Introduction
There are four main reasons why a young person may leave home:
Because information about reasons 1 to 3 are all found in the 2015 GHS, one may determine (a) if all reasons exist and (b) if any reason exists. It would have been desirable to integrate reason 4 in the same way, but this is not possible since information on it is drawn from a different data set.
None of the first three reasons are decisive in determining whether a young person leaves home. A young person may be employed and remain at home. A young person who has ever married or cohabited may become single again and live at home or, more rarely, married and cohabiting people may remain at home. Some of these young people may have been absent from home for a spell, but these departures are treated here as excursions. A young person may have lost both parents, but may be living with grandparents or great grandparents.
The analysis here is what demographers call a ‘period’ analysis. It does not represent the experience of any actual cohort, but of a hypothetical cohort experiencing at successive ages the conditions in 2015 (or 2016 in the case of migration).
Analysis
The analysis will be conducted using survival curves against age. The survival curves are as follows:
Survival curves 1 to 4 for either gender are plotted in Figure 1, and survival curves 5 to 7 are plotted in Figure 2.
Figure 1 displays the survival curve until each individual event occurs: employment, entry into union, loss of both parents and migration since the age of 15. Figure 2 displays survival to the first event (excluding migration) as the “Any” curve, survival until all three events have occurred (the “All” curve), and survival at home until leaving (the “Stay” curve). Figure 2 confirms the statement in the introduction that none of the three events are decisive in causing young people to leave home, since the “Any” curve lies below the “Stay” curve.
A further investigation involves the use of two conditional probabilities. Given two states A and B, the conditional probability is defined as the probability that A occurs given that B occurs. Essentially, the conditional probability ignores what happens if B does not occur.
The table in the Appendix, which sets out the relationship between events and whether young people have stayed at home or left, indicate that the state (excluding migration) most closely correlated with departure is whether the young person has ever been in a union. It is clearly superior for women and as good as employment for men. So the first conditional probability to be considered is the probability of having left, given that the person has ever been in a union. This conditional probability is graphed along with the unconditional probability of having left in Figure 3.
The gaps between the conditional and unconditional probabilities indicate the influence of ever having been in a union on leaving. The effect starts at a younger age for women and is stronger than the effect for men up to the late 20s.
Finally, one may ask what the effect of employment is on being, or ever having been, in a union. The relevant conditional probability is the probability of ever having been in a union given that the person is in employment. Figure 4 sets out conditional and unconditional probabilities.
Figure 4 shows that employment improves the probability for men, but not for women.
A couple of final points:
Summary and conclusions
The analytical findings can be summarized as follows:
Charles Simkins
Head of Research
charles@hsf.org.za
APPENDIX
The relationship between events and staying or leaving 

Stay 
Leave 
Total 

Men 

Not employed 
82,0% 
18,0% 
100,0% 

Employed 
36,3% 
63,7% 
100,0% 

Women 

Not employed 
71,0% 
29,0% 
100,0% 

Employed 
38,9% 
61,1% 
100,0% 

Men 

Never in union 
73,3% 
26,7% 
100,0% 

Ever in union 
20,9% 
79,1% 
100,0% 

Women 

Never in union 
79,8% 
20,2% 
100,0% 

Ever in union 
15,0% 
85,0% 
100,0% 

Parent(s) not lost 
65,0% 
35,0% 
100,0% 

Both parents lost 
47,9% 
52,1% 
100,0% 

All conditions not met 
65,1% 
34,9% 
100,0% 

All conditions met 
100,0% 
100,0% 

No condition met 
86,4% 
13,6% 
100,0% 

Any condition met 
39,6% 
60,4% 
100,0% 
Notes:
[1] See Charles Simkins, What does the 2016 Community Survey tell us about internal migration? and What does the 2016 Community Survey tell us about immigration and emigration of the foreign born? both published on 12 April 2017