Search This Blog

Sunday, July 15, 2012

Performance of Droppers in IIT Kanpur

IITs permit only two attempts in the Joint Entrance Examination. You can sit for the exam in the year you pass 12th, and the year after that. Why not be open and allow everyone. How does it matter if a student who passed 12th class 5 years ago can sit in JEE, pass it, and study in an IIT.

IITs used to allow this till 2006. So I took out data for the 2006 batch, and looked at their performance in the first year (when the courses are common, and hence no bias of some department being more liberal or tough or competitive would matter).

In 2006, if we consider the students who had also given 12th class in 2006 (that is no drop year), the number of students were 184, and the average CPI of these students was 7.9.

Those who gave 12th class in 2005 (one year drop), the number of students were 217, and the average CPI of these students was 7.2.

For 2-year droppers (12th in 2004), 81 students had an average CPI of 6.8.

For 3-year droppers (12th in 2003), 35 students had an average CPI of 6.4.

Remaining 18 students (12th in 1999 to 2002, or 4 to 7-year droppers) had an average CPI of 6.4.

If we just look at the students doing 12th in 2005 and 2006, more than 25% of the students doing 12th in 2006 have a CPI of 9.0 or higher, while less than 10% of the students doing 12th in 2005 have a CPI of 9.0 or higher.

I guess this is too small a sample and that too from many years ago to warrant any conclusions, but if similar data was available from earlier years from all IITs, and it showed a similar trend, it would justify IITs' decision in 2006 to allow only two attempts at JEE.


Abi said...

Thanks for sharing these very interesting data!

Can you also post the number (or fraction) of students in each category?

[For example, do the first-timers -- "no drop year" in your classification -- constitute 43 percent of the incoming class at IIT-K -- like they did in the overall JEE cohort?]

Dheeraj Sanghi said...

@Abi, I have given the numbers, 184+217+81+35+18 for different number of drop years, a total batch size of 535. Freshers were 184, which is around 34%, and not 43% which was the overall average of freshers in the entire JEE qualified group.

Pavan said...

I respect your views. But I donot completely agree with taking CPI as criterion to judge this.

Also, I did not quite understand what you are trying to show here. You are comparing people who cleared JEE (although with multiple attempts) to people who did not (but might have if you did not allow multiple attempts to some people). If you compare their JEE ranks (all gave same exam at same time), people who made it through obviously performed better. Should we give preference to people who might have qualified than the ones who actually qualified 'cos they get good CPI later?

Philosopher said...

Mr. Sanghi does this analysis means that CPI is the sole criteria that decides a successful student ?

Dheeraj Sanghi said...

@Pavan, there is enough data to show that the performance in JEE does not correlate with performance in IITs. And one is trying to see if there is a better way to admit students. (A better predictor function, if you want to look at it in terminology of my previous blog on admission processes.

Since it is impossible to compare those who study in IITs with those who don't, the only way to improve admissions is to see whether there is a pattern that we can detect amongst those who are not performing well, and then modify the admission process to discourage or eliminate such students. It is understood that such a process is imperfect, and therefore should not be used to distinguish between two sets with very small differences in performance. But it also has to be understood that the "ideal" possibility of studying and comparing those get admission and those who don't is impossible.

So, if you insist that such a study be done, then it amounts to saying that there be no change in the admission process ever in any university anywhere in the world, even though there is no justification that the current system is ideal. This is something that I refuse to accept.

Why CPI. This is certainly a weak point of the argument. CPI should certainly be not the only and even the most important metric in considering performance. But there too, should CPI have no value at all in judging performance of the student. And again, the suggestion is that between the freshers and 1-year droppers, though the CPI difference is significant, let us not debar the 1-year droppers, considering the limitations of this type of study and the limitations of considering CPI as the metric. However, when we consider the difference between freshers and 2+ drop years, it is very large, and therefore, unless you want to ignore CPI completely, it appears that in a composite performance metric (if one can design that), the 2+ year droppers would most probably not be able to compete with freshers.

Ankur Gupta said...

Sir, I don't agree with you on restricting the numbers of attempt for JEE aspirants because

there are many aspirants who actually start their preparation after 12th std due to lack of information/money/exposure/family support.

Also,a lot of people from small towns can't afford coaching along with 12th std, so they normally go to individual tutors and lag behind compared to Bansal's, FIITJee etc students.
I know my batchmates who started their preparation at 11th Std when my father wasnt aware of IIT at all.

