The adverse effect of the COVID-19 pandemic (and associated lockdown) on mental health, caused by increased financial and health-related anxiety, is now widely acknowledged. There is however less clarity regarding the implications of the easing of lockdown restrictions on mental health. The easing of lockdown restrictions is expected to reduce the financial stressors, provided that employment and income respond positively to the easing of lockdown. On the other hand, the increase in economic activities may elevate the subjective risk of contracting the virus thereby exacerbating health-related concerns. Using waves 2 and 3 of the NIDS-CRAM data, this paper examines the changes in the distribution and risk of screening positive for depressive symptoms in South Africa as the COVID-19 lockdown restrictions are relaxed from level 3 to 1. We also decompose the change in the proportion of respondents that screen positive for depressive symptoms between the two periods using Oaxaca-Blinder decomposition. This method decomposes the difference in proportion into components that are attributable to differences in characteristics (composition effect) and the difference in coefficients (return effect). Our analysis shows that, despite the easing of lockdown restrictions, both the distribution and the risk of screening positive for depression has increased in wave 3, relative to wave 2. The decomposition analysis shows that an overwhelming part of this change is explained through returns to covariates (or behavioural response). Furthermore, detailed decomposition shows that the returns to and changes in the composition of variables that are positively correlated with low social-economic status explains the increase in depressive symptoms, despite the easing of lockdown restrictions.