SEX-SPECIFIC RECIDIVISM ESTIMATES IN NIGERIAN CORRECTIONAL SERVICE USING ZELTERMAN AND CHAO APPROACHES ON ZERO-TRUNCATED COUNT DATA

Abstract

In Nigeria, rising crime rates and increasing inmate populations reveal the need
for credible estimation of repeated offenders, particularly where complete
enumeration is difficult. This study estimates the population size of recidivists
in Nigeria using capture–recapture modelling techniques. Inmate population
records obtained from the Nigerian Prison Service were analyzed using zerotruncated count models, specifically the Poisson, geometric, and discrete
Lindley distributions. Both aggregate and covariate capture–recapture
approaches were applied to estimate the hidden population of recidivists. Model
performance was evaluated using standard error criteria. Results indicate that
the Zelterman-DLD estimator demonstrated the best performance, yielding the
smallest standard errors for both aggregate and covariate data. The estimated
inmate population was 22,439.79 (95% CI: 22,092–22,787) with SE=177 for
males and 628.35 (95% CI: 569–688) with SE=30.48 for females. The aggregate
inmate population was estimated at 23,066.95 (95% CI: 22,715–23,419) with
SE= 179.54. Findings further show that covariate-based estimation can serve as
a viable alternative to aggregate data in estimating hidden populations within
the capture–recapture framework. These results provide empirical support for
the application of zero-truncated count models in criminal justice research and
correctional system management in Nigeria.