the European Community Household Panel
3.5 Results Descriptive analysis
3.5.3 Expanded models
The differences in the country coefficients could still allow us to use a pooled model for groups of countries where we include interaction dummies. If we were to use slope dummies for every single covariate in the model, this would increase the number of variables by a factor equal to the number of countries minus one. In a multinomial logit context, such a high number of variables easily leads to problems of parameters stability or of extremely large estimates and makes it much more difficult to handle the model. Additionally, a pooled model would impose restrictions in the structure of the errors between countries.
We find it preferable to estimate the same model for every country. The choice of covariates will be based on the most useful information included in the ECHP dataset and the review of empirical and theoretical review in the previous chapter.
Restrictions to the distribution of sensitive information for Germany, and slight differences in the survey, mean that some variables are sometimes not available for certain survey years or for significant covariates, eg potential income seems to play a significant role.37
For each country we estimated a dichotomous model, then a multinomial logit, and conducted a test to see whether the two categories (moving out without partner and moving out with partner) are indistinguishable. The Wald test for combining outcome categories rejects the null of equality of coefficients at the 10% in most of the males equations and in half of the females equation.38 We then computed the Hausman test of the assumption of the independence of irrelevance alternatives (IIA) for the omitted category of “leaving to unknown destination”
(typically representing a case where the youth is not interviewed after leaving home, so that we don’t know if they are living with a partner or not). The test always rejects the null.39 We have studied whether pooling males and females is acceptable. The answer is generally (but not always) negative, as the equality of coefficients is mostly rejected.
The fit of the models vary depending on the country and gender, and we shall discuss the quality of the results under the subsection for each single group of countries. The pseudo R2 are reported for each country at the bottom of the table with full estimation results (Table 3.22 and following ones, in the appendix).40 However we believe that what matters here are the estimates of the coefficients, ie ie their sign, significance, and particularly their importance, as reflected by
35 In a lim ited n u m b e r o f cases, the very low freq u e n cy o f som e ca te go rie s have m ade it ne ce ssary to drop or collapse
the absolute and relative magnitudes, and also how the coefficients compare country by country, as discussed below.
Southern Europe
Table 3.12 and Table 3.13 shows the results for men, using actual and potential income respectively, and Table 3.14 and Table 3.15 refer to women. Men’s labour income has a positive and often significant effect on departures. This effect is, as expected, stronger for departures into a partnership.
For men’s departures with a partner, we observe that the effect of not having a job is large and negative, and very large relative to the average risk of leaving home, with estimates ranging from minus one percentage point to minus 2.5 pet (for Portugal): not having a job always significantly reduces departures into partnerships for men.41 Interestingly, for men’s single departures and for both destinations for women, coefficients are generally negative but mostly not significant.
Family income has a negative effect on Italian and Spanish males moving out with a partner, but the effect is relatively small. Again this is a very interesting result, that we shall comment in the next section. Living in owned accommodation reduces the risk of departure in several cases, for example for single departures in Greece, Italy and Portugal. For females, we observe that more durables reduce the risk of departure in Italy and Portugal. Overall, the various indicators of family wealth seem to point towards a reduction in the risk of departures.
Potential income has mostly a positive effect, and a significant one. The effects are small in absolute value, although relative to the average risk of departure these can be sometimes important if we compare individuals one or two quartiles away in the income distribution. It is interesting to note that the coefficients on potential income tend to be larger and/or more significant then the coefficients on actual income. For Spanish and Portuguese men moving out alone, the coefficients are indeed positive and significant, while both income and employment status covariates were not significant in the specification with current income. This is true for women, too, as in the case for Greece (alone and married), Spain (both destinations). From a qualitative point of view, what is the importance of these effects? Those children who would have the highest earnings potential if they were in a full-time job are leaving the parental home faster, and a move up two quartiles in the potential income distribution results in up to 1.5-2.0 percentage points increase in departure risk. Clearly, it is not possible to move a given individual across the potential income distribution, however it is interesting to notice that high income potential is related to early departure from home (given the way we construct potential income, this would indicated that the relation runs from potential income to departure, rather than the other way round). Overall, we observe that the effects of current income and potential income are different - disregarding at this stage any issues of identification and measurement,42 the reasons for such difference may be linked to the fundamental difference in what the two
41 Expressed e q u iv a le n tly , being employed at the tim e o f the su rve y in crea se s m e n ’s d e p artu re s into p a rtnerships d e stin a tio n s by a la rge and positive coefficient. Note th a t own in co m e is rarely in higher quartiles, e sp e cia lly for co u n trie s in the “ N o rth ” .
