The analyses performed so far refer to the entire pool of immigrants in the sample. However, the descriptive statistics in Table 18 prove quite a degree of heterogeneity in terms, e.g., of migration histories and countries of origin. Hence, in Sects. 7.1 and 7.2 we disaggregate the overall effect of being an immigrant by cohort of arrival and country of origin, respectively, while in Sect. 7.3 we investigate any potential role of the different combinations of immigrant and natives within the couple. Lastly, in Sect. 7.4 we split the sample period between a pre- and a post-crises sub-period.Footnote 35
Heterogeneity by Cohort of Arrival
Italy has been subject to several waves of immigration, that displayed several differences in terms of economic motivation and family consideration. To capture these differences, we replace the dummy for the immigrant status of the household head with a set of dummies capturing cohorts of arrival.Footnote 36 We assign four dummies, for household heads who migrated before 1980, in the 80s, in the 90s, and in 2000 or after.Footnote 37 While few of the immigrant heads in our sample arrived before 1980 (as shown in Table 18, only 7%, that is only 127 observations) and in the 80s (6%, ie., 116 observations), about 31% (560 observations) arrived in the 90s and 56% (1,034 observations) since 2000.
In Table 9 we present results for the distribution of net wealth. As before, the omitted binary variable identifies households with a native head. The table reveals several patterns. For the pre-1980 cohort, the gap for immigrants is only significant at the median. For the 1980s cohort, the gap is significant at and above the median. The 1990s cohort displays a gap at all quantiles. For the post-2000 cohort, the most numerous one, no gap is detected below the 25th quantile, which suggests that among the lowest quantiles those immigrant families that migrated more recently are not significantly poorer than native ones. The observed heterogeneities across cohorts are broadly consistent with those obtained by Mathä et al. (2011) using only the 2008 wave of the SHIW.Footnote 38
Table 10 applies the same disaggregation to portfolio decisions, showing a variegated picture. For the decision to hold risky assets, the gap is driven by the behavior of the two cohorts that arrived after 1990. This is consistent with the results for the corresponding share. Home ownership reveals a gap for all cohorts, suggesting that immigrants hardly catch up with natives when it comes to buying a house. For mortgages, the first and last cohorts are driving the average negative result. Interestingly, while informal debt holdings were not significant for immigrants overall, now we can spot a significant gap for the 1980s cohort, who somewhat unexpectedly seems to be relying less than natives on such kind of debts.Footnote 39
Overall, heterogeneities across cohorts appear substantial, reflecting the distinct stages of the recent immigration history of the country.
Heterogeneity by Country of Origin
Immigrants from different countries may accumulate and allocate their portfolios differently, possibly to account for shocks in the source countries, or in response to distinct cultural backgrounds. In order to dig further in this direction, we estimate variants of models (1) and (2) where the immigrant dummy is replaced by a set of dummies reflecting an immigrant household head’s country of origin, grouped into seven aggregations (defined in detail in Table 17): EU15 and North America (with about 6% of the household heads), New EU (21%), Other Europe (27%), North Africa (15%), Sub-Saharan Africa (11%), Central and South America (7%), and Asia and Oceania (13%).
We start by considering the correlates of net wealth. Since disaggregated data are only available for the period 2006–2012, for the sake of comparison in Table 11, Panel A we first report a specification involving once again the household head’s immigrant status dummy, but now over the shorter time period. The results are in line with Table 2. Next, in Panel B, we present results by groups of countries, where again the omitted binary variable identifies households with a native head. Despite the fact that the low number of observations for each group tends to decrease the significance of the coefficients, we do observe some interesting heterogeneities. For instance, immigrants from EU15 and North America are not significantly poorer than natives, while for immigrants from other European countries the average pattern is essentially replicated. Immigrants from Central and South America and Asia and Oceania are indeed poorer, but only at and above the median.
In Table 12 we repeat the above analysis for asset holdings. Panel A replicates the specification with the immigrant status dummy over the period 2006-2012 and confirms the results in Table 4. In Panel B we replace the immigrant status dummy with the seven dummies for country groups. The emerging picture is variegated. For instance, the lower probability of holding risky assets for immigrants appears to be equally present in all groups of countries, even though it cannot even be estimated for sub-Saharan Africa, due to the low number of immigrants from that region holding such assets. Indeed, as shown in Table 18, only 10.8% of the sample, that is only 143 immigrant household heads, come from sub-Saharan Africa. By contrast, other results are driven by specific source countries. For instance, the lowest probability of being a home owner, if compared to natives, is observed for households with a head born in a EU new member country, possibly since many of them come as domestic helpers, followed by Sub-Saharan Africa and Asia and Oceania.Footnote 40
To sum up, the above results on wealth and asset portfolios do shed some light on the financial choices of people coming from different source countries and, even though their interpretation is sometimes difficult due to the very small number of observations, they indeed testify substantial heterogeneities by source country, with variegated consequences across different kinds of assets. A comparison with the literature is complicated by the fact that other host countries have very different compositions of the immigrant population, if compared to Italy. In the US, for instance, European and Asian households are often found to behave differently from those from Mexico and Central and South America. Overall, however, a great deal of diversity by source country is always present within the immigrant population.