Also if someone is still dedicated to JEE after 2-3 unsuccessful attempts, then I would appreciate that person being a true fighter and someone who knows what he/she wants to achieve in life and to excel in corporate world this is very important quality rather than CPI

Also when you said average, that means there are lot of droppers who have performed better than freshers so why to punish those guys? What is their fault? Just because they were not able to secure couple of extra marks in this attempt does that mean they don't have right to work hard for next attempt.

Dheeraj Sanghi said...

@Ankur, If we start talking about individual instances of successes, one can also talk about individual instances of failures. So that is neither here nor there. And also, let us not get emotional. There is enough data to show that small towns and rural areas are very well represented amongst the IIT JEE qualifiers.

We cannot handpick 10,000 students for all IITs. So we have to have a process which gives us higher probability of finding better students. I am suggesting one such filter. You may want to suggest fine tuning that filter, but you cannot reject that filter on the basis of individual performances.

The fighter spirit that made that student clear the JEE does not seem to help him/her perform once in IIT.

Kyon likhoon said...

Dheeraj why are u giving so much importance to performance in iit(cpi). Do you have data on most successful iitian and is there any correlation with cpi. As long as there is no correlation the system should take steps to incourage diversity :)pailsen

Dheeraj Sanghi said...

@Kyon likhoon, I have already said that CPI should not be the only metric for performance, and not even the most important one. However, any argument, which if logically followed, leads to the conclusion that there should be no change in admission process at any time, I reject such an argument. It is impossible to carry out the kind of study you are talking about. First, there is no way we will all agree to any single definition of success. Second, there is no way, we can get a large enough and random enough sample of alumni to give correct and verifiable data to conclusively put them in successful category. In fact, a lot of alumni who are not successful are also very introvert and their contact information we will not even have.

So your comment will essentially lead to no change ever kind of situation, since we can never do what you consider as pre-requisite for a change.

Considering the limitation of a CPI based metric, I am only saying that perhaps the exclusion of 2+ year droppers was justified since the CPI difference was very large. A small difference of CPI would not have merited the same conclusion. Now, if you are arguing that CPI is completely random phenomenon, I would want you to think of a better metric, which is practically possible to compute.

Chirag Mittal said...

Sir, I agree with this data but in IIT admissions we should consider the background of the student. A student coming from retarded background would not be able to compete with the ones in metros. But this does not mean we should not put restrictions in admission.

My submission is that IITs should do a detailed study, of similar nature, based on some more parameters like background (Rural/Urban), Type of School (ENGLISH/Regional), Age etc and should come up with more wider criteria. Secondly, I dont agree with the definition of dropouts. IITs should give two chances to a person but he should be free to choose his chances. A student who start preparing after 12th or in 12th and choose not to appear in JEE in 12th should still get two chances.

Johnny said...

Lies, damned lies, and statistics

Results are only as good as the assumptions they are based on.Let's look at the same set for things which were overlooked while meticulously finding a co-relation between years of drop and performance (CPI).

184 freshers = 34.5%
217 one year droppers = 40.5%
81 two year droppers = 15%
53 more than two years droppers = 10%

Now the one year droppers are the students from a batch where JEE has already skimmed the creamy students, the so said freshers from that batch the top 35 percent or so.
Now the boiling question is how come JEE missed 40% potential students. JEE surely isn't inline with internal exams where they are awarded CPIs, is it?

The sample set is already rigged by JEE which was not inline with the academic requirements of internal exams held at IIT (K in this case).

Rather than bringing JEE's standard of examination inline with academic requirements of Institute or vice versa, which in turn should also be inline with future options for the "bright future of the nation"; we are manipulating the procedural criteria for appearing in an examinations.

It's a shame that even in academics rather than giving opportunity to all (and I mean all, no cutting of the tail of Gaussian bell), we downsize the number of applicants based on flawed statistics. For what? We will have lesser number of JEE copies to check?

I know justifications of having similar age-group students in a class to eliminate complex among students, et al. What we are doing here is covering one problem with another one.

Life is much bigger than JEE and CPI and even academics taught in school/college. We have already attached too much value to these things and we are a failing society, a failing country as if this country wasn't already full of cultural and religious dogma. Unless we remove these undue weights on the young generation of this nation, we are not making this a better place for our children.

Amit said...

I completely agree with Chirag that the no. of chances can be limited but the student should be allowed to choose his years because the access to information is still very poor in our country. It will be interesting to analyze the academic performance across city (where class XII was done), Class XII board, family profile.

