not accounted for, or when data on the labour market outcomes of male and female immigrants are pooled. Results from models that control for individual fixed effects and 2
distinguish labour market outcomes by gender suggest that only male immigrants reduce their weekly labour supply when the AUD appreciates. The rest of the paper is structured as follows. Section 2 briefly reviews related literature. Section 3 describes the data and Section 4 explains the empirical modelling approach. Section 5 presents empirical results, while Section 6 presents results from a range of robustness tests. Section 7 reports heterogeneous exchange rate impacts among immigrants. Section 8 additionally discusses the results before Section 9 concludes. 2.
Literature review
This paper is related to two strands of literature. The first and developing body of literature explores the impact of macroeconomic conditions (including exchange rate fluctuations) on various aspects of immigrant behaviour. For example, a number of studies have found that a relative depreciation in the home country currency reduces the probability of return for immigrants (Yang, 2006; Abarcar, 2013) or increases the probability of emigration (Gordon and Spilimbergo, 1999). In addition, Yang (2008) finds that a depreciation of the Philippine peso against the host country’s currency leads to increases in household remittances from overseas, a finding which is consistent with an earlier finding by Faini (Faini, 1994)for foreign workers in Germany. 1 As mentioned above, Nekoei (2013) also provides the first evidence on the impact of exchange rate on labour market outcomes of US immigrants. This literature also examines the impact of other macroeconomic conditions (either at home or in host countries) on migrants’ behaviours. For instance, McKenzie et al. (2014) use a panel dataset of migrants originating from the Philippines to provide evidence that migrant flows are responsive to GDP shocks in destination countries. Studies have provided mixed evidence regarding the impact of home countries’ macroeconomic conditions on immigrants’ wellbeing: while Akay et al. (2016) find that the subjective well-being of German migrants decreases with improvements in their home countries’ macroeconomics conditions, Nguyen and Duncan (2015) demonstrate that Australian immigrants feel happier when their home countries’ macroeconomic conditions improve. This paper also contributes to a second strand of literature that examines the impact of remuneration on the labour supply. This literature so far has not come to an empirical consensus yet. For example, some studies (Camerer et al., 1997; Chou, 2002; Crawford and 1
There is also a quite large body of literature examining the impact of exchange rates from the point of view of firms or industries (Campa and Goldberg, 2001; Klein et al., 2003; Nucci and Pozzolo, 2010; Park et al., 2010; Kandilov and Leblebicioğlu, 2011; Ekholm et al., 2012; Héricourt and Poncet, 2013; Amiti et al., 2014).
3
Meng, 2011; Chang and Gross, 2014) find wage elasticities to be negative, consistent with the prediction of the target earning theory of labour supply (Camerer et al., 1997). In contrast, other studies (Gerald S. Oettinger, 1999; Fehr and Goette, 2007; Stafford, 2015) find wage elasticities to be positive, and more in support of the neo-classical theory of labour supply. 3.
Data and sample
3.1.
Data
We use data from several sources for this study. The first data source comes from Household Income and Labour Dynamics in Australia (HILDA) survey. HILDA is an annual nationally representative longitudinal survey of private households in Australia. In addition, HILDA contains rich information at the individual or household level, including data on sociodemographic variables, incomes, and labour market conditions. We use the first twelve waves of HILDA data for this analysis, covering the period from 2001 to 2012. Historical data on daily exchange rates are taken from the Oanda website. Data for a number of macroeconomic variables used in our research are sourced from the World Bank’s World Development Indicators (WDI) database. Additional macroeconomic data for Taiwan are sourced from the Taiwanese statistics office, as data for Taiwan are not available from the World Bank’s database. 2 3.2.
Exchange rates
Similar to Nekoei (2013), the purpose of this paper is to estimate the causal impact of real exchange rate movements on immigrants’ labour market outcomes. The real exchange rate between Australia and home country c is defined as 𝑒𝑒𝑐𝑐 = 𝐸𝐸𝑐𝑐 x(𝑃𝑃𝐴𝐴𝐴𝐴𝐴𝐴 /𝑃𝑃𝑐𝑐 ), where 𝐸𝐸𝑐𝑐 is the yearly nominal exchange rate for home country c (defined further below), and 𝑃𝑃𝑐𝑐 and 𝑃𝑃𝐴𝐴𝐴𝐴𝐴𝐴
are the yearly consumer price indices for country 𝑐𝑐 and Australia respectively. Exchange
rates are specified as the number of foreign currency units per AUD (for example 20,000
Vietnam Dong (VND)/AUD). 3 Therefore an increase in the exchange rate 𝑒𝑒𝑐𝑐 represents an appreciation of the AUD. For each country and in each year, we construct the yearly nominal
exchange rate 𝐸𝐸𝑐𝑐 as the average of daily exchange rates over the calendar year. Daily exchange rates are derived from the mid-point between the "buy" and "sell" rates from global currency markets. These nominal exchange rates are then used in conjunction with yearly CPI to construct yearly real exchange rates for use in our empirical modelling. 2
See http://data.worldbank.org/indicator. For Taiwan, the data is from National Statistics: http://stat.gov.tw/. We do not use an alternative exchange rate measure (i.e. units of AUD per unit of home currency) because for some currencies (such as Indonesian Rupiah or Vietnamese Dong), such measure results in a loss of precision. 3
4
3.3.
Sample
In our analysis, immigrants are those who were born outside Australia. We restrict the empirical sample to countries with enough observations and to countries with exchange rate data available in any year. 4 We further restrict the sample to individuals of working ages, between 16 and 64. 5 We also exclude individuals with missing information on any variable used in our empirical model. These sample restrictions result in a sample of 25,762 individual-year observations from 4,789 unique individuals obtained over twelve years of data and immigrants from 66 countries (See Table A2 in the online appendix for details). Immigrants in Australia are from a wide variety of countries. Table A2 displays the distribution of countries of birth of Australian immigrants a majority of them come from the United Kingdom (28%), New Zealand (12%), the Philippines (5%), Vietnam (4%), India (3%), South Africa (3%), China (3%), Germany (3%), Italy (2%), Netherlands (2%), and US (2%). The geographical diversity of Australian immigrants provides us with a considerable source of exchange rate volatility during the study period. This is illustrated in Figure 1 which presents the daily nominal exchange rates for selected major home countries of Australian immigrants. Figure 1 also shows, with the exception of the 2008-2009 recession when the AUD depreciated significantly, most immigrants enjoyed favourable exchange rate changes from 2001 to 2012. This is also confirmed in Table 1 which shows that, on average, the nominal AUD appreciated by 4.1% annually from 2001 to 2012. However, due to the Australia’s lower CPI during this period, in real terms, the AUD only increased by about 3.9% per year. Table A2 in the online appendix also shows a large variation in the yearly nominal exchange rate growth during the period, ranging from minus 5.3% (Iraq) to positive 37% (Iran). Controlling for CPI differences between Australia and the home countries of migrants, we still observe significant fluctuations in yearly real AUD exchange rates, ranging
4
In particular, we focus on countries with at least 50 observations surveyed in all years covered in our study period. The results are not sensitive when we increase the number of observations per country to 100. Migrants from the former Yugoslavia were excluded from our analysis because the country was separated into several countries before or during our study period and we are unable to observe which of the new countries the Australian immigrants come from. We additionally exclude 73 individual-year observations from Zimbabwe because the country experienced extreme macroeconomic fluctuations during the study period (for example, its CPI was above 24,000% in 2007). 5 We use 16 as the minimum age restriction to keep our sample comparable with that used by Nekoei (2013). The results are very similar when we increased the minimum age restriction to 24 (See Table A6 in the online appendix – Column 5 and 14).