42 See co rre la tio n c o e ffic ie n ts reported in section 3.4.3.
covariates are measuring: current income includes transitory, random factors, and in any case does not represent the "permanent income" of the young person.
Table 3.12 - Multinomial model o f departure from the parental home -Southern Europe - Males, current income - Marginal effects o f selected covariates
GREECE GREECE ITALY ITALY SPAIN SPAIN PORT. PORT.
Alone Partner Alone Partner Alone Partner Alone Partner Own income -0.00005 0.000043 0.00011 0.000046 0.000036 0.000118 0.00001 0.000317
[0 5 7 ] [0.48] [2.28]** [1.23] [1.33] [2.29]** [0.02] [3.02]***
Zero income -0.007121 -0.00619 0.002705 0.001919 0.000251 -0.004379 -0.00384 -0.001852 Dummy [1.50] [1.02] [0.70] [0.64] [0.13] [1.18] [2.55]** [0.18]
No Job -0.002151 -0.014675 -0.006371 -0.011479 -0.000475-0.008891 0.000244 -0.024439 [0.37] [1.83]* [1.34] [3.27]*** [0.21] [2.15]** [0.15] [1.89]*
Parental income 0.000028 -0.000014 0.000038 -0.000071 -0.000004 -0.000069 0.00002 -0.000078 [0.46] [0.20] [1.01] [1.99]** [0.16] [1.74]* [1 04] [0.84]
Durables 0.001722 0.003608 0.000959 -0.000239 0.000012-0.000512 -0.000277 0.001103 [1.16] [2.23]** [1.24] [0.31] [0.03] [0.66] [0.60] [0.61]
House own -0.009613 -0.001997 -0.003736 0.001203 0.003167 0.003923 -0.002648 0.004136 [2.65]*** [0.40] [1.67]* [0.58] [1.32] [1.19] [2.05]** [0.73]
Rooms per -0.00167 -0.004337 0.001441 -0.003716 -0.000682-0.001821 0.000253 -0.006853 person [0.54] [1.05] [0.69] [2.14]** [0.79] [0.99] [0.24] [1.53]
Mobility 0.001185 0.010129 0.001969 -0.00138 0.002618-0.000768 -0.000116 0.002373 [0.31] [2.26]** [0.61] [0.48] [1.91]* [0.26] L0.08] [0.30]
Source: Own an a lysis on E C H P data. A b so lu te value o f z sta tis tic s in b rackets * sig n ifica n t at 10%; ** sig n ifica n t a
*** sig n ifica n t at 1%. A g e range is 17-34. S a m p le sizes: M en: G ree ce 2858; Italy 6152; Spain 5289; P ortugal 3429.
Table 3.13 - Multinomial model o f departure from the parental home - Southern Eurof.
Males, potential income - Marginal effects o f selected covariates
GREECE GREECE ITALY ITALY SPAIN SPAIN PORT. PORT.
