We examine the gender gap in unemployment duration in a large developing economy that is characterised by high gender inequality in social and economic opportunities. Due to sensitivity concerns, such data are often released by authorities in the form of contingency tables, which consist of multiple categories of count data. To overcome this limitation, we estimate a Bayesian log-linear Poisson regression model and develop the Oaxaca-Blinder (O-B) decomposition based on multi-way contingency tables. Using data from the 2019 Iranian Labour Force Survey, our findings indicate that gender gap can be narrowed most substantially if women with university educations face the same labour market opportunities as men.