19.09.2013 Views

Perspektiv på välfärden 2004 (pdf) - Statistiska centralbyrån

Perspektiv på välfärden 2004 (pdf) - Statistiska centralbyrån

Perspektiv på välfärden 2004 (pdf) - Statistiska centralbyrån

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

ered on three occasions, and there are, hence,<br />

eight years between every point of measurement<br />

(for information about the dataset, see below).<br />

The point of departure for the empirical analysis<br />

is taken in figure 1. Three blocks of factors are<br />

distinguished. The first block contains year of<br />

birth, gender and class of origin. These aspects are<br />

exogenous, internally unrelated and also invariant<br />

over time. The rational for including gender and<br />

class of origin is discussed above, but that is not<br />

the case for year of birth. There are, even though<br />

only a very narrow age span is included in the<br />

analysis, probably significant age differences in<br />

the population at the first point of measurement,<br />

and the only reason for including age is to keep<br />

these differences under control. We do not expect<br />

age to have any long-term impact, but we do expect<br />

that age will affect the situation at t 0 . The<br />

second block in Figure 1 contains information<br />

about labour market affiliation, education and<br />

household formation at t 0 , i.e., these factors are<br />

assumed to capture some of the most important<br />

characteristics that, as discussed above, distinguish<br />

between different categories of youth. The<br />

second block is assumed to be dependent on the<br />

first block; labour market affiliation is, for example,<br />

assumed to be influenced by age, gender<br />

an00d, of course, class of origin. We assume that<br />

the characteristics within block two are internally<br />

related. However, no assumptions about causation<br />

are made. Hence, we are, for example, not assuming<br />

that people have left the nest because they are<br />

employed or that they study and therefore are not<br />

employed. The obvious reason is that the direction<br />

of causation in both cases could be the opposite,<br />

i.e., some people will study because they cannot<br />

find a job and some will work because they left<br />

the nest.<br />

The last block consists of indicators of economic<br />

circumstances and changes over time in<br />

these circumstances, i.e., the basic outcome variables.<br />

What will be measured are income level<br />

and the incidence of economic difficulties at t 0 .<br />

We 0assume that income level and economic<br />

difficulties covary, but we do not assume a causal<br />

relationship. It is further assumed that change in<br />

income and change in economic problems over<br />

time are affected by the starting level at t 0 . Finally,<br />

we suppose that change in income relates to<br />

change in economic difficulties; that is, if the<br />

income increases we believe that the incidence of<br />

economic problem will decrease, but we do not<br />

assume a unidirectional causal relationship, as it<br />

could be the case that economic problems lead<br />

people to increase their earnings. The model facilitates<br />

an analysis of the impact of class, gender<br />

and other vital circumstances on the economic<br />

situation during the youth period itself and the<br />

Youth<br />

development over time as the transition into<br />

adulthood is made.<br />

Data, method and operationalizations<br />

Data sources<br />

The analyses are based on data from the annual<br />

Swedish Survey of Living Conditions (ULF) (Häll<br />

and Vogel 1997). ULF is based on a representative<br />

sample of the Swedish population aged 16 to<br />

84. In 1979, a panel approach was introduced and<br />

since then a sub-sample has been re-interviewed<br />

every eighth year. Thus, it is now possible to analyse<br />

a three-wave panel covering a period of 16<br />

years. To maximize the available sample size,<br />

data from 1979, 1980, 1981 and 1982 are merged.<br />

The panel samples from these four years were reinterviewed<br />

a second time in 1986, 1987, 1988,<br />

1989 and 1990 and a third time in 1994, 1995,<br />

1996, 1997 and 1998 2 . The average non-response<br />

rate was 14.4 per cent during 1979-82, increased<br />

to an average of 21.0 per cent for 1986-90 and<br />

21.3 per cent during 1994-98. For the sake of<br />

convenience, the first panel wave will be labelled<br />

t 0 , the second t 1 and third t 2 . The working sample<br />

has been limited to those who were between 19<br />

and 25 years old at the first wave, and accordingly<br />

between 35 and 41 years old when data for the<br />

third wave were collected. The working sample is<br />

made up of 1,435 respondents, representing those<br />

who participated at all three waves.<br />

Operationalizations of variables<br />

The main outcome variables are income and an<br />

indicator of economic difficulties. Income is<br />

measured as the individuals’ income before tax<br />

recalculated to 1998's prices. Hence, what is analysed<br />

is the overall capacity to earn an income<br />

regardless of income source and taxation. The<br />

income measure should not be viewed as a direct<br />

indicator of economic well-being. Taxes will of<br />

course impact the amount of money left for consumption,<br />

and we do not know the household’s,<br />

i.e., the consumption unit’s, income or size. By<br />

focusing on the0 individual income, we do not<br />

risk blurring the analysis by the substantial<br />

changes in disposable income and, even more so,<br />

in equivalent disposable income that are caused<br />

by changes in household composition. The distribution<br />

of the income variable is shown in table 1.<br />

Instead of using income, direct information<br />

based on survey questions will be used as an indication<br />

of economic standard, or to be more cor-<br />

2<br />

The sample from 1979 is about twice as large as the<br />

latter sample, and the panel is split so that half the<br />

sample were re-interviewed in 1986 and the other half<br />

in 1987.<br />

71

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!