5
from minus 2.8% (Iraq) to positive 34% (Iran). 6 These large fluctuations in the real exchange rates between countries over the study period and within countries over time thus validate our empirical strategy that exploits the changes in real exchange rates across home countries over time to identify the causal impact of exchange rate volatility on the immigrants’ labour market outcomes. [Table 1 and Figure 1 around here] Summary statistics by gender for the immigrant’s characteristics and labour market outcomes are presented in Table 1. Table 1 shows that about 53% of our sample is female. On average, immigrants in the sample are around 43 years old and male immigrants are slightly older (four months) than female immigrants. We also observe immigrants have lived in Australia for about 24 years and males have spent about four months in the host country longer than females. With exceptions of education and health, we notice significant gender differences in terms of all other characteristics such as language background and marital status. In particular, as compared to female immigrants, male immigrants are more likely to come from an English Speaking Background (ESB) country or be single. Gender differences in labour market outcomes are even more pronounced as they are all statistically significant at the 1% level. Consistent with empirical findings from the labour economics literature (Altonji and Blank, 1999), males in our sample are more likely to work and, conditionally on working, earn more than females. Our analysis additionally suggests that, while both working males and females desire to work less, males desire to work about one hour per week less than females. Finally, males in our sample are more likely to be self-employed and are less likely to change occupation than females. These gender differences indicate a need for a separate analysis by gender of immigrants as found in previous immigration literature (Adsera and Chiswick, 2007; Amuedo-Dorantes and De la Rica, 2007). 4.
Econometric model
We first follow Nekoei (2013) in estimating each of a series of labour market outcomes 𝑌𝑌𝑐𝑐𝑐𝑐𝑐𝑐
of immigrant 𝑖𝑖 from home country 𝑐𝑐 at time 𝑡𝑡 as follows:
𝑌𝑌𝑐𝑐𝑐𝑐𝑐𝑐 = 𝛼𝛼𝑐𝑐 + 𝛼𝛼𝑡𝑡 + 𝛽𝛽𝑒𝑒𝑐𝑐𝑐𝑐 + 𝑋𝑋𝑐𝑐𝑐𝑐𝑐𝑐 𝛾𝛾 + 𝑍𝑍𝑐𝑐𝑐𝑐 𝛿𝛿 + 𝜀𝜀𝑐𝑐𝑐𝑐𝑐𝑐
6
(1)
In principle, changes in nominal exchange rates among EURO zone countries (Ireland, Austria, Belgium, France, Germany, Netherlands, Finland, Italy, Malta, Portugal, Spain, Cyprus, Greece, and Latvia) should be the same at any point in time. However, due to the unbalanced nature of our panel sample, the average growth in nominal exchange rates is not the same for those countries for the whole 2001-2012 period.
6
In equation (1), 𝑒𝑒𝑐𝑐𝑐𝑐 represents the real exchange rate between Australia and home country c; 𝑋𝑋𝑐𝑐𝑐𝑐𝑐𝑐 is a vector of individual time-varying characteristics; 𝑍𝑍𝑐𝑐𝑐𝑐 is a vector of other
macroeconomic variables; and 𝜀𝜀𝑐𝑐𝑐𝑐𝑐𝑐 is a zero-mean error term. Home country fixed effects (𝛼𝛼𝑐𝑐 ) are included to capture time-invariant heterogeneity in immigrants’ labour market outcomes that may relate to countries of birth. Equation (1) additionally includes time fixed
effects (𝛼𝛼𝑡𝑡 ) to control for any shocks that are common to all countries in a given year. The
resulting identifying variation thus comes from changes in real exchange rates across home
countries over time. In seeking to replicate the approach of Nekoei (2013), we first treat our panel as a repeated cross-section, and estimate (1) using pooled least squares regression with time and home country effects. We then exploit the panel nature of our data to control additionally for individual fixed effects (𝛼𝛼𝑖𝑖 ) through the following fixed effects (FE) regression 7:
𝑌𝑌𝑐𝑐𝑐𝑐𝑐𝑐 = 𝛼𝛼𝑡𝑡 + 𝛼𝛼𝑖𝑖 + 𝛽𝛽𝑒𝑒𝑐𝑐𝑐𝑐 + 𝑋𝑋𝑐𝑐𝑐𝑐𝑐𝑐 𝛾𝛾 + 𝑍𝑍𝑐𝑐𝑐𝑐 𝛿𝛿 + 𝜀𝜀𝑐𝑐𝑐𝑐𝑐𝑐
(2)
The inclusion of individual fixed effects addresses a concern that the unobservable characteristics of immigrants could be correlated with some of the control variables in the labour market outcome equations such as the duration of stay in Australia, immigrants’ health status, or wealth. Failing to control for these unobservable traits could lead to potential biases in the estimated impact of exchange rates on labour market outcomes (Wooldridge, 2010). 4.1.
Labour market outcome variables
The main labour market outcomes examined in this study include: (1) whether the respondent was employed (including self-employed); (2) the number of weekly working hours conditional on working; (3) hourly wage rates 8 and (4) weekly wages conditional on working for wages. These four outcomes were measured in the week prior to the survey time. We also use annual wage earnings in the last financial year as the fifth labour market outcome. 9 Because HILDA survey does not have direct information about yearly labor supply, in order to provide results comparable to those presented in the previous work by Nekoei (2013) we divide the annual wage earnings by hourly wage rates to calculate the sixth labor market outcome, namely annual hours spent on wage work. Note that differences in the timing of the hourly wage rates and annual wage earrings may cause measurement errors in the calculated
The inclusion of individual time-invariant heterogeneity (𝛼𝛼𝑖𝑖 ) necessarily incorporates the unobservable country fixed effects (𝛼𝛼𝑐𝑐 ) included in (1). 8 Hourly wages are calculated from dividing the weekly wages by the weekly hours spent on working for wages. 9 In Australia, the financial year runs from 1st July to 30th June of the following year. All earnings are before tax. 7
7
annual working hours, which results in a reduction of the number of wage earners for whom annual wage working hours are calculated. 10 All outcomes, except the employment decision, are adjusted for CPI, using the 2001 CPI as the base, and measured in log form. Since exchange rates are also measured in log form, the coefficient estimates of exchange rate variable (𝛽𝛽) in equation (1) or (2) can be interpreted as the elasticity of an outcome (for instance, weekly wages) with respect to exchange rates.
It is possible that labour market structures (such as requirements about part-time or full-time working hours or labour market frictions) may not allow immigrants to fully respond to the exchange rate fluctuations as they desire (Altonji and Paxson, 1988; Dickens and Lundberg, 1993). To capture this possibility, we construct a variable called “hours desired to work more” as the difference between the desired and actual weekly working hours per week. 11 A positive value for this variable would indicate evidence of underemployment as working individuals desire to work more than they actually do (Wooden et al., 2009). We also capture the impact of exchange rate fluctuations on other labour market outcomes such as participating in the labour market at a particular mode (full-time, part-time or self-employed) or changing occupation since last interview. 4.2.
