On October 30, 1995, the citizens of Quebec went to the polls to decide the future of their province. They were asked to vote “Yes” or “No” on whether Quebec, a predominantly French-speaking province, should secede from Canada and become a sovereign country. The “No” side was declared the winner, but only by a thin margin. Immediately following the vote, however, allegations began to surface that the result was closer than it should have been. In particular, the ruling separatist Parti Québécois, whose job was to decide which ballots were rejected, was accused by the “No” voters of systematic electoral fraud by voiding thousands of “No” votes in the predominantly allophone and Anglophone electoral divisions of Montreal. Cawley and Sommers examined whether electoral fraud had been committed by running a regression, using data from the 125 electoral divisions in the October 1995 referendum. The dependent variable was REJECT, the percentage of rejected ballots in the electoral division. The explanatory variables were as follows:
• ALLOPHONE: percentage of allophones in the electoral division
• ANGLOPHONE: percentage of anglo phones in the electoral division
• REJECT94: percentage of rejected votes from that electoral division during a similar referendum in 1994
• LAVAL: dummy variable equal to 1 for electoral divisions in the Laval region, 0 otherwise
• LAV_ALL: interaction (i.e., product) of LAVAL and ALLOPHONE
The estimated regression equation (with t-values in parentheses) is
Predicted REJECT = 1.112 + 0.020 ALLOPHONE
(5.68) (4.34)
+ 0.001 ANGLOPHONE + 0.223 REJECT94
(0.12) (2.64)
- 3.773 LAVAL + 0.387 LAV_ALL
(-8.61) (15.62)
The R2 value was 0.759. Based on this analysis, Cawley and Sommers state that, “The evidence presented here suggests that there were voting irregularities in the October 1995 Quebec referendum, especially in Laval.” Discuss how they came to this conclusion.