Alone Partner Alone Partner Alone Partner Alone Partner Potential income -0.000023 0.0004 0.00037 0.000529 0.000097 0.000118 0.000138 0.00053 [0.23] [3.82]*** [0.91] [0.82] [1.94]* [1.16] [2.79]*** [2.76]***
Parental income 0.000068 -0.000095 0.00001 -0.000093 -0.000009-0.000095 0.000022 -0.00017 [1.02] [1.05] [0.24] [2.58]*** [0.34] [1.78]* [1.66]* [1.49]
Durables 0.000865 0.004675 0.001031 -0.000123 0.000062 -0.000343 -0.000985 -0.00113 [0.50] [2.44]** [1.25] [0.16] [0.13] [0.36] [2.03]** [0.50]
Rooms per -0.001526 -0.008077 0.000759 -0.003152 -0.000581 -0.001365 -0.000139 -0.01164 person [0.40] [1.54] [0.35] [1.88]* [0.58] [0.57] [0.10] [1.93]*
Parental health -0.001294 -0.005259 -0.00219 -0.000964 -0.00055 0.005613--0.000037 -0.00341 [0.26] [0.92] [0.76] [0.55] [0.41] [2.02]** [0.03] [0.56]
Mobility 0.002791 0.013119 0.002825 -0.002547 0.002415 0.000159 ■-0.000614 -0.00727 [0.62] [2.52]** [0.86] [0.85] [1.49] [0.04] [0.30] [0.68]
Source: Own an a lysis on E C H P d a ta. A b so lu te value of z sta tistics in b ra cke ts * s ig n ifica n t at 10%; ** sig n ifica n t at 5%;
*** sig n ifica n t at 1%. A g e ra n g e is 17-34. S am ple sizes: Men: G reece 2762; Italy 5668; Spain 5100; Portugal 3321.
117
Table 3.14 - Multinomial model o f departure from the parental home - Southern Europe -Females, current income - Marginal effects of selected covariates
GREECE GREECE ITALY ITALY SPAIN SPAIN PORT. PORT.
Alone Partner Alone Partner Alone Partner Alone Partner Own income 0.000017 0.00028 0.000126 -0.000011 0.00001 -0.000022! 0.00001 0.000227
[0.13] [1.15] [2.31]** [0.09] [0.91] [0.16] [0.69] [1.14]
Zero income 0.001757 -0.01357 0.00458 -0.018222 0.00001 -0.0187551-0.000011 -0.01133 Dummy [0.23] [1.00] [1.02] [2.51]** [1.04] [2.70]*** [0.62] [0.90]
No Job -0.00476 -0.01644 0.001446 0.008412 -0.000017’ -0.00817 -0.000014• -0.00095 [0.60] [0.89] [0.26] [0.86] [2.19]** [0.90] [0.65] [0.05]
Parental income 0.000084 -0.00019 -0.000014 -0.000029 0.00001 -0.000131 0.00001 0.000251 [1.14] [1.18] [0.34] [0.34] [0.61] [1.31] [0.21] [1.67]*
Durables -0.00084 -0.00229 -0.001923 -0.004095 0.000001 -0.00112 -0.000003i -0.00915 [0.68] [0.75] [1.91]* [2.26]** [0.52] [0.66] [0.61] [3.04]***
Parental health 0.004422 0.016285 -0.000177 -0.001887-o.oooooeI 0.005022 -0.000004 0.02186 [0.87] [1.42] [0.07] [0.38] [0.77] [0.99] [0.48] [2.91]***
Unemployment -0.0002 -0.00248 0.000109 0.00105 -0.000001 0.000595 0.000004 0.001831 Rate [0.26] [1.40] [0.69] [2.91]*** [1.43] [1.31] [2.85]*** [1.19]
Change in Un. 0.000947 -0.00532 -0.003091 0.001877 -0.000001 0.001314 0.000005 -0.00394 [0.26] [0.73] [2.21]** [0-77].... [0.15] [0.60] [0.75] [0.79]
Source: O wn an a lysis on ECHP data. A b so lu te va lu e o f z sta tistics in brackets * sig n ifica n t at 10%; ** sig n ifica n t c
*** sig n ifica n t at 1%. A g e range is 17-34. S a m p le sizes: W o m e n : G reece 2021; Italy 4947; Spain 4480; P ortugal 2
Table 3.15 - Multinomial model o f departure from the parental home -Southern Eurof Females, potential income - Marginal effects o f selected covariates
GREECE GREECE ITALY ITALY SPAIN SPAIN PORT. PORT.
