A quick glance at the bolded rows shows that the most-Democratic districts — many of which are majority-minority districts New York, Philadelphia and Los Angeles — still haven’t voted. We could try to use what we know about the demographics of districts that have already voted — e.g. that Trump does better in areas with lower incomes, lower levels of education, etc. — to project the results of these contests. But the extreme scarcity of Republicans in these places makes that sort of forecasting difficult.

Specifically, we don’t have a good read on who the Republicans in these districts are, how regional and local culture will shape the preferences of these voters, whether not-so-strong blue state GOP organizations have any influence on these voters, or if campaign effects (e.g. an extra-strong mobilization effort from the Ted Cruz campaign) could flip a district away from a candidate who is a more natural fit. Additionally, blue district Republicans who have already voted often had to choose between four or five candidates rather than three, so that difference makes it more difficult to use results from those districts to predict future results.