To shed further light on the potential determinants of the observed heterogeneity, we investigate how country-specific preferences may explain it. To do so, we rely on a country-level dataset assembled by Falk et al. (2018) and based on data collected in 2012 through the Global Preference Survey in 76 countries. We focus on those preference traits that they report as particularly relevant for our outcomes of interest, that is, measures of time and risk preferences. One shortcoming of our approach is that we can only rely on country of origin information by broad country groups. For each country group, we construct average measures of patience and risk taking that we then assign them to each household head born in that group. We assign to natives the reported values for Italy. The results suggest that patience, rather than risk taking, is the economic preference trait behind several portfolio decisions. In particular, being born in a country group that in aggregate displays more patience is positively associated with a higher probability to invest in risky and foreign assets (Table 12, Panel C) and own a business and valuables (Table 32, Panel C).Footnote 41 Albeit these findings are merely suggestive, they point to the potential relevance of cross-country preference variation for economic and financial decisions.Footnote 42
The Influence of Spouses and the Role of Intermarriage
The results from the previous sub-sections focus on the immigrant status of the household head, consistently with his/her responsibility for the financial choices of the household. However, within a household, the primary decision maker may well be influenced by other family members, and especially by the partner within a couple. In particular, a couple can involve two immigrants, or else an immigrant and a native, or two natives. In case of a mixed couple, it may also matter whether the household head, as opposed to the partner, is the immigrant. To account for all the possible combinations and assess their influence on financial decisions, we focus first on a sub-sample of households including a couple. As explained in Sect. 4, over this sample we then define four dummy variables denoting households including a couple of natives (both natives), a couple of immigrants (both immigrants) or a mixed couple, further distinguishing whether the immigrant is the household head (mixed immigrant head) or the spouse (mixed immigrant spouse).Footnote 43
In Table 13 we present results for the distribution of net wealth. Preliminarily, in Panel A we present the immigrant status dummy alone as in Table 2 but now, for the sake of comparison with the following specifications, the regression is run over the sub-sample of households including a couple. If compared to Table 2, where all households are included, some differences do emerge. The gap in wealth with respect to natives is larger in size and is significant also at the 25th quantile, while it becomes insignificant at the 10th quantile. Since in this specification we cannot distinguish whether the immigrant household head has a native or an immigrant spouse, the observed effect is a weighted average of the effect of Mixed Immigrant Head and Both Immigrants.
In Panel B of Table 13 we can verify if the composition of a couple by immigration status does matter. The reference is a couple involving two natives. We show that, for couples where both partners are immigrants, the gaps in wealth captured by the immigrant status dummy are largely confirmed. However, when we look at mixed couples, we find that those with an immigrant head are not significantly different from natives apart from the lowest quantile (where the effect is marginally significant and positive), while those with an immigrant spouse are poorer than natives along the entire wealth distribution. In other words, the average effect displayed in Panel A is to be attributed not only to couples of immigrants, but also to mixed couples where the head is a native. In Panel C again we break the average effect displayed in Panel A in order to further distinguish whether, within a mixed couple, the gender of the head also matters, to reveal that the weaker position of mixed couples with a native head and an immigrant spouse is confirmed independently of gender considerations, even though the gap with respect to natives is much larger for couples with a male immigrant spouse, that is, couples with a female native head. Likewise, when the immigrant is the head, gender does not modify previous conclusions.
In Table 14 we repeat the above analysis for asset decisions. Panel A replicates Table 4 over the sub-sample of couples, yielding very similar results with the exception of the decision to hold a mortgage, where immigrant status no longer displays a significant coefficient.Footnote 44 In Panel B we replace the immigrant status dummy with the set of dummies capturing pure vs mixed couples, where a couple of natives is the omitted category. For couples including two immigrants, the results mirror those in Panel A. However, for mixed couples, the coefficients are never significant, with the only exception of the decision to hold risky assets when the immigrant is the head. In other words, the financial decisions of mixed households are largely indistinguishable from those of native households, independent of the immigrant status of the head or the spouse. One possible explanation for these findings is that, through intermarriage, immigrants might have gone through an assimilation process, prior and/or during marriage, that makes them more similar to natives even with respect to financial choices. However, this effect might not be precisely estimated due the very limited number of observations (only 195 mixed households in the sample). The distribution of household heads by gender, with a prevalence of males, may also be part of the explanation, as addressed in Panel C, where we observe different patterns across each investment decision. For instance, the lower participation in risky assets is explained, within mixed couples with an immigrant head, by those couple where the immigrant head is a male. With regard to home ownership, on the other hand, mixed couples where the immigrant spouse is a male actually outperform natives. Informal debts are significantly lower for mixed couples with a female immigrant head.