Alone Partner Alone Partner Alone Partner Alone Partner Potential income 0.000159 0.000827 -0.000015 0.000015 0.000001 0.000259 0.000024 0.000282
[1.62] [3.73]*** [0.14] [0.06] [2.53]** [1.24] [0.58] [1.23]
Parental income 0.000103 -0.0002 -0.000006 -0.000089 0.00001 -0.0002 0.000029 0.000269 [1.39] [1.10] [0.15] [0.96] [0.21] [1.89]* [1.01] [1.71]*
Durables -0.001095 -0.00292 -0.001961 -0.004188 0.000011 -0.00198 -0.00108 -0.00988 [0.83] [0.88] [2.32]** [2.14]** [0.00] [1.09] [1.65]* [3.15]***
Parental health 0.005175 0.023087 -0.000034 0.000563 -0.000008 0.006018 -0.00025 0.024613 [1.01] [1.87]* [0.01] [0.11] [0.75] [1.11] [0.19] [3.03]***
Mobility 0.001663 -0.00982 0.001837 0.008954 0.000006 0.011281 0.000729 -0.00536 JO. 36] [0.781 [0.54] [1.26] [0.67] [1.65]* [0.40] [0.45]
S ource: Own a n a lysis on E C H P data. A b s o lu te va lu e o f z statistics in b rackets * s ig n ifica n t at 10%; ** sig n ifica n t at 5%;
*** s ig n ifica n t at 1%. A g e range is 17-34. S a m p le sizes: W om en. G reece 1930; Italy 4625; Spain 4313; P ortugal 2478.
Northern Europe
Overall, our results appear more diverse and mixed for the countries outside Southern Europe and Scandinavia. To some extent, the coefficients are less well determined when the estimation is conducted country by country, and the standard error are not small, so that the coefficients could be potentially large.
There are several significant coefficients, but it is difficult to find a clear pattern across countries.
The effect of own income is not as significant as for Southern Europe, exceptions being UK men moving out alone and Irish men moving into partnerships. Unexpectedly, we observe that having no job appears to increase departure rates in both Ireland and the UK, rather than decrease them as we found for Southern Europe.
Family income has a positive effect on UK and German males departures, while living in owned accommodation decrease risk in Ireland for both genders and destinations (and in several other groups).
High levels of unemployment appear to delay departures in the UK and Ireland (no regional data is available for Germany), and an increase in unemployment further reduces departures in both countries, particularly Ireland.
While the results for the UK and Ireland are consistent with each other and show a clear effect of income, employment and regional unemployment rates on departures, there appears to be no clear pattern across Northern countries, given that the results for the other countries are less well determined. The analysis for some of these countries appears to suffer from relatively small panel numbers, as young people tend to leave home faster and therefore drop out of the sample at risk, and one can also question the quality of the datasets in relation to confidentiality restrictions and high attrition rates.
