America’s Dark Voting Legacy

Sanjay Rangavajjhala
8 min readNov 24, 2020



It is generally accepted that the electoral college does not treat every citizen equally.

Indeed, when viewing a heat map of the electoral votes per capita, one can clearly see the discrepancies between states. States like Wyoming have disproportionately more votes per capita than other states, like California or New York.

Legend is Electoral votes per Million Residents

When placed side by side with a heat map of urbanicity, it becomes clear that there is a pattern by which states are favored by the electoral college. In states with larger population densities, there seems to be a lack of representation for its citizens. California and New York are some of the darker states on this density map, while Wyoming is almost indistinguishable.

Legend is Population per Square Mile

This relationship between population density and electoral college representation is easy to understand by just looking at these heat maps. However, could this relationship extend into population demographics as well? As the tweet below by former Secretary of Labor Robert Reich aptly indicates, most of us have an inkling about which demographic groups benefit from electoral college representation.

Tweet by Robert Reich, former US Secretary of Labor

Our team hypothesized that the bias of electoral representation in low-density states could also be indicative of the systematic disenfranchisement of certain demographics. The demographics of urban areas are generally understood (more minority populations, younger, etc.), but we wanted to quantify which demographics are actually being disenfranchised, and by how much, using census data.


We began by calculating the electoral vote per capita for the voting population of each state, much like what had been done for the red heat map above.

Data gathered by the US Census Bureau (

These values of electoral votes per capita for each state could then be multiplied back to the individual demographics to ascertain how many votes each demographic was responsible for in each state.

We could then sum the number of electoral votes for each demographic across the entire US, to get a representation of how many votes each demographic received. This became our “Actual” column below.

So the next logical question is, what is our “Expected” column. To derive this, we ignored the electoral votes per state completely. Instead, we made the assumption that if a demographic made up 48.5% of the population of the US (as in the case of males), they should also get 48.5% of the 538 available electoral votes. This roughly translates to 260 votes in this case.

It is important to note that any difference between the expected and actual values are solely the result of an error in the way electoral votes are partitioned per state. Giving more electoral votes per capita to some states over others meant some demographics got more votes than others as well.


The value of “Actual” minus “Expected” is depicted below. A positive difference means a demographic receives more electoral votes than they “should”. Likewise, a negative difference is indicative of the disenfranchisement of the respective population.

A little context about the y-axis. The numbers you see may not seem like a lot, but in 2019 each electoral vote was equivalent to about 44,000 people. As such, a disenfranchisement of 0.1 electoral votes can be equated to throwing away the votes of almost 4,500 members for a given demographic. As we have seen most recently in the 2020 election, these margins are more than enough to decide elections.

The largest differences seem to be in gender, race, and income
The largest differences seem to be in gender, race, and income

Note that this is a zero-sum environment. You cannot add votes to one segment of the population without stripping votes from another.

After visualizing the data, we found the greatest disparities among gender, race, and income. Intrigued, we decided to see how these specific subcategories had changed over time, based on data from 2017 and 2018.

We derived the data for 2017 and 2018 using the exact same aforementioned method. Below, we graph the difference between “Actual” and “Expected” over time for each demographic split.

When looking at gender, we can see that males have consistently been in the positive, while females have consistently been negative. This indicates that the electoral college consistently favors one gender: male.

This next graph shows the delta based on income. We see a general negative trend in the representation of people below the poverty line, indicating the disenfranchisement of this population will only get worse with time without a fundamental change in our system of voter representation.

Finally, the delta based on race. While most races tend to hug 0—meaning they are relatively unaffected by electoral college representation— the white population seems to experience a very large and growing positive delta, at the severe cost of the Hispanic population.

Note: There is an observable anomaly in 2018, which presents itself as significantly more erratic changes in the demographic representations. This should not have been caused by 2018 being a voting year, as the populations of demographics don't change based on if it is a voting year or not. Rather, we think it may be attributable to certain government policies that took effect during that year, such as those concerning immigration and welfare.

Looking Forward

In most of these graphs, but arguably most clearly in the comparison of racial demographics, we see a trend of certain demographics getting an increasingly positive delta, while others are going further into the negatives with time. This is no coincidence. It is obvious that the value of votes is tied to more than just population density.

As US News reports in May 2019, America is only becoming more urban. The key demographics which largely compose these urban areas— minorities, people of color, low-income individuals— are accordingly becoming more concentrated in these urban hubs across the nation. The disproportionate representation of states by the electoral college only makes these populations more and more disenfranchised, as seen in the trends of the graphs above.

The consistent pattern of “rich, white men” getting disproportionately more votes does not seem new. Remember when the only people allowed to participate in our democracy were “white, landowning (rich) men”? It seems that there is still a legacy of the flawed principles on which our country was first founded: that some votes matter more than others.

Keep in mind, we have not even touched the disenfranchisement of votes using a winner-take-all system per state. In the most recent election, as currently reported by the Associated Press, Californians cast over 5.7 million (more than half the mid-western states combined) votes for the Republican Party. Yet how many electoral votes from California went red? 0. The overwriting of these votes points to a fundamental flaw in the winner take all system: we need more granularity in how electoral votes are assigned.

Possible Solutions

I will continue to refer back to this concept of granularity, so it may be helpful to define it. Granularity is how precise the electoral college is in depicting the true proportions of the population. It helps to think of it like pixels: the more pixels you have, the more accurate the picture is.

To make the “picture” clearer, the most obvious solution is to increase the number of “pixels”, or in this case electoral votes. This would prevent rounding errors in assigning electoral votes to states and could standardize the “electoral votes per capita” across states.

Another approach would be to implement mixed allotment nationwide. This is simply getting rid of a winner-take-all system on the national level and pushing it down to the district level. Nebraska and Maine already use this, which is why some electors from these states may vote blue while the other may vote red: it all depends on which district they come from.

Electoral map of California by County

If implemented in California, instead of all 55 electoral votes going to the Democratic Party, 22 of them would go to the Republicans. This is because 22 of the 53 congressional districts swung red(the extra 2 electoral votes are assigned based on the popular vote of the state).

However, then comes the logical question of “If we are reducing the granularity to the district level, why not do better and take it to the individual level?” In essence, determining our election through a popular vote. This is the ultimate granularity we could reach, and the only true way to ensure each vote cast is given the same value.

Actually getting to a popular vote directly is near impossible in the current political climate: it would require a Constitutional amendment. However, there are creative ways around this, like the National Popular Vote Interstate Compact (NPVIC).

The NPVIC is a legally binding agreement between states which pledges to give all of their electoral votes to the winner of the popular vote. This is an efficient way to bypass the electoral college completely, and in theory, states with a cumulative total of only 270 electoral votes would need to sign up in order to ensure that the winner of the popular vote is also the winner of the electoral vote. As I write this article, states with a cumulative total of 196 electoral votes have signed this coalition, which is 70% of the threshold of 270. Once that value is reached, the electoral system would essentially be rendered inert.

There are creative solutions to this issue, but one thing remains certain. The current electoral college is incredibly flawed, and it will only get worse unless we take action now. Political action from a grassroots level can be incredibly powerful in affecting this very change. Staying informed, contacting your representatives, and voting are some basic civic duties we can use to better this system and make every vote matter.


Milligan, Susan. “Rural vs Urban America”. US News.

U.S. Census Bureau. (2020). 2017–2019 American Community Survey 3-year Citizen, Voting-Age Population By Selected Characteristics[CVS Data file]. Retrieved from

National Popular Vote, 30 Oct. 2020,