Other explanatory variables
Other explanatory variables include age (and its square), duration of stay in Australia (and its square), education, English Speaking Background (ESB), 12 marital status and health status of the individual immigrants. We purposely do not include some economic variables that would be reasonably affected by home country’s macroeconomic fluctuations such as household non-labour income and home ownership status.13 Household characteristics in the models include the number of co-residing members of various age cohorts. We additionally control for differences in working conditions across regions by including the regional unemployment
10
Due to differences in sample size, log (annual wage earnings) may not equal to the sum of log (hourly wage rates) and log (annual wage working hours). 11 The desired working hours are derived from the question “In total, how many hours a week, on average, would you choose to work? Again, take into account how that would affect your income”. Note that this question is only asked for individuals employed (including self-employed) in the last week prior to the survey time. We exclude the self-employed from the calculation of this variable because their reports of actual and desired working hours may not be realizable. For a similar reason, the self-employed are not included in the calculation of the weekly working hours variable. 12 English Speaking Background countries include the United Kingdom (UK), New Zealand, Canada, US, Ireland and South Africa. See Table A1 for details of variable definition. 13 We thank a referee for this suggestion. These variables are sometimes used in the labour economics literature to examine the effect of wealth on labour supply. Unreported results show that our results are not sensitive to the inclusion of these variables in the regressions.
8
rate, regional relative socio-economic advantage index, and state dummies 14 in the labour market outcome equations. We also control for the heterogeneity in the time of survey by controlling for year and month fixed effects. To capture assimilation profile of the immigrants, in pooled models, we also include dummy variables for various groups of immigrants with time of arrival in five-year-bands. Note that all variables representing the length of stay in Australia are not identified in our fixed effects models because they have already included other three time-dimension variables (i.e. immigrant’s age, year dummies, and individual FE). In addition, our fixed effect models which control for individual-specific heterogeneity associated with arrival cohorts also capture cohort-specific unobserved characteristics affecting immigrant’s labour market outcomes (Borjas, 1999). We recognize the possible differences in labour market patterns and assimilation between Australian males and females (Meng and Gregory, 2005; Chiswick and Miller, 2013) by analysing males and females separately. 15 For ease of interpretation, we use Ordinary Least Squared (OLS) to estimate all equations. 16 Standard errors are clustered at the country level to account for possible correlation within countries of origin (Moulton, 1986; Bertrand et al., 2004; Cameron and Miller, 2015). 5.
Empirical results
5.1.
Results for a pooled sample of males and females
For comparison purposes with the US study by Nekoei (2013), we first report exchange rate effects where we do not separate the sample by gender. Results from this exercise are reported in Table 2. Consistent with the US study, we also find a statistically insignificant impact of exchange rates on the work probability from pooled results which do not control for individual fixed effects (Column 1 to 4 in Table 2). In line with findings reported in the work by Nekoei (2013), pooled results for two annual labour market outcome variables also suggest a negative and statistically significant (at least at the 10% level) effect of exchange 14
The inclusion of state/territory dummies also accounts for possible internal migration patterns. Our data show that about 13% of immigrants moved interstate each year. 15 The assimilation patterns of male and female immigrants have also been found different in other countries such as European countries (Adsera and Chiswick, 2007) and Spain (Amuedo-Dorantes and De la Rica, 2007). Indeed, in our data, unreported results from a Hausman - type test reject the null hypothesis of the equality of the estimates for males and females at the 1% level for all labour market outcomes, suggesting that labour market behaviors of males and females should be modelled separately. It should be noted that Nekoei (2013) does not analyse by gender. However, in an Appendix Table (Nekoei, 2013 Table A3), he shows that estimation results are largely the same for males and females. 16 Logit estimates for the binary outcome variable (whether the individual worked in the week prior to the survey) are largely similar to those OLS estimates presented here. Logit results for pooled data are available upon request. Unfortunately, the fixed effect Logit model for this outcome failed to converge, probably because of the large number of dummy variables used.
9
rates. However, in terms of the size, Australian estimates are much higher than those reported in the US study. Particularly, Australian estimates are about three (two) times greater than US estimates for annual working hours (annual earnings). Table 2 (Column 5 to 8) additionally shows that controlling for individual FEs turns the exchange rate effects on these two annual labour market outcomes to statistically insignificant. Other FE results in Table 2 indicate a negative and statistically significant effect of exchange rates on weekly working hours, weekly wages and the probability of working full-time. 17 We however do not observe any statistically significant exchange rate effect on hourly wage rates, desired working hours or the probability of being self-employed or changing occupation. [Table 2 around here] As explained above, labour market outcomes of immigrants should be modelled separately by gender of immigrants. Therefore, in what follows we will report results separately for males and females. 5.2.
Results for main labour market outcomes
Table 3 presents exchange rate estimates on main labour market outcomes from both pooled and fixed effects models. We discuss the pooled results first. Our pooled results (Table 3 – Column 1 for females and Column 9 for males) show that exchange rate fluctuations have a statistically significant (at the 5% level) and opposite impact on the employment decision of female and male immigrants. In particular, for females, a 10% appreciation of the AUD reduces their labour market participation probability by about 0.7 percentage point. This impact is quite substantial (equivalent to a 1.1%18 reduction in mean working rates for female immigrants) given an already low labour market participation rate for females (65%) and the large appreciation of the AUD during the study period (3.9% per annum). By contrast, a 10% appreciation of the AUD increases the working probability of males by 0.5 percentage point. Our finding of a statistically significant (but opposite) impact of the exchange rate appreciation for females and males is new to the literature since the previous work finds that an appreciation of the USD does not have any significant impact on the working probability of US male and female immigrants. [Table 3 around here] 17
Unfortunately, Nekoei (2013) only presents results for yearly labour market outcomes so that we cannot compare our results on weekly labour market outcomes to his. 18 This figure is calculated as (0.07/0.65)*100=1.1% where 0.65 is the average working rate for female immigrants in our sample.
10
Pooled results also show that, conditional on working, in response to an appreciation of the AUD, male and female immigrants appear to reduce their weekly working hours and weekly wage earnings. The estimates, however, are statistically significant (at the 5% level) for males’ weekly wages only. These estimates indicate that, in response to a 10% appreciation of the AUD, male immigrants reduce their weekly wage earnings by about 1%. When labour market outcomes are measured over the last financial year, estimates from pooled regressions suggest that annual working hours and annual earnings are much more responsive to exchange rate fluctuations than before. However, estimates are statistically significant (at the 5 (10) % level) for annual working hours for females (males) only. Now in response to a 10% appreciation of the AUD, female (male) immigrants reduce their annual working hours by about 2.0 (1.0) %. Thus, estimates of annual wage working hours for Australian female immigrants are in line with those reported for their US counterparts in terms of the sign and significant level (Nekoei, 2013). However, in terms of the magnitude, while estimates for Australian males are largely similar to the estimates for the US males and females, the Australian estimates for females are much higher than those reported for the US where in response to a 10% USD appreciation, US female immigrants reduce their annual working hours by only about 0.5%. These findings are interesting given that immigrants in our sample have spent a longer time (about 8 years) in the host country than those in the US study. Furthermore, our finding of a negative but statistically insignificant impact of exchange rate fluctuations on annual wages (for male and female immigrants) is not in line with the previous US finding of a negative and statistically significant impact. We now turn to discuss FE results (Column 5 for females and Column 13 for males). The F test statistics confirm that FE models are preferred to OLS models and this is the case for all labour market outcomes. 19 These test results suggest that there are some unobservable timeinvariant individual characteristics which are correlated with the exchange rate fluctuations and labour market outcomes at the same time and these characteristics should be controlled for when modelling immigrant labour market outcomes. Indeed, FE results which control for those unobservable individual characteristics show significant changes to our previous findings. In particular, for both females and males, estimates turn from statistically significant to insignificant for work decision and annual working hours. As such, FE estimates point to a 19
F statistics are calculated in the models not allowing robust standard errors because it is very complicated to do so. For brevity, F statistics are not reported here but they will be available upon request. Our results for standard errors (reported in parentheses) are about the same between pooled and FE regressions, indicating that insufficient variation in explanatory variables is not a problem for our data (Allison, 2009).
