This graph shows the answer to the generic Congressional ballot question, “This November, do you plan to vote for a Democratic or a Republican candidate in your Congressional district?” This question is highly predictive of the actual popular vote. Also, for each percentage point of movement away from 2014, each party’s seat count should change by about three seats (with some uncertainty).
We draw data from the Huffington Post feed. Our version of the database is here, and is updated regularly using the script convert_huffpost_csv.py, originally written by Ryan Buckley. On any given day, the graph shows the median of the last 3 polls, or all polls ending within the previous 21 days, whichever is greater. If the same organization performed multiple polls, then only the most recent poll is used. The rule is implemented using the MATLAB script House_timeseries.m.
Note that because of partisan gerrymandering and other factors, Democrats would have to win the popular vote by a substantial margin to take control of the House of Representatives. For example, the national House popular vote in 2012 (blue line) gave a margin of 1.2% for Democrats, yet Republicans currently control the chamber, 234-201. The margin needed for 2014 is shown using a green line, and comes from an estimate by Harry Enten.
The only period during which a change to Democratic control would have been likely is in October-November 2013, during the Congressionally-triggered government shutdown. This was immediately followed by a sharp swing toward Republicans during the rollout of the Affordable Care Act and Healthcare.gov.
Note that the generic Congressional ballot question and the Senate Meta-Margin are uncorrelated. The House and the Senate do not move together during the campaign.