Association estimated	Standardized estimate	SE	p‐Value
Age 21 homelessness predicted by
Study cohort	−0.211	0.138	.127
Female (referent: male)	−0.020	0.041	.627
Age (years)	0.254	0.142	.074
Family socioeconomic status	−0.081	0.041	.050
Age 21 antisocial behaviour predicted by
Study cohort	−0.045	0.094	.632
Female (referent: male)	−0.289	0.024	<.0001
Age (years)	0.121	0.094	.197
Family socioeconomic status	−0.067	0.026	.009
Age 21 victimisation predicted by
Study cohort	−0.088	0.101	.381
Female (referent: male)	−0.296	0.026	<.0001
Age (years)	0.076	0.100	.450
Family socioeconomic status	−0.092	0.028	.001
Age 23 homelessness predicted by Age 21
Homelessness	−0.038	0.090	.673
Age 23 antisocial behaviour predicted by Age 21
Homelessness	0.037	0.063	.557
Antisocial behaviour	−0.009	0.044	.883
Age 23 violence victimisation predicted by Age 21
Homelessness	−0.018	0.067	.790
Violence victimisation	0.053	0.048	.268
Age 25 homelessness predicted by Age 23
Homelessness	−0.075	0.120	.531
Age 25 antisocial behaviour predicted by Age 23
Homelessness	0.007	0.085	.932
Antisocial behaviour	0.044	0.059	.464
Age 25 violence victimisation predicted by Age 23
Homelessness	0.120	0.090	.180
Violence victimisation	−0.097	0.067	.147
Correlations specified in the model
Age 21 homelessness with Age 21
Antisocial behaviour	0.185	0.051	<.0001
Violence victimisation	0.270	0.051	<.0001
Age 21 antisocial behaviour with Age 21
Violence victimisation	0.611	0.027	<.0001
Age 23 homelessness with Age 23
Antisocial behaviour	0.315	0.064	<.0001
Violence victimisation	0.404	0.061	<.0001
Age 23 antisocial behaviour with Age 23
Violence victimisation	0.714	0.027	<.0001
Age 25 homelessness with Age 25
Antisocial behaviour	0.415	0.068	<.0001
Violence victimisation	0.473	0.069	<.0001
Age 25 antisocial behaviour with Age 25
Violence victimisation	0.534	0.037	<.0001

Model fit statistics: χ ² (41, N = 2328) = 41.83, p = .4346, Comparative Fit Index = 0.999, Tucker‐Lewis Index = 0.999, Root‐Mean‐Square Error of Approximation = 0.003. Correlations among exogenous demographic characteristics are not estimated in the model; the observed correlations between these variables are taken into account by Mplus. SE = standard error. Homelessness, antisocial behaviour and violence victimisation (coded 0 = no, 1 = yes). Female (coded 0 = male, 1 = female).