11
statistically insignificant impact of exchange rate fluctuations on all main labour market outcomes for Australian females and males. Overall, our results which control for individual heterogeneity suggest that main labour market behaviours of male and female Australian immigrants are not responsive to the exchange rate fluctuations. Other results from Table 3 show that the effects of exchange rate fluctuations are not contaminated by the impact of other home country’s macroeconomic conditions (such as GDP per capita (results reported in Column 2, 6, 10 and 14 of Table 3) or unemployment rates (Column 3, 7, 11, 15 of Table 3) since estimates for the exchange rate variables are largely unchanged when other macroeconomic variables are introduced to the model. However, the inclusion of region-time dummies 20 to account for the possibility that exchange rate movements are correlated across regions turn estimates of some weekly labour market outcomes for males from statistically insignificant to significant (Column 4, 8, 12, and 16 of Table 3). For example, FE estimates now suggest that in response to a 10% appreciation of the AUD, male immigrants reduce their weekly working hours (weekly wages) by about 1.3 (1.9) % (column 16). 5.3.
Results for other labour market outcomes
Pooled estimates for the “hours desired to work more” variable (the first row – column 1 of Table 4) indicate that working female immigrants desire to work more when the AUD appreciates. For them, a 1% appreciation of the AUD makes them desire to work about 2.02 hours per week more. We also note that the pooled estimate of exchange rate on the “hours desired to work more” variable is quite stable when we include other macroeconomic variables or region-time FEs (Column 2 to 4) in the regressions. These results when viewed with earlier results that female immigrants do not reduce their actual weekly working hours in response to an AUD appreciation suggest some labour market structures may exist that prevent working female immigrants from working more. FE results (Column 5 to 8) indicate a smaller and less statistically significant impact of exchange rate on the hours desired to work more for females. Finally, controlling for both individual FE and region-time dummies turns the estimate to statistically insignificant, suggesting that female immigrants do not desire to work more when the AUD appreciates. Similarly, we do not observe any statistically 20
We thank a referee for his or her comment which leads us to use this arguably more robust specification. We do not include home country–time FEs in our regressions because within a country, exchange rate movements are highly correlated (as can be seen from Figure 1). Indeed, when we do this, we get very strange results. We use the World Bank’s classification to group countries into seven regions: East Asia & Pacific (the base group), Europe & Central Asia, Latin America & Caribbean, Middle East & North Africa, North America, South Asia, and Sub-Saharan Africa.
12
significant impact of exchange rates on the hours desired to work more by males from all specifications. [Table 4 around here] Other pooled results from Table 4 suggest that an appreciation of the AUD statistically significantly (at the 10% level) decreases the probability of working full-time for females only. However, also for females, controlling for region-time FEs or individual FEs turns the exchange rate impact on the probability of working full-time from statistically significant to insignificant. In contrast, controlling for individual unobserved heterogeneity turns the exchange rate impact on the probability of working in a particular labour force status from statistically insignificant to significant (at the 5 (10) % level for worked part-time variable (self-employed) for males (column 13). FE estimates with region-time dummies (column 16) show that, in response to a 10% appreciation of the AUD, male immigrants increase their part-time (self) employment probability by about 0.9 (0.8) percentage points. This finding is consistent with our earlier finding that only male immigrants reduce weekly labour supply when the AUD appreciates. This finding when viewed with an early finding of an insignificant impact of exchange rate on the probability of being employed suggests that male immigrants respond to exchange rate appreciation not by exiting the labour market but by reducing their work intensity. Finally, we do not find any significant impact of exchange rate volatility on the probability of changing occupation for both males and females. 5.4.
Timing of the impact
It is possible that immigrants may need some time to adjust their labour market decisions to exchange rate volatility. To account for the possible dynamics of exchange rate fluctuations we introduce lagged exchange rates to the equation (2). Estimates for different lags of exchange rates are obtained from separate regressions and results for main (other) labour market outcomes variables are reported in Table A3 (A4) in the online appendix. Results from these tables show that all current labour market outcomes of female immigrants are not responsive to exchange rate fluctuations which took place in the previous year or two years ago. An exception is observed where an appreciation of the AUD in the last two years is found to decrease their probability to work full-time and the effect is statistically significant at the 5% level (Column 2 of Table A4 in the online appendix). For males, estimates for lagged exchange rates are often less pronounced in terms of the magnitude or statistical significance level than those for current exchange rates obtained earlier.
13
6.
Robustness checks
6.1.
Return immigrants
We first check whether exchange rate volatility influences the decision to return of the immigrants in our sample. To do this, we estimate a model similar to the model (2) where the dependent variable is replaced by an indicator taking value 1 if the immigrant moves overseas (and hence is not surveyed in that year) and zero otherwise. We include the same list of explanatory variables as described above for the model (2). We include either contemporaneous or lagged exchange rates. Results (Table A5 in the online appendix) indicate that an appreciation of the AUD increases the probability of return but statistically insignificant. We also find no effect of previous year exchange rate fluctuations on the current decision to return for both males and females.21 Only the 2-year lag of exchange rates is found to marginally (at the 10% level of significance) affect the probability of return in current year for male immigrants. Results of this robust check suggest that our earlier findings on the impact of exchange rate fluctuations on the labour market outcomes are not sensitive to non-random sample attrition due to return immigration. 6.2.
Threats to individual FE identification approach
Despite the advantages of the individual FE models of removing time-invariant unobserved characteristics, the individual FE approach cannot address reverse causality. Nevertheless, it is unlikely that individual labour market activities affect exchange rate fluctuations so reverse causality is not an important concern in our models. However, there are still concerns that our empirical models may miss some important time-variant variables which are correlated with both exchange rate fluctuations and immigrants’ labour market outcomes. One may anticipate that a regional macroeconomic crisis that affects both Australia and a country where its immigrants come from to be one such unobservable factor. Correlated weather shocks could be another factor. These two events could affect demand for immigrants as well as exchange 21
These results are not contrary to the findings of Yang (2006) and Abarcar (2013) because we use different samples of immigrants and different exchange rate shocks. In particular, immigrants in our sample are permanent and spend a much longer time in the host country than those in other studies (i.e. most of immigrants are temporary in the study of Yang (2006) and immigrants are recently arrived and have spent on average two years in Australia in the work by Abarcar (2013)). It has been shown that immigrants who spend longer time in the host country are less affected by the home country shocks (Nekoei, 2013; Nguyen and Duncan, 2015; Akay et al., 2016). Indeed, Abarcar (2013) also finds that exchange rate shocks have no impact for those who initially stated their desire to stay in Australia. Moreover, we study small and frequent exchange rate fluctuations whereas other studies analyse large and infrequent shocks (such as the Asian crisis) which are more likely to influence return migration.
14
rate movements at the same time, causing the estimates of exchange rates on immigrants’ labour market outcomes 22 to be biased. We address the variable omission bias concern by additionally including more variables describing such possible factors in the regressions. Previously, in our extended regressions, we included some variables representing business cycles at home countries such as GDP growth and unemployment rates in addition to the existing exchange rate variable to the regressions and found our results robust to the inclusion of such macroeconomic variables (See Section 5). In this section, to measure home country macroeconomic shocks, we additionally introduce the home country’s GDP per capita and terms of trade to the regressions and find the exchange rate estimates are largely unchanged (See columns 6, 7, 15 and 16 of Table A6 in the online appendix). Furthermore, to control for weather shocks from home countries, we include a set of variables describing the frequency and magnitude of natural disasters occurring at home countries during the survey year. 23 These variables include the number of natural disasters, the number of people affected and the amount of damage to property, crops, and livestock occurred because the natural disasters happened during the survey year. 24 Results of this robustness check (reported in columns 8 and 17) suggest that our findings are largely not driven by weather shocks either. 25 Overall, the above robustness checks confirm that exchange rate has an effect separate from that of other macroeconomic and weather shock variables, and increase the credibility of the identification strategy. 6.3.