Table 3.16 - Multinomial model o f departure from the parental home - Northern Europe - Males, current income - Marginal effects o f selected covariates ________________________________
G E R M A N Y G E R M A N Y B E LG IU M BE LG IU M UK UK IR E LA N D IR ELA N D
A lo n e P a rtn er A lone P artner A lone Pa rtn er A lone P artner
Own income 0.000004 0.000033 0.000001 -0.000025 0.000611 0.000207 0.000079 0.000005 [0.04] [0.47] [0.62] [0.22] [1.77]* [0.71] [0.75] [1.88]*
No Job -0.00138 -0.00714 -0.000002 -0.021573 0.04846 -0.02814 0.021805 0.000359 [0.20] [1.32] [0.28] [1.54] [2.15]** [1.52] [2.05]** [1.38]
Job security 0.00181 -0.00162 0.000001 0.001547 0.007355 -0.00574 0.004061 0.000061 [1.17] [1.42] [0.62] [0.75] [1.71]* [1.80]* [1.98]** [1.23]
Parental income 0.000193 0.000067 0.000001 0.000023 0.000645 -0.00011 0.000076 0.000003 [1.80]* [0.99] [0.62] [0.22] [2.29]** [0.41] [0.87] [1.13]
Durables -0.00145 0.001561 0.000001 -0.001477 0.00477 -0.00604 0.004401 0.000054 [0.64] [0.94] [0.43] [0.67] [0.73] [1.01] [2.21]** [1.00]
House own -0.01485 -0.00402 0.000004 -0.000008 0.008251 0.000237 -0.02776 -0.00048 [3.04]*** [1.21] [0.83] [0.00] [0.45] [0.02] [4.56]*** [1.89]*
Accommodation 0.007452 -0.00469 -0.000002 0.008494 0.034554 -0.00767 0.010099 0.000286 short of space [1.28] [1.01] [0.34] [1.43] [2.78]*** [0.64] [2.00]** [1.48]
Mobile 0.000001 0.002724 0.056047 0.028877 -0.01872 0.000002
[0.10] [0.38] [2.98]*** [1.02] [2.23]** [0.01]
Unemployment 0.000001 0.000118 -0.00707 0.005323 -0.02304 -0.00689 [0.65] [0.20] [1.75]* [1.49] [2.45]** [2.57]**
Change in Un. -0.000019 -0.012292 -0.01691 0.001586 -0.02001 -0.00491 [1.60] [0.78] [1.45] [0.15] [2.04]** [2.57]** Europe - Males, potential income - Marginal effects o f selected covariates
G E R M A N Y G E R M A N Y UK UK IR E LA N D IR E LA N D
Table 3.18 Multinomial model o f departure from the parental home Northern Europe -Females, current income - Marginal effects o f selected covariates______________________________________________________
G E R M A N Y Own income 0.000264 -0.00016 0.000074 -0.00043 0.000185 0.000423 -0.000064 0.000249
[1.42] [0.70] [0.36] [1.53] [1.36] [0.91] [0.36] [2.03]**
No Job -0.01471 -0.02446 0.001022 -0.05311 0.004311 0.000657 0.001546 0.000976 [1.07] [2.09]** [0.09] [2.00]** [0.65] [0.02] [0.13] [0.14]
Parental income 0.000275 -0.00021 -0.000036 0.000006 0.000038 0.000464 -0.000159 0.000027 [2.06]** [1.27] [0.42] [0.03] [0.49] [1.03] [1.24] [0.36]
House own -0.01038 -0.00354 -0.01229 -0.00797 -0.0083 0.02498 -0.016534 -0.01532 [1.48] [0.46] [2.70]*** [0.66] [1.95]* [1.08] [2.01]** [2.34]**
Mobile -0.01751 0.001925 0.00482 0.069146 -0.008246 -0.01879
[1.47] [0.14] [0.76] [2.36]** [0.91] [2.06]**
Unemployment -0.00078 -0.00074 -0.000082 0.002612 -0.03231 -0.03211 [1.08] [0.56] [0.09] [0.45] [2.45]** [3.91]***
Change in Un. -0.0115 -0.01188 0.003415 0.010202 -0.045877 -0.03107 [1.32] [0.53] [1.27] [0.58] [3.33]*** [3.41]*** - Females, potential income - Marginal effects o f selected covariates_____________
G E R M A N Y G E R M A N Y
Change in Un. 0.000423 -0.00435 0.000435 -0.0022 -0.05098 -0.05461 [0.04] [0.37] [0.57] [0.08] [3.20]*** [4.19]***
Nordic Europe
Due to high departure rates at young ages, the Nordic countries feature very limited sample sizes. We found that separate regressions yield very few significant coefficients and rather large standard errors, so we found it is necessary to use a pooled model, which turns out to be better determined.43
Own income is significant for single males and can yield relatively large effects: a move from the first quartile to the third quartile brings about an increase of the departure risk of almost 4 percentage points. Potential income is overall significant and the effects are larger: for men’s single departures the effect is nearly 50% larger, since a move in potential income of two quartiles now means slightly more than 5 percentage points; for departures into partnerships of both men and women potential income is now significant, while actual income was not - for women we see that a move higher in the distribution of potential income by 50 centiles brings about an increase in the departure risk of approx. 6.7 percentage points.
