Our Election Lab midterm forecast has been bullish about a GOP Senate majority for a while. After the latest updates, it still favors the GOP, but less so.

A key reason is that polls are taking on greater weight in the forecast. Although both the House and Senate forecasts still incorporate our statistical model’s expectations — that is, expectations based on factors like presidential approval, state partisanship, incumbency, etc. — at this point our Senate forecast weight the polls more heavily in races with substantial polling. The polls will take on increasing weight in the forecast over the next month or so.

Moreover, now that the Alaska primary is over, we have incorporated polls for every competitive race. In most cases, this process has simply reinforced our model’s expectations, but there are some exceptions (see Alaska and Michigan below). Now that most forecasters and handicappers are relying heavily on the polls, the various forecasts are also converging — compare us and The Upshot, for example.  This is something we’ve long expected.

Here’s a brief rundown of the key Senate races:

Alaska. This state is what’s leading the overall forecast to be less pessimistic for Democrats. We haven’t been trying to predict primary winners, so our forecast in Alaska had been relying on the underlying statistical model and not accounting for polling data. Now that Dan Sullivan has won the Republican nomination, our forecast incorporates two additional pieces of information: Sullivan’s previous experience in elective office and the available Alaska polls. Sullivan has not won elective office; his positions in Alaska were appointed. This favors Democratic Sen. Mark Begich. Moreover, the most recent polls give Begich a solid lead over Sullivan. Because the forecast is relying more and more on polls as the election approaches, these polls count for a lot.

But there is a big caveat here: There haven’t been any polls since Sullivan’s nomination. What we’ve seen after the primary in a couple other states — Iowa and Georgia, notably — is rapid gains by the candidate who first faced a competitive nomination battle. Joni Ernst’s and David Perdue’s numbers have increased since their nomination. For this reason, we may yet see Begich’s lead narrow and the race appear more competitive.

Arkansas. Early on, our model suggested — and most observers believed — that Democratic incumbent Sen. Mark Pryor was vulnerable. Everything since that point has confirmed that. Here’s how I’d put it: Since May, only three of 14 polls have shown Pryor in the lead. One of those was Pryor’s own poll, one was the Democratic Senatorial Campaign Committee’s, and one was the Arkansas Democratic Party’s. To be sure, some of the polls showing Pryor behind were Republican-sponsored polls. But some were not. Until there is a spate of independent polls showing Pryor ahead, Republican Rep. Tom Cotton is the favorite.

Colorado. Back in April, we saw Colorado within range for the GOP, thanks in part to their nomination of relatively experienced candidate, Rep. Cory Gardner. The polls are generally in line with that, suggesting a close race, although there hasn’t been a poll in nearly a month. We still consider Democratic Sen. Mark Udall the favorite here.

Georgia. Our model has always been pessimistic about Democrats’ chances in this state. Republican David Perdue’s recent rise in the polls has only confirmed that. This is a clear case of the polls moving toward the underlying fundamentals of the race. As a consequence, The Upshot’s forecast has also shifted in this direction. Democrat Michelle Nunn may ultimately prove to be a good candidate — even one who outperforms the fundamentals — but right now that doesn’t look like enough to win. For more on this race, see Ben Highton’s recent post.

Iowa. This is a state where our forecast continues to be different than others. The statistical model sees the GOP as favored. Joni Ernst’s surge in the polls is thus another case of the polls moving toward the underlying fundamentals (and may therefore have less to do with her style of campaigning, hugging, etc.). The Pollster average has Democratic Rep. Bruce Braley only slightly ahead. Our polling average has Ernst slightly ahead, though both our average and Pollster’s are close to 50-50. Nevertheless, because our model and polling average favor Ernst, the forecast tilts toward the Republicans. This will be an interesting race to watch.

Louisiana. Interestingly, the underlying fundamentals give the Democrats a worse shot in Louisiana than in Arkansas, but Sen. Mary Landrieu is outperforming that expectation. To be sure, she has been tied or behind in most recent polls. But the polls are, on average, close to a statistical tie. Of the Democratic Senate candidates trying to win in the Deep South, she might be the one most likely to pull it off.

Michigan. Although national conditions and the state’s partisanship tilt Michigan toward the Democrats, basic information about the candidates’ political experience actually tilted our April forecast to the GOP. Since then, however, Democratic Rep. Gary Peters has consistently bested Republican Terri Lynn Land in the polls. In some sense, Land “should” be doing better. So if Peters ends up winning easily, then it could be, as some have speculated, that Land’s defeat was due to her weaknesses as a candidate.

North Carolina. This is another race where our forecast is different than others. Our statistical model has suggested that Democratic Sen. Kay Hagan is favored to win reelection. The nomination of Republican Thom Tillis doesn’t really help the GOP, at least in terms of his previous political experience: It’s unusual for a state legislator to beat an incumbent senator. Of course, Tillis is not a back-bencher but the speaker of the North Carolina House. But that may come with liabilities for him, too. Recent polls suggest that Hagan is leading, which is in line with the model’s prediction and makes our overall forecast optimistic for her. That said, the inclusion of a third-party candidate, Sean Haugh, may complicate the polling in that race.

As we’ve noted all along, our forecast will change as new information comes in, particularly from polling. It’s important to note that, based on our analysis of earlier Senate elections, trends in the polls as of now tend to be trends in the direction of the election’s outcome. So a spate of new polls pushing in one direction should be a good indicator of where a race is headed.

But for the moment, the Republicans are very likely to control at least 48 seats after the elections, including Georgia. With just three seats out of the ones discussed above, they would take control of the Senate. That’s why we still see them as having a better than 50:50 chance to do it.