Other robustness checks
We also examine the robustness of our results to alternative selections of country, year, age, and measure of exchange rates. First, UK immigrants represent a largest share (28% as seen in Table A2) of all immigrants in Australia. We gauge whether the results change when UK 22
Recall that in all regressions, we control for local labour market conditions (such as regional unemployment rates and socio-economic index) to further address that a concern that such events may affect the immigrants’ labour market outcome via changing local labour market demand for immigrants. 23 We are grateful to a referee’s suggestion which leads us to further control for other time variant variables. 24 These variables are achieved from the International Disaster Database (Guha-Sapir et al.). 25 Two exceptions are observed. First, the exchange rate impact is no longer statistically significant for males’ weekly working hours and wages. Second, the estimate of exchange rate on males’ hours desired to work more is negative and statistically significant at the 1% level. However, the estimate is unreasonably high. Our further investigation suggests that including both region-time dummies and disaster variables in the regression may cause multi-collinearity problems because these variables are highly correlated across regions and overtime. As such, they should not be introduced together in regressions. Indeed, when we exclude region-time dummies but still include weather variables in the regression, the estimate of exchange rate on males’ hours desired to work is much smaller (- 0.81) and not statistically significant at any conventional level.
15
immigrants are excluded from the regression. Results of this experiment (reported in Column 2 and 11 of Table A6 in the online appendix) are very similar to the baseline results (rereported in Column 1 and 10 of Table A6 for ease of comparison), suggesting that our results are not driven by the UK immigrants. 26 Second, using a larger sample of 26,315 observations from 113 countries for them we are able to find sufficient macroeconomic variables we get results (reported in Column 3 and 12 of Table A6) that are very similar to the baseline results. Third, we check the sensitivity of our results to the recent global financial crisis which caused the AUD to drop significantly in 2008. Results (Column 4 and 13 of Table A6) do not greatly vary from the baseline after excluding the years 2008 and 2009 from our sample. Fourth, our prior findings are largely unchanged when we apply the regression (2) to a sample of older Australian immigrants aged between 24 and 64 (Column 5 and 14 of Table A6). Finally, we check whether the effects are sensitive to the measure of exchange rates. The timing of measurement of exchange rate and labour supply variables would be an issue. In our study, real exchange rates are measured at a calendar year basis (i.e. from 1st January to 31st December each year). By contrast, labour supply variables are measured for either a week or a financial year prior to the survey time. In HILDA, the interviews are conducted annually with most of interviews occurring in September and October. 27 Therefore using the exchange rate and labour supply variables which are measured at the same year may not capture the full impact of real exchange rates on labour market outcomes. In this sensitivity test, we use the exact survey date to link with the daily exchange rates to measure the average exchange rates for a full 12 months prior to the survey time and find similar results (See Column 9 and 18 of Table A6). 28 7.
Heterogeneity among immigrants
Above, using FE models, we found that only male immigrants as a whole adjust their weekly labour market outcomes in response to exchange rate fluctuations. It may be the case that immigrants with different socio-economic background respond heterogeneously to exchange rates fluctuations. We investigate the heterogeneity of the impact by estimating the regression (2) for two sub-groups, separated by each variable of a series of variables which represent 26
An exception is the positive effect of exchange rates on the probability of working of males which turns from statistically insignificant to highly significant (at the 1% level). 27 In particular, 50, 24, 13, and seven % of immigrants in our sample were interviewed in September, October, August, and November, respectively. 28 In this experiment, we still use yearly CPI to calculate real exchange rate because CPI on a shorter time horizon (such as quarterly) is not readily available for all countries in our sample. As already mentioned, we also include survey month dummies in this exercise.
16
socio-economic background of the immigrants, their ties with home countries or return probabilities. These variables include the immigrant’s age, the duration of stay in Australia, marital status, the presence of children, education level, citizenship status29, the presence of a close family member (i.e. parents and siblings) overseas, whether the immigrant speaks a language other than English at home, and whether the immigrant reports that he or she speaks English very well. 30 In addition to the above individual characteristics, we also consider the immigrant’s home country characteristics such as whether the country is an ESB country, the physical distance between the home country and Australia, whether the country is classified as a high income country by the World Bank, the home country’s democracy index, and the country’s remittance inflow/GDP ratio. 31 For each of non-binary variables (for example, age, the duration of stay in Australia, the air distance from the home country to Australia, the home country’s democracy index, and the country’s remittance/GDP ratio), sub-groups are defined relative to the median of the respective male and female sample. We focus our heterogeneous analysis on the log of weekly wages because we find its estimate statistically significant for males and the annual wage earning outcome because the previous US study by Nekoei (2013) reports heterogeneous analysis for it. 32 Exchange rate estimates in the weekly wage regressions by sub-groups (panel B of Figure 2) suggest that the negative exchange rate impact on this outcome observed earlier for male immigrants is mainly driven by those who are young, recent immigrants, highly educated, speak other languge at home, have any sibling overseas, or come from countries which are classifed as non-ESB, more geographically distant from Australia, or low income countries because exchange rate estimates for those sub-groups are greater (i.e. more negative) or more statistically significant. However, taking the statistical differences of exchange rate estimates by subpopulations into account indicates the impact of exchange rate fluctuations is not statistically 29
Questions about citizenship are only asked once for all respondents, starting from wave 2 for all respondents and only for new entrants from wave 3. Similarly, questions about residential locations of parents and siblings are only surveyed in Waves 8 and 12. We use the panel nature of our data to fill in missing information for these variables in other waves. It is possible that these variables change overtime that our data cannot capture. Unfortunately, HILDA does not provide the exact overseas locations of family members so that we cannot identify whether family members live in the immigrants’ home countries. 30 While the last two language indicators are highly correlated (in our data, their correlation coefficient is minus 0.5 and statistically significant at the 1% level), we examine them separately because they may represent immigrants’ ties with home countries or their assimilation to Australia differently. 31 The remittance/GDP ratio is averaged over the study period (i.e. 2001-2012) because, for some countries, data are not available for all years studied. Similarly, the democracy index, which is provided by the Economic Intelligent Unit with a higher index representing a higher level of democracy, is averaged over the 2006-2012 period. We use Google map to measure the air distance between Australia and the home country. 32 Unreported results for remaining labour market outcomes do not show any statistically significant differences between sub-groups or a pattern that could easily be interpreted.
17
significantly different by all characteristics we consider here. 33 Also consistent with our earlier finding that females did not adjust their weekly labour supply when the AUD appreciates, we find almost all sub-group estimates are statistically insignificant (panel A of Figure 2). [Figure 2 around here] By-sub-group estimates for exchange rate impacts on annual wage earnings (results are reported in Figure A1 in the online appendix) are in line with our earlier results obtained for the whole population of males and females. These sub-group results confirm that male and female immigrants indeed did not statistically significantly adjust their annual labour supply in response to exchange rate fluctuations. We also do not find any significant evidence supporting for differential impact of exchange rate by sub-groups for this labour market outcome. 34 8.