No significant effect is found for employment status and job security, clearly indicating that current labour market conditions may not be as important as potential ones.
There is some indication that parental income increases departure rates for partnership destinations, which could indicated financial transfers from parents to children setting up a partnership.44
Overall, our results are consistent with the limited literature on household formation for Nordic Europe, which showed little effect of current income on departures, but it is important to note that the use of potential income points to a much more significant role of income variables than previously thought.
43 T his m odel is an e xp a n d e d ve rsion o f the one presented in T ab le 3.9.
44 O th e r co va ria te s or co n tro l va riab le s are not p a rticu la rly in sightful and are relegated to the appendix.
Table 3.20 Multinomial model o f departure from the parental home Nordic countries pooled -Marginal effects o f selected coefficients
Actual income Potential income
Males Males Females Females Males Males Females Females
Alone Partner Alone Partner Alone Partner Alone Partner
Own income 0.000737 0.000085 -0.00007 0.000447 [3.73]“ * [0.51] [0.18] [1.03]
Potential income 0.00105 0.00054 -0.00022 0.001331
[2.32]** [1.91]* [0.76] [3.06]***
Zero income 0.026305 -0.01428 -0.01174 0.010005
Dummy [2.52]“ [1.43] [0.82] [0.57]
No Job 0.000222 -0.01006 0.001633 -0.01442
[0.02] [0.92] [0.08] [0.60]
Job security -0.00013 -0.00206 0.000725 0.004343
[0.05] [0.94] [0.19] [0.86]
Student -0.02637 -0.00953 -0.01176 -0.05712 -0.04405 -0.00939 -0.00484 -0.08163 [2.91]“ * [1.15] [1.11] [3.59]“ * [3.45]*** [0.97] [0.40] [3.81]***
Parental income 0.000146 0.000306 -0.00003 0.000474 0.000226 0.000316 -0.00038 0.000903
[0.96] [2.05]** [0.11] [1.32] [0.95] [1.69]* [1.22] [1.88]*
House owned 0.004949 0.004207 -0.00418 -0.04347 0.007572 -0.0005 -0.0213 -0.06223
[0.57] [0.52] [0.32] [2.89]*“ [0.55] [0.05] [1.30] [2.86]***
Mobile -0.00285 0.002488 0.021042 0.016787 0.000492 -0.00179 0.027693 0.028546
[0.29] [0.28] [1.79]* [0.99] [0.04] [0.20] [1.89]* [1.36]
S ource: Own an a lysis on EC H P data. A b so lu te value o f z sta tistics in brackets * sig n ifica n t at 10%; **
s ig n ifica n t at 5%; *** sig n ifica n t at 1%. A g e range is 17-29. S am ple sizes: A ctu a l incom e m odel: Men: 1880, W o m e n : 1325; po ten tia l in co m e m odel. M en: 1158, W o m en : 791.
3.6 Discussion
In this section we relate our results to the literature survey and discuss their significant in the context of our research objectives stated in the introduction and expanded in the final part of chapter 2.
