The conclusion concerns what the author believes will win votes. Whether a candidate would actually attempt to enact reform is not relevant, because the argument concerns the effects of supporting such reforms. One might be able to win votes simply from support.

There’s nothing flawed about drawing opposite conclusions if the evidence supports those differing conclusion. The evidence shows a discernible pattern in the northeast, but no discernible pattern elsewhere. So there’s nothing flawed about drawing different conclusions.

We have no reason to think the conclusion is based on a “small” sample. The sample includes the results of votes over elections in the past 10 years. This could be dozens of elections. We have no idea whether the sample is too small.

There’s a correlation between winning more votes in the northeast and supporting tax reform. The author assumes that the support for tax reform is what caused candidates to win more votes. This doesn’t have to be true. So we can’t conclude that supporting tax reform will attract additional voters.

There’s nothing inherently flawed about drawing a conclusion from data that’s ten years old. In any case, the argument isn’t based “solely” on data that’s ten years old. The data concerns elections throughout the last 10 years.