Discussions
Our empirical results show controlling for individual fixed effects leads to a change in the impact of exchange rate shocks on some labour market outcomes. As mentioned above, immigrants from different countries are likely to differ in both ability and motivation; it is also possible that immigrants from the same country arriving at different points in time differ from one another in terms of their unobservable characteristics. While the OLS regression model (1) with home country and cohort FEs helps control for such unobservable characteristics, it cannot control for other individual unobservable characteristics that may be associated with exchange rate fluctuations and individual labour market decisions at the same time. One such characteristic could be the immigrant’s propensity to transfer in response to exchange rate fluctuations (Yang, 2008). Another related unobserved individual characteristic would be the immigrant’s propensity to save with respect to exchange rate movements (Bauer and Sinning, 2011). The immigrant’s propensity to consume could be another unobserved
33
Visually, the 95% confidence intervals which do not include zero indicate a statistically significant (at the 5% level) estimate. The statistically significant differences in the estimates by sub-population are indicated visually by the observation that the 95% confidence intervals do not overlap. An alternative approach to test for the statistical differences in sub-group estimates is to introduce an interaction term between one of the above mentioned socio-economic background variables and the log of exchange rate to the equation (2) and test for the statistical significance of that interaction term. Unreported results from this alternative approach are largely similar to the results reported here. 34 An exception is that females with and without children appear to react differently in response to exchange rate fluctuations (panel A – Figure A1 in the online appendix). Particularly, in response to an AUD appreciation, while females with children (representing about 70% of our sample) tend to reduce their yearly wages, those without children appear to increase their yearly wages.
18
factor that would be correlated with both exchange rate movements and labour market outcomes (Devereux et al., 2012). Failing to account for such individual unobservable characteristics may bias the estimates of exchange rate impacts on labour market outcomes. Unfortunately, the direction of the potential bias caused by different sources of individual unobservable characteristics is ambiguous because we lack strong empirical evidence on how exchange rate impact varies by immigrant characteristics. This work has highlighted the importance of controlling for individual heterogeneity when modelling the labour market behaviour of immigrants. While this finding may be specific to our data and application, it is consistent with evidence presented in previous studies demonstrating the importance of controlling for unobserved individual heterogeneity in modelling labour market outcomes of immigrants (Borjas, 1985; Hu, 2000; Duleep and Dowhan, 2002; Dustmann and Fabbri, 2005; Lubotsky, 2007; Cobb-Clark et al., 2012; Breunig et al., 2013; Abramitzky et al., 2014). Some recent studies which use a similar approach to ours to examine the impact of other macroeconomic variables (such as regional unemployment rates) on individual behaviours (e.g. labour market outcomes, health outcomes, or health behaviours) have also suggested the benefits of controlling for individual time-invariant unobservable characteristics (Dávalos et al., 2012; Currie et al., 2015; Foged and Peri, 2015). It is interesting to observe that using a largely similar measure of immigrants’ yearly labour market outcomes, Australian study (i.e. this study) provides little evidence supporting findings of a negative and statistically significant impact of exchange rate appreciation on immigrants’ labour supply presented in the previous US study by Nekoei (2013). Besides differences in our treatment of FEs as discussed above, another possible explanation for our differences in findings is that as immigrants in the two countries are not the same, neither are their behaviours (Antecol et al., 2003; Antecol et al., 2006; Chiswick et al., 2008; Clarke and Skuterud, 2013). Relative to the US, Australia maintains a skilled immigrant selection policy producing immigrants with different human capital characteristics (Antecol et al., 2006). Furthermore, differences in the socio-economic environments that immigrants live in may be another factor contributing to the differences in our findings. One of such socio-economic environments could be labour market flexibility as represented by hiring and firing practices, flexibility in wage determinations and redundancy costs. It has been evidenced that employment in countries or industries with a higher labour market rigidity is less responsive to exchange rates shocks (Alexandre et al., 2015). As the labour market is less flexible in 19
Australia than in the US (Cuñat and Melitz, 2012; Schwab and Sala-i-Martín, 2015), our findings of the non-responsiveness of yearly labour market supply to exchange rate fluctuations are consistent with the macroeconomic evidence presented by Alexandre et al. (2015). While our preferred FE results (presented in Table 2 and 3) suggest that exchange rate effects on most labour market outcomes are not apparently dissimilar between male and female immigrants, estimates on the weekly working hours and wages convey that male and female immigrants appear to differ in how much they respond when the Australian dollar appreciates. 35 Specifically, we find only male immigrants reduce their weekly labour supply when the Australian dollar appreciates. This gender difference is consistent with usually documented labour market disadvantages of female immigrants over male immigrants in Australia (Miller and Chiswick, 1985; Beggs and Chapman, 1988; Breunig et al., 2013). Australian female immigrants who are often secondary migrants in skilled-visa streams either struggle to join the workforce or end up in low-paid jobs. It would be possible that their weaker labour market skills may also prevent them from adjusting their labour supply when the Australian dollar appreciates. 9.
Conclusion
Using various groups of immigrants from 65 countries of origin and panel data for twelve years, we have investigated how Australian immigrants adjust their labour market activities in response to plausibly exogenous exchange rate fluctuations. Our results from models which do not control for individual unobservable time-invariant characteristics and do not examine labour market outcomes by gender of immigrants show a similar exchange rate impact on yearly labour market outcomes as found in a previous US study by Nekoei (2013). Results from models controlling for individual heterogeneity however suggest that, unlike immigrants in the US, those in Australia as a whole do not reduce their yearly labour market outcomes when the local currency appreciates. Our work further uncovers that only male immigrants reduce their weekly labour market activities in response to an appreciation of the host country’s currency. In particular, in response to a 10% appreciation of the AUD, Australian male immigrants reduce their weekly working hours (weekly wage earnings) by about 1.3 (1.9) %.
35
Results from tests similar to those used in Section 7 reach a similar conclusion.
20
This work has highlighted the importance of controlling for individual heterogeneity when modelling the labour market behaviour of immigrants. Future work may extend the topic to other countries’ data or study the impact of exchange rate fluctuations on other aspects of immigrant behaviour such as consumption, saving and transfers.