I think family profile plays a major role ... I think children of IIT/IIM/Engg./Civil servants/other-higher-education people do lot better @ both IIT-JEE and IIT because of the guidance they receive.

Mukul Gupta said...

I would say that CPI is not the correct criterion for this comparison however, i wonder what you will get if u do a similar analysis for students entering with reservation quota.

Since they are getting the same education now (at almost no cost) they should be overwhelming the others. although this takes ur blog to a new direction, but might make a few things clear to deciders. Do they deserve to be there?

ofcourse this is a different issue, and you might not be able to explain things, in a rational way keeping every reader happy about it.

Unknown said...

Well I think a person who has the courage to fight further and achieve is any-day better than the one who has never seen a defeat in life.

Because life is not always going to be easy for anyone not atleast for the one who we aims to bring big changes in this world.

Dheeraj Sanghi said...

@Mukul, I don't see how reservation is relevant to the discussion. Within each category of admissions, we have to find the best way of admitting students.

Digbijoy Nath said...

@ Ankur Gupta, sorry if I am misinterpreting, but your statement that you appreciate those as 'fighters' who drop 2-3 times to clear JEE, is something I found to be unacceptable. So what if a student drops 5 yrs or 7 yrs ? Is he/she still a fighter according to you? According to me (it's my opinion, need not be correct), he/she is being a silly person with no foresight/guidance.

The question is: is the 2-3 or 5-7 yrs drop, WORTH it, in the eventual run? Is the sheer loss of 2-3 irrecoverable years worth what we eventually become in life? The IIT brand name is certainly an enviable and coveted one, no doubts, but what I am curious to know is - if instead of dropping for like 3 or 5 yrs, he/she took a different route (NIT, DCE, etc.) and achieved in life almost at par with what his/her dropping-counterpart achieved after losing 3-5 yrs, then is it justified to lose those precious 3-5 yrs ?

Why I am trying to argue is that - I see, and I know, a lots of NITians or alumni from other non-IITs who

1) earn almost the same pay-cheks in similar MNCs as IITians after graduation,

2) clear CAT/GMAT with near 100 percentile and study in top business schools in India/USA (as many IITians)

3) do research/PhD with similarly strong profiles as IITians in top US schools.

Of course, this does NOT mean that IITs are indistinguishable from non-IITs, because that would be hilarious. What I am trying to argue, with proper justification & reasons, is that, Prof. Sanghi's CPI argument/statistics notwithstanding, losing out 3-5 yrs of precious time for studying engineering ONLY in IITs, might not be a worthy and logically justified decision.

As usual, it being my personal opinion, need not be correct :-)

MN said...

As others mentioned, we may want to run an analysis on the socio-economic background of the folks who entered IIT in '06.

Being from the '06 batch, I can cite a few examples - one person from my batch used to drive tempos in UP in evening to pay for his coaching, another was a farmer who completed his BSc and then joined IIT, a third used to teach at a coaching institute to support his family, etc. All of them came not from their 1st or 2nd attempt, but 3rd, 4th...

Such stories were simply missing for those who came straight to IIT from school.

No parent from an affluent background will allow his kids to put in 3rd, 4th attempts.

Since we want to be data driven, just count the number of students who came to IIT from top Delhi schools (e.g. DPS RK Puram, etc) in their 3rd and 4th attempts.
Answer = 0

My 2 cents...

Riju Singhal said...

I think if we had to use "one" criterion, then as of now, CPI (if not comprehensive) is still a fair criterion. Has the average CPI increased since 2006 though?
Regarding droppers, I just have a thought to share. A lot of students, in the year that they drop and prepare for JEE, get a chance to explore advanced text books, and develop deeper understanding. IIT-JEE is supposed to be an exam where you'd face original, unexpected problems, and to solve those, concepts should be crystal clear. Therefore deeper understanding, and giving more time to subjects help. This may or may not be possible if preparation for class 12th exams goes together with that of JEE. The problem however is, that once you enter an IIT, most often in exams you don't need clear concepts, rather a thorough revision of assignments works. The exams (with some exceptions) no longer test the capability to handle unknown situations. This is something that is worrying and might be the reason for most droppers to have lower CPIs.

harry said...

with the scrapping of IITJEE, which had two attempts and AIEEE, which had three attempts, replaced by a single exam JEE, which should have a single and same eligibilty, i.e. for JEE mains and advanced.Now 12th passouts of 2011 have a chance at AIEEE, but, no attempt at .Dont you think that they should be eligible for JEE main and advanced too??

