Testing for complementarity when strategies are dichotomous

Open innovation: Are inbound and outbound knowledge flows really complementary?

“…Miravete and Pernias () argue that, even if we had found a positive correlation, we could not have concluded the existence of complementarity, as they claim it is not possible to extend Arora's () reduced form test for the existence of complementarity to evaluate the relationship between a couple of dichotomous strategies, because it might lead to an incoherent simultaneous discrete response model.…”

Complementarity and information technology adoption: Local area networks and the Internet

“…However, positive correlation in the error terms is neither a necessary nor sufficient condition for identifying complementarities (Arora, 1996;Athey and Stern, 1998). Moreover, as Miravete and Pernıas (2008) show, the result in Arora and Gambardella (1990) cannot be extended to the case of binary variables, because the model is incoherent as well.…”

Competitive Pressure and the Adoption of Complementary Innovations

“…No assumptions are made about these potentially complementary relations and our estimates will determine them regardless of whether the strategies are continuous, such as the scale, or discrete, as in the case 20 Arora and Gambardella (1990) first computed similar correlations to test for the existence of complementarity although the theoretical foundation of this test is due to Holmström and Milgrom (1994). Miravete and Pernías (2009) show that such a simple correlation analysis cannot capture the effect of complementarity when decision variables are, as in the present case, dichotomous. Indeed, these correlation measures will only capture correlation due to unobservable heterogeneity.…”