21
References
Abarcar, P., 2013. The Return Motivations of Legal Permanent Migrants: Evidence from Exchange Rate Shocks and Immigrants in Australia: MPRA Paper No. 47832. Abramitzky, R., Boustan, L.P., Eriksson, K., 2014. A Nation of Immigrants: Assimilation and Economic Outcomes in the Age of Mass Migration. Journal of Political Economy 122, 467-506. Adsera, A., Chiswick, B.R., 2007. Are there gender and country of origin differences in immigrant labor market outcomes across European destinations? Journal of Population Economics 20, 495-526. Akay, A., Bargain, O., Zimmermann, K.F., 2016. Home Sweet Home? Macroeconomic Conditions in Home Countries and the Well-Being of Migrants*. Journal of Human Resources forthcoming. Alexandre, F., Bação, P., Cerejeira, J., Portela, M., 2015. Exchange Rates, Employment and Labour Market Rigidity. The World Economy forthcoming. Allison, P.D., 2009. Fixed effects regression models: London : SAGE. Altonji, J.G., Blank, R.M., 1999. Chapter 48 Race and gender in the labor market, In: Orley, C.A., David, C. (Eds.), Handbook of Labor Economics: Elsevier, pp. 3143-3259. Altonji, J.G., Paxson, C.H., 1988. Labor Supply Preferences, Hours Constraints, and Hours-Wage Trade-offs. Journal of Labor Economics, 254-276. Amiti, M., Itskhoki, O., Konings, J., 2014. Importers, Exporters, and Exchange Rate Disconnect. American Economic Review 104, 1942-1978. Amuedo-Dorantes, C., De la Rica, S., 2007. Labour market assimilation of recent immigrants in Spain. British Journal of Industrial Relations 45, 257-284. Antecol, H., Cobb-Clark, D.A., Trejo, S.J., 2003. Immigration Policy and the Skills of Immigrants to Australia, Canada, and the United States. The Journal of Human Resources 38, 192-218. Antecol, H., Kuhn, P., Trejo, S.J., 2006. Assimilation via prices or quantities? Sources of immigrant earnings growth in Australia, Canada, and the United States. Journal of Human Resources 41, 821840. Bauer, T., Sinning, M., 2011. The savings behavior of temporary and permanent migrants in Germany. Journal of Population Economics 24, 421-449. Beggs, J.J., Chapman, B.J., 1988. Immigrant Wage Adjustment in Australia: Cross Section and TimeSeries Estimates*. Economic Record 64, 161-167. Bertrand, M., Duflo, E., Mullainathan, S., 2004. How Much Should We Trust Differences-inDifferences Estimates? The Quarterly Journal of Economics 119, 249-275. Borjas, G.J., 1985. Assimilation, Changes in Cohort Quality, and the Earnings of Immigrants. Journal of Labor Economics 3, 463-489. Borjas, G.J., 1999. Chapter 28 The economic analysis of immigration, In: Orley, C.A., David, C. (Eds.), Handbook of Labor Economics: Elsevier, pp. 1697-1760. Breunig, R., Hasan, S., Salehin, M., 2013. The immigrant wage gap and assimilation in Australia: does unobserved heterogeneity matter? Economic Record 89, 490-507. Camerer, C., Babcock, L., Loewenstein, G., Thaler, R., 1997. Labor Supply of New York City Cabdrivers: One Day at a Time. The Quarterly Journal of Economics 112, 407-441. Cameron, A.C., Miller, D.L., 2015. A Practitioner’s Guide to Cluster-Robust Inference. Journal of Human Resources 50, 317-372. Campa, J.M., Goldberg, L.S., 2001. Employment versus wage adjustment and the US dollar. Review of Economics and Statistics 83, 477-489. Chang, T., Gross, T., 2014. How many pears would a pear packer pack if a pear packer could pack pears at quasi-exogenously varying piece rates? Journal of Economic Behavior & Organization 99, 117. Chiswick, B.R., Le, A.T., Miller, P.W., 2008. How immigrants fare across the earnings distribution in Australia and the United States. Industrial and Labor Relations Review 61, 353. Chiswick, B.R., Miller, P.W., 2013. Negative and Positive Assimilation By Prices and By Quantities. IZA Discussion Paper No.7389. Chou, Y.K., 2002. Testing alternative models of labour supply: Evidence from taxi drivers in Singapore. The Singapore Economic Review 47, 17-47.
22
Clarke, A., Skuterud, M., 2013. Why do immigrant workers in Australia perform better than those in Canada? Is it the immigrants or their labour markets? Canadian Journal of Economics/Revue canadienne d'économique 46, 1431-1462. Cobb-Clark, D., Hanel, B., McVicar, D., 2012. Immigrant Wage and Employment Assimilation: A Comparison of Methods. IZA Discussion Paper Number 7062. Crawford, V.P., Meng, J., 2011. New York city cab drivers' labor supply revisited: Referencedependent preferences with rationalexpectations targets for hours and income. The American Economic Review 101, 1912-1932. Cuñat, A., Melitz, M.J., 2012. Volatility, labor market flexibility, and the pattern of comparative advantage. Journal of the European Economic Association 10, 225-254. Currie, J., Duque, V., Garfinkel, I., 2015. The Great Recession and Mothers' Health. The Economic Journal 125, F311-F346. Dávalos, M.E., Fang, H., French, M.T., 2012. Easing the pain of an economic downturn: Macroeconomic conditions and excessive alcohol consumption. Health Economics 21, 1318-1335. Devereux, M.B., Smith, G.W., Yetman, J., 2012. Consumption and real exchange rates in professional forecasts. Journal of International Economics 86, 33-42. Dickens, W.T., Lundberg, S.J., 1993. Hours Restrictions and Labour Supply. International Economic Review 34, 169-192. Duleep, H.O., Dowhan, D.J., 2002. Insights from longitudinal data on the earnings growth of US foreign-born men. Demography 39, 485-506. Dustmann, C., Fabbri, F., 2005. Gender and ethnicity—married immigrants in Britain. Oxford Review of Economic Policy 21, 462-484. Ekholm, K., Moxnes, A., Ulltveit-Moe, K.H., 2012. Manufacturing restructuring and the role of real exchange rate shocks. Journal of International Economics 86, 101-117. Faini, R., 1994. Workers remittances and the real exchange rate. Journal of Population Economics 7, 235-245. Fehr, E., Goette, L., 2007. Do Workers Work More if Wages Are High? Evidence from a Randomized Field Experiment. The American Economic Review 97, 298-317. Fertig, M., Schurer, S., 2007. Labour Market Outcomes of Immigrants in Germany: The Importance of Heterogeneity and Attrition Bias. IZA DP No. 2915. Foged, M., Peri, G., 2015. Immigrants' Effect on Native Workers: New Analysis on Longitudinal Data. American Economic Journal: Applied Economics Forthcoming. Gerald S. Oettinger, 1999. An Empirical Analysis of the Daily Labor Supply of Stadium Venors. Journal of Political Economy 107, 360-392. Gordon, H.H., Spilimbergo, A., 1999. Illegal Immigration, Border Enforcement, and Relative Wages: Evidence from Apprehensions at the U.S.-Mexico Border. The American Economic Review 89, 13371357. Guha-Sapir, D., Below, R., Hoyois, P., 2009. EM-DAT: The CRED/OFDA International Disaster Database – www.emdat.be – Université Catholique de Louvain – Brussels – Belgium. Héricourt, J., Poncet, S., 2013. Exchange Rate Volatility, Financial Constraints, and Trade: Empirical Evidence from Chinese Firms. The World Bank Economic Review. Hu, W.-Y., 2000. Immigrant earnings assimilation: estimates from longitudinal data. American Economic Review, 368-372. Hum, D., Simpson, W., 2004. Reinterpreting the performance of immigrant wages from panel data. Empirical Economics 29, 129-147. Kandilov, I.T., Leblebicioğlu, A., 2011. The impact of exchange rate volatility on plant-level investment: Evidence from Colombia. Journal of Development Economics 94, 220-230. Klein, M.W., Schuh, S., Triest, R.K., 2003. Job creation, job destruction, and the real exchange rate. Journal of International Economics 59, 239-265. Lubotsky, D., 2007. Chutes or ladders? A longitudinal analysis of immigrant earnings. Journal of Political Economy 115, 820-867. McKenzie, D., Theoharides, C., Yang, D., 2014. Distortions in the International Migrant Labor Market: Evidence from Filipino Migration and Wage Responses to Destination Country Economic Shocks. American Economic Journal: Applied Economics 6, 49-75.