Chintalagiri Shashank said...

Dr.Sanghi, a relevant question which I think those who make up the IIT community should be asking is whether such analysis was performed and presented before or during the discussions for the 2006 decision.

Personally, this decision itself is not of great import to me. I wouldn't have bothered to spend a year of my life preparing for an entrance exam, of all things.

What I do find somewhat irksome about the decision making process that we use is that this kind of data driven analysis seems to be employed only tangentially, and usually after the fact. There's this "act first, rationalize later" attitude that seems to have set in, which has turned much of the decision making process seem, atleast to someone in the position of a student, like an exercise in shotgun decision making.

life goes on said...

In my view, these stats only show that droppers are too exhausted after 3-4 years of preparations for JEE. It would be interesting to see how the CPI looks in final years?

Anyways, good to see some analysis, even though it is half baked.

Devesh Tiwari said...

Though this data looks interesting, I am afraid there is one piece of information which we are overlooking.

Number of students with 2 or more drop years is 134, which almost one-fourth of the population size.

So, if one fourth of the population does not perform well for some reason, then do we think it is reasonable to exclude them from admission process all together, and call that we made a "positive" change or a "justified" change?

Some of these 25% students may be actually performing better than first timers. May be a detailed distribution plot will help here.

If the aim is to choose the best, then we got to give chance to these 2+ drop years applicants because
some of them may be performing better than first timers. That is my hypothesis, so if your data says otherwise I will take back my reasoning.

I liked that change because it may potentially reduce the coaching culture. Some will argue that it only makes things worse, because now you have to prepare from 7th grade may be.

In any case, I am not entirely convinced yet that CPI data in just avg. format can be used to add justification to that change.
Again, if none of the 2+ drop years students perform better than first/second timer, your analysis is sound and complete.

@Chirag: I had similar opinion. why not let applicants choose those 2 attempts just like IAS and other civil services exams.

As I understand, it is more of administrative challenge. Tracking attempts is very hard, given people make fake profiles etc., esp when they are not sure of cracking the exam.

Again, I am not saying it can't be done. I am sure Dr Sanghi knows of other challenges in implementing that scheme, and will share with us.

Sivaramakrishnan said...

One possible way to interpret the data is that the students who can better cope with academic workload and expectations qualify in earlier attempts.

An interesting possibility is that "droppers" who do end up with a good CGPA actually ended up clearing the JEE (or came very close) in the previous attempt, and wanted to improve their rank the next time around.
@Prof. Sanghi: I wonder if you can test any such correlation in the data.

Dheeraj Sanghi said...

@Devesh, As I said in response to an earlier comment, we perhaps need a more refined filter. And, of course, we are discussing this under the assumption that if we had done a larger study, the results would be similar, and CPI is the right criteria to use.

If some people in this group are doing better in the other group, it calls for a careful evaluation, and coming up with an alternative filter, or a more refined filter. But suppose there is none. Or at least the admission team and several experts who were contacted (hypothetically) could not come up with a refined filter. Would it not be ok to still make the change.

I think one of the problems that I have with several of the comments is the following. We are selecting 10,000 students. If we come up with a new filter, which will cause 1 or 10 or even 100 students to be different in the two scenarios, then the immediate response will be, why are you changing if things will remain largely the same. On the other hand, if the modified filter is such that only 9000 students will get selected in both cases, and 1000 students will be different in the two scenarios. Then people will point out just 1 or 2 amongst those 1000 who will not be selected but either are truly good, or have an emotional appeal about them (son of rickshaw puller), and people will say that any scheme that excludes this one student is bad.

If we were to accept this argument that no good person (and who defines good) who could be selected by the current method could be denied admission by the new method, otherwise new method is bad, then there can never be any meaningful changes to the admission process (and for that matter any process that deals with large number of people).

Dheeraj Sanghi said...

@Sivaramakrishnan, Unfortunately, we have no idea about previous attempts of the students. I have no access to JEE data of any year, and in any case, I do not think that JEE asks for this information, or has capability of correlating applications of two years to find out this information.

Manu Thambi said...

Hi Dheeraj,

I have been following your posts on JEE with a lot of interest. I mostly agree with what you say, but in this case I think you are off.

Let us assume for the sake of argument that CPI measures exactly what is important for "success".

