The question of interest groups influence is fundamental to our understanding and evaluation of political systems and processes. The definition and measurement of influence is however one of the most serious challenges to empirical studies of interest groups. Some argue that it is a mission impossible to observe the diffuse concept of influence. Nevertheless, more innovative methods have been developed to manage this challenge, but we do not know to what extent these different measurement methods agree, and therefore to what extent we can compare different studies and accumulate knowledge about which groups are influential and why. This article argues that, if different measures of influence correlate, we can have more faith in our measures and perhaps by triangulation come even closer to measuring at least important aspects of interest group influence. Therefore, the article is set out to test measurement agreement across different measures of interest groups' influence. It focuses on studies of many cases (large-N studies) and uses measures based on (i) survey data and (ii) documentary data. The Danish Parliament is the empirical setting of the study. First, the article reviews different definitions of influence and outlines the definition of this study. Here, influence is understood as control over political outputs, such as bills or parliamentary debates. Second, the article discusses strengths and weaknesses of different measures of influence. Survey data can illuminate informal venues of influence but may be biased by strategic replies from the interest groups. Documentary data can track the impact of group activity on political outputs in a more unbiased manner but can only uncover formal venues of influence. Put together, these indicators may provide a more valid measure of influence. This, however, requires the indicators to correlate. The article tests the agreement between these two sources of data across three different measures: (i) group activity; (ii) agenda-setting influence; and (iii) legislative influence. Agreement is strongest with relation to activities and weakest with relation to legislative influence, which was also expected. In general, the analyses show that even though measurement agreement is low, it is promising for future studies that different measures of influence are strongly and significantly correlated. The article finds no clear indications of some group types being less 'honest' in their responses. We do not need to be very suspicious towards specific categories of groups, such as business for instance. The article does obtain results indicating that the formulation of response categories is very important to the answers we obtain and also important to the agreement between survey and documentary data. These findings implicate that measuring - aspects - of influence is not necessarily a mission impossible. As different measures correlate, we can compare different studies and thereby arrive at more general conclusions, and we may improve our studies by combining more measures of influence and also by paying even more attention to the design of the surveys we use.
