A Comparison of Micro-objective Non-parametric Statistical Matching Techniques for the Integration of Different Survey Data

Lack of information is often reported as one of the main study limitations in transport research. Other data sources provide an important opportunity, as the information from these data sources might be fully exploited when they are integrated using an appropriate mathematical framework. This paper determines which data integration techniques can be adopted in the context of road safety, and which criteria should be used to assess the quality of the integrated dataset. The paper contributes to the literature by providing a comparison of different techniques that allow the tabulation of the joint micro-data. To compare the suitability of different Statistical Matching techniques, data from the 2012 Belgium Road Safety Institute (BRSI) Attitude Survey and data from the 2010 Belgian National Household Travel Survey (BELDAM) are integrated. In particular, the socio-economic factor "Income" is being matched to the BRSI Attitude Survey data. The paper discusses and compares different non-parametric micro-objective Statistical Matching techniques, in which the “Income” variable from the BELDAM survey was added to the BRSI Attitude Survey. Based on similarity/dissimilarity measures, the Rank hot-deck Statistical Matching yielded the best result, which is in line with the existing literature. Moreover, from the results, it can be concluded that statistical data matching is a practically applicable technique, which can help overcome the lack of additional information. Further research should focus on using the integrated datasets with an attempt of expanding the scope of variables, and explicitly testing for omitted variable bias.