We are trying to compare the following two scenarios:
A. Droppers are allowed.
B. Only current year students are allowed.

case A:
535 students are admitted, of which 351 are droppers.

case B:
The 351 droppers are replaced by people further down in the rank list for year 2006. (let us call the set S)

So the relevant question to ask is whether S on average have higher CPI than the droppers. One way to get an upper estimate for the CPI of S, is to look at the CPI of the bottom, say 25, JEE rankers among the current year students (bottom 25 out of 184).

If they have a lower CPI (which I suspect), assuming JEE ranks have some predictive value, then case B would result in a higher overall batch CPI, than case A.

Of course there are other costs for admitting droppers, which have to be weighed against the benefits.
1. More work involved in giving JEE to more students.
2. The wasted 1+ year of droppers working lives which could have been spent pursuing a career path outside IIT.

Dheeraj Sanghi said...

@Manu, Great to hear from you. You are, of course, absolutely right, but strictly within the example you gave. The situation we have is slightly different, though. I have already said (in response to Pavan) that due to limitation of the study (and what you have pointed out is exactly the limitation - we don't compare the IIT students with those who could have come in), we should not disallow one year dropper. So we have 134 out of 535 who are being disallowed. They form 25% of the class, and they need to be compared with some people who would have come from freshers' list. Note that the average of these 134 students is only 6.6 approximately. So we only need to check whether the bottom 25% (as per JEE ranks) of the freshers' list had a CPI better or worse than 6.6. And this was indeed the case that the freshers' CPI was far better than the average of 6.6.

cipher said...

Hi Dheeraj,

Following up on your last comment and Manu's earlier comment, you seem to be justifying disallowing ONLY students with attempts > 2 by comparing these students with the lower end of students (in terms of CPI performance) with attempts = 1. Shouldn't you be comparing students with attempts > 2 with the students with attempts <= 2 in order to justify this? In terms of absolute CPI differences, students difference b/w CPIs of students with attempt = 1 and attempt =2 is larger than difference b/w CPIs of students with attempt =2 and attempt=3.

Sorry for the lose language, I hope I was able to make my point.

cipher said...

Dheeraj, You seem to be supporting the exclusion of 3rd timers by comparing them with 1st timers (You are ok with 2nd timers). Why not compare 3rd timers with the tail of 2nd timers or the overall tail of 1st and 2nd timers? Also the absolute CPI difference b/w 1st and 2nd timers is much higher than b/w 2nd and 3rd timers. So isn't there a less stronger case for starting the exclusion from the 3rd attempt?

Dheeraj Sanghi said...

@cipher, well the reason to do that was that latter statistics showed that 2+ year droppers were replaced by freshers, keeping the number of 1-year droppers as same. And all this analysis is being done post facto anyway. I am sure, if such an analysis was done prior to 2006, we would have looked at the numbers slightly differently. (But then if this analysis was done by JEE across all IITs and across a few years, they would have had much more data and perhaps more interesting, and perhaps finer filters could have been thought of.) Right now, we are just trying to justify a political decision by a small post facto study.

Devesh Tiwari said...

Quoting Dr. Sanghi:

"If we were to accept this argument that no good person who could be selected by the current method could be denied admission by the new method, otherwise new method is bad, then there can never be any meaningful changes.."

I agree to that and I am all open to new changes, in case I sounded otherwise. As I said earlier, I liked that change, I just didn't like this justification.

My problem with this CPI based justification is not with the metric CPI either.

Rather I am afraid how your method can be used to justify absurd new changes.

Consider the following: Today, MHRD minister decides to bar UP board guys from writing IIT-JEE. 5 years later, one IIT shows the data that in 2011 UPBoard students had an average CPI of only 6.4, and CBSE board students had average CPI of 8.9. So, it was Okay to exclude them, though UPBoard students were 25% of the IIT class in 2011. But we didn't let any UPBoard student write IIT-JEE exam in 2012.

Now, let's not pick on the choice of UPBoard and CBSE Board. They just represent two different groups of students. One group is doing better than the other.

I am not against "calling for careful evaluation" when one group is doing better than the other. But that call does not need to be a hasty decision, which we try to justify later to our own souls.
I would say "call for evaluation" does not mean an immediate change, with no intent of experimentation.

I could never understand why the following two options were not tried before forcing 2-drop-year

1. Letting students decide when to exhaust their attempts : I understand tracking attempts is non-trivial (e.g. IAS, UPSC etc.). I am sure there are other challenges that you know of and probably you can share.

2. Adding negative weight-age in JEE score after second attempt : it sends out the same message -- discourage more than 2 attempts.