We can compare our results with those derived for other countries. For instance, for the US, Cobb-Clark and Hildebrand (2006a) focus exclusively on households including a couple and do not include mixed households among immigrant ones, since they expect them to behave like native-born households. Thus, they do not distinguish, as we do, between mixed households headed by an immigrant rather than a native. For Germany, Sinning (2007) adopts a classification similar to ours and finds that, in terms of portfolio diversification, pure immigrant households perform at the bottom, followed by mixed households with an immigrant head and mixed households with a native head.
To sum up, the results in this sub-section document complex interactions between the patterns of intermarriage, the responsibility of making financial decisions, and the gendered division of roles within the household. Moreover, these interactions are likely influenced by the cultural background associated with different source countries, as highlighted in the previous sub-section.
The Impact of the Great Recession
We now investigate whether the financial crisis has influenced how immigrant households behave if compared to native ones. To this end, the sample is split into two sub-samples, where 2006 and 2008 are interpreted as pre-crisis waves, while 2010, 2012, and 2014 are interpreted as post-crisis waves. The choice to assign 2008 to the pre-crisis sub-sample can of course be questioned. However, it can be defended on several grounds. First of all, responses for each survey wave are collected at the very beginning of the following year, that is, for 2008, in early 2009. Since the real effect of the crisis on GDP manifested itself, for the case of Italy, only in 2009, with a dramatic drop of 6%, it is reasonable to assume that survey respondents, at the beginning of 2009, had not yet perceived it. In other words, even though 2008 witnessed a turmoil in financial markets, culminating in September with the bankruptcy of Lehman Brothers, 2008 was not yet, at least for Italy, a recession year. The relative stability of the real economy as of 2008 is also confirmed by data on the rate of unemployment, which was then still at 6.7%, increased to 7.8% in 2009, and then continued its growth until 2014, when it reached 12.9%. Moreover, in Italy the banking sector showed a remarkable resilience, at least in the immediate aftermath of 2008, while the decline in house prices manifested itself only after the initial financial shock and developed very gradually.
In Table 15 we present quantile regressions for wealth, separately for the pre- and post-crisis sub-samples (the relevant term of comparison is the full sample in Table 2). While before the crisis immigrant status is associated with a non-significant gap at all quantiles, after the crisis wealth gaps are consistently larger, with significance levels that replicate those over the entire 2006–2014 time period, apart for the lowest quantile which is now insignificant, so that the effect in Table 2 is attributable to the after-crisis sub-sample. Thus, these results point to a worsening of the financial conditions of immigrant households after the crisis, relative to natives.
Turning to asset allocation, in Table 16 again we present separate regressions, for each outcome of interest, for the pre- and post-crisis sub-samples (with Table 4 as term of comparison). The results reveals that for some financial decisions the gaps for immigrants, if compared to natives, are relatively stable before and after the crisis.Footnote 45 This is the case, for instance, for the decisions concerning risky assets, which show similar coefficients over the two time periods. However, the gap in home ownership actually becomes smaller after the crisis, while it disappears for mortgages, possibly because native households as a consequences of the crisis also have reduced home ownership and mortgages. Informal debt holdings, on the other hand, become more likely for immigrants after the crisis.Footnote 46
To sum up, the financial crisis worsened the conditions of immigrant households, relative to native ones, in several dimensions, including wealth holdings and financial fragility. After the crisis, immigrants also appear to rely more on informal debt channels.Footnote 47
Even though the evidence we report is only descriptive, as it cannot capture a causal impact for the recession, our results are consistent with the assessment of the consequences of the crisis for Italian immigrants in Colombo and Dalla Zuanna (2019). It is useful to relate our results also to those obtained by Amuedo-Dorantes and Pozo (2015) by comparing 2006 and 2010 for US households. They find that post-crisis wealth losses for immigrants were particularly large for the middle and top wealth quartiles, which is broadly consistent with our findings. Moreover, they show that this outcome was driven by differences across assets, with greater losses in primary housing ownership and primary housing values. Again, this pattern broadly mirrors our results. However, it should be highlighted that, while the housing market crash in the US led the recession, as previously mentioned in Italy the decline in house prices manifested itself quite gradually after the initial financial shock.