Abstract | Incremental Dynamic Analysis (IDA) is a widespread approach to evaluate the seismic response of structures by means of nonlinear dynamic analysis. It employs a single set of ground motion records and scales them up until dynamic instability is reached.The objective of IDA is to obtain the statistics for different Engineering Demand Parameters (EDPs) that gauge the response of the structure. It is well known, however,that the earthquake scenarios dominating the site hazard differ with the intensity of the ground motion and, therefore, the spectral shape of the expected records should reflect this change. Hence, the response of the structure, in general, would be better estimated in a multiple stripe analysis (MSA) framework by multiple sets of records selected to be hazard consistent to different intensity levels via, for example, the conditional spectrum (CS)method. Despite this issue, IDAis still a standard tool in seismic design and assessment. This study tries to quantify the level of bias that derives from using a single set of randomly selected records in IDA versus a multiple set of CS-based hazard consistent records in a MSA setting. We further explore different alternatives of CS for record selection to implement in IDA including a ‘multi-level’ approach,which combines the seismic properties of multiple intensity levels in a single record set. This proposed approach provides a trade-off between the more accurate but more complex method of MSA versus the conceptually and practically simpler IDA.