TCPD Columns

The Privilege of Education

Education eludes participation in politics. This is a common notion that has been making circles from small Indian households to Big Publishing Houses. The intent of this piece is to test this hypothesis, by parsing it with relevant data. The analysis will be focused on answering 2 questions: 

  1. Is there space for educated politicians in Indian politics? 
  2.  How do educated politicians perform as opposed to their lesser-educated counterparts?

The Dataset reports the level of education for each participant in the election by using data from the MyNeta portal which categorizes education into 11 distinct levels.

1.   5th Pass

2.  8th Pass

3.  10th Pass

4.  12th Pass

5.  Graduate

6.  Graduate Professional1

7.   Post Graduate2

8.  Doctorate

9.  Literate

10. Illiterate

11. Others

I have chosen the data from the 2014 General Elections from LokDhaba. General Election results from 2019 had to be ignored because > 90% of the data pertaining to education was missing.

In the 2014 election, there seems to have been a similar level of participation from the higher-educated as opposed to lower-educated candidates. However, this distribution has high levels of missing values (first bar in Figure 1) which might be attributed to losses in data collection. Overall, there does not seem to be an inherent visual bias towards any particular education level.

Figure 1: Number of Candidates ~ Education Level

Now that we have looked at the participation of educated politicians, we shift gears and look at the Performance of these individuals. The performance will be looked at by measuring 3 key metrics:

1.   Recontest: The probability of politicians getting a ticket again to run from the same constituency will be checked. The option to recontest signals that the party believes the candidate will perform in the new election.

2.  Win Probability: This will tell us the probability that the candidate will win the election

3.  Margin Percentage for winners: This showcases the margin of votes by which the candidate won against the second most popular candidate. This is a proxy for the candidate’s popularity in the constituency.


To look at the differential impact of education on recontesting, I have run a Logit model to compute probabilities of recontesting. Logit regressions allow for the model to fit

non-linearly, which is ideal given that ‘Recontest’ is a binary variable. The Y-axisin this bar chart represents the average probability of ‘recontestation’ of a candidate given their

education level.

Figure 2: Probability of Recontesting ~ Education Level

Looking at Figure 2, we can see that the low-education levels correspond to a lower probability of recontesting elections. The results for higher education levels (Graduate and Beyond) are all statistically significant at the 1% significance level.

Probability of Winning

To determine the probability of winning, we have, as before, run a Logit Model given that Winning is a binary variable. The Y-axis4 represents the average probability of winning the election given the candidate’s education level.

Figure 3: Probability of Winning ~ Education Level

Looking at Figure 3, we can see that politicians with higher education enjoy a higher probability of winning the election in their respective constituencies. The results for the higher educated candidates are statistically significant which is visible in the very thin confidence intervals of the bars.

Margin Percentage of Winners

In the case of modelling margin percentages, I have, first, filtered the data for only winners. Then, I ran a linear regression model, unlike the previous 2 instances where a

logit model was used, as margin percentage is a continuous variable. So, in this case, the Y-axis or the bar height will represent the margin of votes (in %) that the candidate wins the election in comparison to the candidate who came second.

Figure 4: Margin Percentage of the Winner ~ Education Level

Looking at Figure 4, we can see that politicians with higher education do win by a better margin than their lesser-educated counterparts, especially the ones who have a

doctorate in this case. The effects are statistically significant, more so for the winners with higher education levels which is also seen visually with the narrow confidence intervals.


In all the cases above, we saw that politicians with higher education enjoyed a lot of benefits vis-a-vis their performance in the election; they are more likely to recontest an election as their education level increases; they have a much higher chance of winning the election, and they win that election with a higher margin of votes.

However,  there is one major confounder that cannot be accounted for: Wealth. Higher education, especially at very high levels such as Post Graduation or Doctorate is largely only accessible to the wealthy in India. So, income here may be driving the aforementioned trends. There is definitely more to be carried forward with this hypothesis with newer, more relevant data which is not available at the moment. Notwithstanding the possible impact of wealth, the results above are striking enough to challenge the popular notion that education precludes participation in politics.


I would like to thank the TCPD team who were very responsive and involved with the research in this piece. I am also grateful to the LokDhaba team for making important data public.

About the Author

Akshar Katariya is a student at Ashoka University and a data nerd who is interested in applied analytics across many disciplines.


My Neta Portal; 

TCPD LokDhaba for 2014 GE; 

1 This will be abbreviated to Professional in Plots and Graphs for visual purposes

2 This will be abbreviated to PG in Plots and Graphs for visual purposes


This article belongs to the author and is independent of the views of the Centre.