The paper attempts to examine the economic effect of changing age structure in Bangladesh in the demographic dividend while being cautious of aggregation bias. One of the major contributions of this paper to the estimation of the demographic dividend literature is its use of a disaggregated dataset to produce a representative estimation of demographic dividend and compare among different education groups. The findings of the paper shed light on the debate on the sources of the first demographic dividend—whether this dividend comes from a pure age structure factor or represents an education dividend. When the economic profiles are disaggregated by levels of education, the Economic Support Ratio (ESR) decreases compared to the estimates when it only classified by ages. After estimating the first demographic dividend, the paper disaggregates the dividend into education effect and age effect using the Das Gupta decomposition technique. Results show that the size of the dividend is driven largely by age effects, while the education effect has been negative in Bangladesh for the past decades. The negative education effect indicates the aggregation bias in the estimates of support ratio if data is not disaggregated at that level.