By Berwood Yost

It was widely reported that Pennsylvania voters were switching to the Republican Party in large numbers. News reports at the end of March told us that nearly 215,000 voters switched parties in 2016 and most of them, about 125,000, had registered Republican.

This reporting fit nicely into the national storyline depicting high excitement among Republicans and the power of Donald Trump to attract new voters.

Turns out those numbers were wrong.

The Pennsylvania Department of State released revised numbers on April 3 showing that only 150,000 voters had switched parties, with 90,000 registering as Republicans.

The startling fact about these stories, both old and new, is that each provides an inaccurate and misleading picture of the state's voters.

The Department of State reports numbers of new and online voter registrations. Combining these with the party-switchers shows that there are now 222,000 new Democrats and 214,000 new Republicans.

Not much of a Republican surge, is it?

But while there is no GOP voter surge statewide, there is an important story that has been missed by almost everyone. Republican registration is surging, but only in some parts of the state.

In the southwest, large majorities of new voters in Greene, Fayette, Washington, and Westmoreland Counties have registered as Republicans. The same is true in the northeast, where large majorities of newly registered voters in Luzerne and Lackawanna Counties have registered Republican.

So there is a significant movement to Republicans among voters who live in areas well known for having a lot of "Reagan Democrats." These are Trump voters nationally, and they are likely to be Trump voters in Tuesday's Pennsylvania primary.

But these shifts in a few counties probably do not mean the Republican Party is substantially better positioned to win state-wide in November. That's because even after all this shuffling around, there are still about one million more registered Democrats than registered Republicans in the state.

And that matters. Today, fewer and fewer voters split their votes between the two parties, which means political party registration is the best way to understand how someone will vote.

The other missing part of the story is that there is no context for what these changes might mean. In fact, the shuffling between the parties we've seen this year has not changed the overall electorate as profoundly as did party switching in previous elections.

In 2008, for example, Democrats nearly doubled their voter registration advantage over the Republicans. In November 2007, Democrats had about a 640,000 voter registration advantage. By April 2008, that advantage had stretched to a million voters.

So neither the patterns in the current registration data nor the patterns we've seen compared with previous registration changes provide any evidence that a widespread change in voter alignment is taking place in the state, regardless of the news stories that say otherwise. It just isn't happening.

How could the media and their expert commentators be so mistaken? The answer could be the real "Trump effect": Members of the media were captured by the compelling national storyline about the Trump candidacy. Knowingly or not, they substituted what they knew was happening elsewhere for what they thought might be happening in Pennsylvania. That's a too-common mistake when many of us make quick judgments, particularly about those things that seem to confirm what we think we know.

But the media's essential function as our democracy's information gatekeeper is too important to rely on snap judgments or stories that fit information into a commonly accepted way of thinking.

The irony in this story - and perhaps its lesson as well - is that the rise of Trump is embedded at least in part in the decline of trust in the media. How else can someone who makes wild and unfounded claims so frequently get away with it?

The media doesn't help itself or our democratic process when it gets the basic facts wrong, whatever the reason.

Berwood A. Yost is the chief methodologist for the Franklin and Marshall College Poll, director of the Floyd Institute for Public Policy Analysis, and director of the Floyd Institute's Center for Opinion Research. byost@fandm.edu