* In the enhanced Presidential ballot CSV files, the first 9 columns are the exact values as supplied by the NYT feed (as in the raw CSV), but the subsequent columns are some additional calculated values which are useful in seeing the changes (sometimes called deltas) between batches. Once you've downloaded an earlier version, we recommend applying ' pretty formatting' to both JSON files and then comparing them in a visual diff tool like WinMerge to compare the differences. Perform a search for the JSON URL to see if archived copies are available. Some of these are available in the Wayback Machine. Researchers may also be interested in obtaining previous versions of the JSON data to verify whether NYT or Edison made later modifications to the result feeds. It appears to only be available for Florida, Michigan and Pennsylvania, for some reason.) (This JSON file appears in a different structure and may include extra data. We recommend the Firefox browser for exploring JSON data as it has a simple tree interface for drilling down into the data as well as keyword search. These contain only a core subset of the JSON data.

The NYT does not provide CSV files of the data, but you can generate CSV files below if you find these easier to work with. Read Understanding the NYT Data for more information on common pitfalls. If there are any instances where the data does not match the Secretary of State's final certified results, the SoS data should take priority. The numbers for ballot batches are low-resolution in that they are only given in percentages and rounded to three decimal places (there's no exact ballot counts, except for the final result). The following JSON data comes from the New York Times, and appears to be a transformed subset of the paid data stream from Edison Research, combined with a few entries from the Associated Press. Secondary Ballot Races Results by CountyĬoming soon. If you spot unusual anomalies, discuss them in the comments below.

Read Understanding the NYT Data for more details on the low-quality data provided by NYT and common pitfalls. This is due to the fact that earlier batches of ballots have a greater impact on the cumulative margin than the later batches. It's expected that we would see high volatility in the red and blue lines at the start of the counts, and a smoothing effect towards the end.

Thanks to for developing this interactive chart. Estimated percent of ballots counted (eevp)

Winmerge compare json