23
Meng, X., Gregory, Robert G., 2005. Intermarriage and the Economic Assimilation of Immigrants. Journal of Labor Economics 23, 135-174. Miller, P.W., Chiswick, B.R., 1985. Immigrant Generation and Income in Australia*. The Economic Record 61, 540-553. Moulton, B.R., 1986. Random group effects and the precision of regression estimates. Journal of Econometrics 32, 385-397. Nekoei, A., 2013. Immigrants' Labor Supply and Exchange Rate Volatility. American Economic Journal: Applied Economics 5, 144-164. Nguyen, H.T., Duncan, A., 2015. Macroeconomic fluctuations in home countries and immigrants’ well-being: New evidence from Down Under. Bankwest Curtin Economics Centre working paper number 15/2. Nucci, F., Pozzolo, A.F., 2010. The exchange rate, employment and hours: What firm-level data say. Journal of International Economics 82, 112-123. Park, A., Yang, D., Shi, X., Jiang, Y., 2010. Exporting and Firm Performance: Chinese Exporters and the Asian Financial Crisis. Review of Economics and Statistics 92, 822-842. Schwab, K., Sala-i-Martín, X., 2015. Global Competitiveness Report, 2014-2015: World Economic Forum. Stafford, T.M., 2015. What Do Fishermen Tell Us That Taxi Drivers Don’t? An Empirical Investigation of Labor Supply. Journal of Labor Economics 33, 683–710. Wooden, M., Warren, D., Drago, R., 2009. Working Time Mismatch and Subjective Well-being. British Journal of Industrial Relations 47, 147-179. Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel Data, 2 ed. Cambridge, Mass: MIT Press. Yang, D., 2006. Why do migrants return to poor countries? Evidence from Philippine migrants' responses to exchange rate shocks. The Review of Economics and Statistics 88, 715-735. Yang, D., 2008. International Migration, Remittances and Household Investment: Evidence from Philippine Migrants’ Exchange Rate Shocks*. The Economic Journal 118, 591-630.
24
Table 1: Summary statistics by gender
Age (years) Length of stay (years) English speaking background country (%) Bachelor degree or higher (%) Single (%) Disable (%) Employed (%) Weekly working hours (hours) Hourly wage rates (AUD/hour) Weekly wages (1000 AUD) Annual wage working hours (hours) Annual wages (1000 AUD) Hours desired to work more (hours/week) Full-time employed (%) Part-time employed (%) Self-employed (%) Occupation change (%) Number of observations
Female
Male
(1) 43.37 24.30 44.09 25.29 12.72 18.05 63.42 32.81 20.31 0.40 1531.33 18.85 -1.76 34.89 26.22 9.44 20.09 13521
(2) 43.69 24.66 51.60 25.92 17.98 18.51 81.39 42.74 22.35 0.75 2258.07 34.07 -2.77 69.70 9.28 14.22 16.27 12241
All immigrants (3) 43.52 24.47 47.66 25.59 15.22 18.27 71.96 38.01 21.38 0.56 1912.68 26.08 -2.29 50.90 18.43 12.01 18.05 25762
Male-Female (4) 0.32** 0.36** 7.51*** 0.63 5.26*** 0.46 17.97*** 9.92*** 2.03*** 0.35*** 726.74*** 15.22*** -1.01*** 34.81*** -16.95*** 4.78*** -3.82***
Notes: Sample of individuals age 16-64. Working hours, earnings, and “hours desired to work more” are conditional on working. All labour market outcomes, except the employment decision, are adjusted for CPI, using the 2001 CPI as the base. Tests are performed on the significance of the difference between the sample mean for male and female immigrants. The symbol *denotes significance at the 10% level, **at the 5% level, and ***at the 1% level. See Table A2 in the online appendix for detailed statistics by country.
25
Table 2: Impact of exchange rates on labour market outcomes - Pooled sample Pooled results Dependent variable Employed
Weekly working hours (log)
Hourly wage rates (log)
Weekly wages (log)
Annual working hours (log)
Annual wages (log)
Desired working hours
Worked full-time
Worked part-time
(1)
(2)
FE results
(3)
(4)
(5)
(6)
(7)
(8)
-0.01
-0.01
-0.01
-0.01
0.02
0.02
0.02
0.03
[0.02]
[0.02]
[0.02]
[0.02]
[0.02]
[0.02]
[0.02]
[0.03]
{25762}
{25762}
{25762}
{25762}
{25762}
{25762}
{25762}
{25762}
-0.08**
-0.08**
-0.08**
-0.13***
-0.06*
-0.06*
-0.06*
-0.08**
[0.03]
[0.03]
[0.03]
[0.04]
[0.03]
[0.03]
[0.03]
[0.03]
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
-0.04
-0.04
-0.04
-0.04
-0.02
-0.02
-0.02
-0.03
[0.05]
[0.05]
[0.05]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
-0.12**
-0.12**
-0.11**
-0.17***
-0.08**
-0.08**
-0.08**
-0.11***
[0.05]
[0.05]
[0.06]
[0.06]
[0.04]
[0.04]
[0.03]
[0.04] {15746}
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
{15746}
-0.15***
-0.15***
-0.14***
-0.17***
-0.09
-0.10*
-0.09
-0.04
[0.04]
[0.04]
[0.04]
[0.06]
[0.06]
[0.06]
[0.06]
[0.08]
{14841}
{14841}
{14841}
{14841}
{14841}
{14841}
{14841}
{14841}
-0.17**
-0.17**
-0.16**
-0.17*
-0.09
-0.09
-0.09
-0.08
[0.07]
[0.07]
[0.08]
[0.09]
[0.07]
[0.07]
[0.07]
[0.08]
{16645}
{16645}
{16645}
{16645}
{16645}
{16645}
{16645}
{16645}
0.64
0.62
0.62
1.03
0.73
0.77
0.46
0.21
[0.94]
[0.93]
[0.94]
[0.94]
[0.90]
[0.87]
[0.89]
[0.90]
{16279}
{16279}
{16279}
{16279}
{16279}
{16279}
{16279}
{16279}
-0.05*
-0.05*
-0.05*
-0.08**
-0.06*
-0.06*
-0.05
-0.06*
[0.03]
[0.03]
[0.03]
[0.04]
[0.03]
[0.03]
[0.03]
[0.03]
{23533}
{23533}
{23533}
{23533}
{23533}
{23533}
{23533}
{23533}
0.03
0.03
0.03
0.06*
0.07**
0.07**
0.07**
0.08**
[0.03]
[0.03]
[0.02]
[0.03]
[0.03]
[0.03]
[0.03]
[0.04]
{23533}
{23533}
{23533}
{23533}
{23533}
{23533}
{23533}
{23533}
0.02
0.02
0.02
0.04*
0.03
0.03
0.03
0.04*
[0.02]
[0.02]
[0.02]
[0.02]
[0.02]
[0.02]
[0.02]
[0.02]
{18556}
{18556}
{18556}
{18556}
{18556}
{18556}
{18556}
{18556}
0.05*
0.05*
0.04
0.03
0.05
0.05
0.04
0.03
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.03]
[0.04]
{15429}
{15429}
{15429}
{15429}
{15429}
{15429}
{15429}
{15429}
Year and month dummies
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Country dummies
Yes
Yes
Yes
Yes
N/A
N/A
N/A
N/A
Self-employed
Changed occupation
GDP growth (%) Unemployment rates (%)
Yes
Yes Yes
Region-year dummies
Yes Yes
Individual FE
Yes Yes
Yes
Yes
Yes
Notes: Pooled results (columns from 1 to 4) are from the regression (1) while FE results (remaining columns) are from the regression (2). Estimates are obtained from separate regressions. Other explanatory variables include age and its square, education, marital status, health status, the number of co-residing members of various age cohorts, the regional unemployment rate, regional relative socio-economic advantage index, state dummies, and year and month dummies. Pooled regressions also include gender, ESB, duration of stay in Australia, migration cohort dummies, and home country fixed effects. Number of observations is in curly brackets. Robust standard errors clustered at the country of origin level in parentheses. *** p
