This study investigated the impact of the universal subsidy program on households in Mwandama millennium villages in the Zomba district of Malawi. This was done by comparing yield before the agriculture intervention and after the intervention. There is a clear yield difference between the before the intervention period and the after the intervention period. Additionally, there is a rising trend in yield from 2005/06 to 2009/10. These results show that the universal subsidy program has had a significant impact on sustaining the food security status in study area. It also sought to establish whether the interventions that the Millennium Villages Project (MVP) had been providing to the project site have had an effect on the food security status in the area .The difference between the means was used to statistically determine whether there was a difference between the mean yield in the treatment and control groups. Dzaone, a non-intervention area, was used as a control group while Mwandama research village (MV1) was the treatment group. MVP yield data was analyzed using STATA version 10.1 and the treatment size was tested using a t-test statistical model. It has been empirically revealed that the computed t-value was much higher than the observed t-value, signifying statistical significance. This means that the interventions that the MVP had a significant effect on the food security status in the area. This implies that in a country where land and labor are major constraints, it is important for the government to approach food security in an integrated way by providing all determinants of food security in a single package including modern inputs, investment in infrastructure (road, market and irrigation), agricultural extension activities, social safety net programs and investment in skills training and education for farmers. One of the major limitations to this study was the nature of the data which made it impossible to use regression analysis, an alternative method to independent t-test for significance levels. The data that was available for use was mean yield per village as opposed to yield per household which would have allowed for any econometric analysis.