We develop a new test of simultaneous versus sequential complementarities by using a discrete-time binary model for survival-type panel data on the adoption of one, both, or neither of two technologies while controlling for unobserved heterogeneity by instrumenting key variables. We examplify using data on the adoption of CAD and CNC technologies by US metalworking plants. Results indicate significant complementarities. Simultaneous adoption is in this case more likely than adoption of the two technologies in isolation. Sequential adoption is significant but in only one direction. Prior adoption of CNC has a positive effect on the posterior adoption of CAD. We find no evidence of the opposite.