The limited literature on Southern Europe, as surveyed in chapter 2, showed that being employed has a significant positive effect on departure. Many papers in the literature do not express their results as marginal effects, which makes it more difficult to compare results. In general, the literature finds that income is positively correlated with departure rates. Although our results for Southern Europe are generally consistent with that research, we do see that, by using marginal effects in the reporting of the results, the magnitude of the effects becomes much clearer compared to the results seen in the literature. Furthermore, we noted that in many cases the researchers did not distinguish clearly between income and employment. Our results for the South show that the effect of employment is significant and large relative to the small departure rates observed in the South, but it is still true that being employed increases the risk of departure by only around 1-2 percentage points, which can perhaps not be large enough in itself to be useful for, say, policy purposes. The effect of having a job is, as expected, stronger for departures into a partnership.45
The effect of income is small when we control for employment status, and we believe that the relatively strong effect of income reported in some literature (Aassve et al. (2001), amongst others) may be in fact reflecting the effect of employment.
For the rest of the ECHP sample, we find that the effects of employment and income are not always significant, but we do not seem to find very well determined results for some of the countries. For the UK and Ireland income does play a role, and for the Nordic countries we find a strong positive effect of income for men’s single departures, and the magnitude is clearly larger than for the coefficients estimated in Southern Europe.
We find some evidence that parental income has only a weak effect, but such effect is negative in the South and positive in the North. Previous research did also find no large effect of parental income, but the difference in sign has not been studied extensively. This is an important result that strengthens the findings of Manacorda and Moretti (2000) for Italy. Again, the effects are small in the South, and it is difficult to conclude that, overall, parental income plays an important role in affecting departures.
The difference in the size of the impact of income between the South and the North and the knowledge that current income may, in any case, not be the most important measure of income, prompted us to investigate the role of potential income. We find that potential income generally has a positive effect, although not always a large one. In the North potential income appears to be more important than in the South, where the significant coefficients for Spain and Portugal in fact represent very small effects. The fact that potential income has such a small or no effect in
45 S o m e tim e s, the s p e cifica tio n o f th e m odel is so d ifferent th a t the m a g n itu d e o f the e ffe cts ca n n ot be com pared. For exam ple, A a ssve e t al. (2001) use interaction q u alitative d u m m ie s to an a lyse Ita ly’s de p artu re rates and find th a t being e m p lo ye d and on a high sa la ry has a very large e ffe ct on d e p artu re rates. S in ce the au tho rs do not include age in the d e p artu re e q u atio n and b e ca use old e r Italians are m uch m ore likely to be em p lo ye d and on high incom es, the large e ffe ct fo u nd m ay in fa c t be ca p turin g m ostly an age effect.
3.6 Discussion
In this section we relate our results to the literature survey and discuss their significant in the context of our research objectives stated in the introduction and expanded in the final part of chapter 2.
The limited literature on Southern Europe, as surveyed in chapter 2, showed that being employed has a significant positive effect on departure. Many papers in the literature do not express their results as marginal effects, which makes it more difficult to compare results. In general, the literature finds that income is positively correlated with departure rates. Although our results for Southern Europe are generally consistent with that research, we do see that, by using marginal effects in the reporting of the results, the magnitude of the effects becomes much clearer compared to the results seen in the literature. Furthermore, we noted that in many cases the researchers did not distinguish clearly between income and employment. Our results for the South show that the effect of employment is significant and large relative to the small departure rates observed in the South, but it is still true that being employed increases the risk of departure by only around 1-2 percentage points, which can perhaps not be large enough in itself to be useful for, say, policy purposes. The effect of having a job is, as expected, stronger for departures into a partnership.45
The effect of income is small when we control for employment status, and we believe that the relatively strong effect of income reported in some literature (Aassve et al. (2001), amongst others) may be in fact reflecting the effect of employment.
For the rest of the ECHP sample, we find that the effects of employment and income are not always significant, but we do not seem to find very well determined results for some of the countries. For the UK and Ireland income does play a role, and for the Nordic countries we find
For the rest of the ECHP sample, we find that the effects of employment and income are not always significant, but we do not seem to find very well determined results for some of the countries. For the UK and Ireland income does play a role, and for the Nordic countries we find