An analysis of votes cast in the Georgia lieutenant governor’s race this past November suggest the results certifying Republican Geoff Duncan as the winner “are in substantial doubt,” according to a professor of statistics at the University of California-Berkeley.

Philip Stark, associate dean of UC-Berkeley’s Division of Mathematical and Physical Sciences and an expert in election statistics and post-election auditing, said he discerned a noticeable disparity in Duncan’s contest with Democrat Sarah Riggs Amico between undervote rates of paper ballots and those cast on the state’s obsolete electronic voting machines. Stark said his analysis—included in an affidavit filed in a pending legal challenge to Duncan’s election—“strongly suggests that malfunction, misconfiguration, bugs, hacking, or other error or malfeasance caused some [direct-recording electronic] voting machines not to record votes in the lieutenant governor’s contest.”