Dissertation Submitted in Partial fulfillment of the requirements for the Degree of
Master of Science in Accounting and Finance (Msc A&F) of Mzumbe University
Microfinance institutions (MFIs) were established to fill the gap in the financial system in Tanzania that emerge following rejection of poor entrepreneur (marginalized group) by commercial banks in accessing debt capital. These MFIs provide an avenue whereby small and medium scale entrepreneur can now acquire capital to start or expand their business. Therefore researcher become interested to finds the impact of loan from MFIs on the performance of SMEs in Mbeya urban
This was the survey study covered 100 SMEs found in Mbeya urban particularly Uyole, Sido and Mwanjelwa markets. Researcher adopted cross-sectional design due to limitations of dataset obtained from the respondents. They claimed to have no records for past financial year and therefore researcher had to collect only data that was available during the time he was making survey for data collection exercise from SMEs managers/owners.
The study used both primary and secondary data, and in order to estimate impact of loans from MFIs on performance of SMEs researcher employed OLS regression because dataset contain continuous values and errors are independent and identically distributed (errors have equal variance), therefore according to Carter-Hill et al.(2001), suggest that, the model that contain continuous values as well as with residual that are uncorrelated and with equal variance can be correctly estimated by using the OLS regression analysis. Empirical findings of the study revealed that, loan has a direct impact on the performance of SMEs as the variables sales, assets and the number of employees has increased by applying a proxy of a natural logarithm of a variable debt (lndebts)
Following the results of the study it is high time for further research to be conducted on the study area by using panel data methodology because it allow controlling for unobservable heterogeneity of individual firms as well as makes it possible to exclude biases deriving from the existence of individual effects (Gujarat, 2004).