I am afraid this change was implemented because of other motives (e.g. reducing number of applicants, diluting coaching culture, putting these drop years to better use -- B.Sc.? etc.). I can only guess, and my guess is that poor performance of droppers was not the primary motive for that change. Then why do we want to justify that using your data?
Your data has a lot of value and that should be used for future experimentation not for justifying whatever we did in the past, and that's just my opinion.

Actually, if we implemented the system Dr. Barua (or MHRD) had proposed where 10+2 performance heavily counts towards JEE. After five years, one can show the data that the students with good CPI had done well in the school as well, so this change was acceptable and a "positive" change. But, we all know that there are many things that are not right with that proposal.

In fact, we already know that there exists a good enough correlation between CPI and 10+2 performance. So, using the same avg CPI method/data , Dr. Barua would be able to justify his proposal five years down the line, if we implement his proposal today.

Dheeraj Sanghi said...

@Devesh, If there is a hypothetical board whose students manage to get into IIT but consistently perform poorly, why should we not have a filter which bars those students. But, of course, it will have to be done in a way which is politically palatable. We will have to study more as to why those students are performing poorly. Can we convert that lack of preparation into a general eligibility, or make changes to the entrance test that it checks for those topics more thoroughly, instead of saying that students of a particular board are not eligible.

For example, suppose we find that most students of a particular board
are not performing well. We look into it and find that this board has decided that a student need not study any language in 11th and 12th classes. And we realize that the language skills (as opposed to English skills) of these students is rather poor. Instead of barring this board, we could say that only those students are eligible for IITs who have studies P, C, M, and one language in 11th and 12th classes. Or we could change the question paper pattern to have long passages on science and then ask questions based on those passages even in PCM test.

And to give an example as to why such a possibility is not absurd, IIT Kanpur already has such a policy for PG admission. We do not admit students who have not done any practical work. (So, distance education based degrees, if they do not have any project and lab work, we don't consider them, where as most other universities and even IITs will consider them eligible.)

By the way, there is a difference between the data and analysis that I am presenting versus what one could do with data of five years hence. I am giving data of 2006 and not of 2012, to justify decision of 2006. Ideally, this should have been data of 2001 to 2005 from multiple IITs, but still it is not as bad as using data of 2012 to justify an action of 2006.

Tracking multiple attempts is difficult, but more importantly, if, for a moment, you believe in this study, it is independent of the number of attempts, and only depend on the year of attempt, and it is showing (in a limited way) that students giving JEE 2 years after 12th should be discouraged compared to students giving JEE 1 year after 12th.

I, actually suggested to JEE at that time, that instead of barring someone completely, they could give some penalty marks to the student with each year passing. But, it was felt too difficult to explain politically.

(In fact, I am a big fan of encouraging and discouraging certain types of students through use of enhanced and reduced credits. I have, in my earlier articles before I started blogging, had suggested that we can give extra credits to people doing 12th class from village schools, to girl students, to students whose parents are not educated, and so on, perhaps even for other things like Olympiads. But too controversial.)

I am glad that you refer to the current debate and the current proposals that have gone through IIT Council, and NIT Council, etc. One of the arguments of the other camp has been that pretty much anything any body proposes, it is shot down as not supported by strong scientific study, and that there is not much study one can do in these matters, and hence their experience is all that should count. And our argument has been that yes, ideal statistical analysis cannot be done, because it involves comparing students with non-students. But still we need to collect some data, we need to do some analysis, and we must interpret the results by first understanding the limitations of our study. We are constantly pointing to studies in this space, our own data, and our own limited analysis.

While this blog was a very tiny study, but in general, if one were to keep rejecting any study that is less than perfect, one is essentially supporting the argument that any change should be done with "experience" and not with "data and analysis."

Shishir said...

@dheeraj,though you have bravely attempted to know the validity of a decision through statistical correlation analysis, I think you have overlooked a vital point of the same.

There could be other variables affecting the CPI and they too may have a strong correlation with CPI, thus confusing which variable is the true cause of low CPI. My suspicion is that drop in dropper's CPI may be due to the complex in her mind that she is not upto the mark as she couldn't get in in the first attempt. This lack of confidence , if not overcome early, definitely may affect her performance. So it is not dropper, per se, who is weak. It is the mental outlook of the dropper, which may not have anything to do with the her intelligence,which affects the CPI.

Therefore , unless this cause and effect thing is backed by some exhaustive qualitative analysis, i.e,unless we 'know' that there is indeed some kind of relationship between droppers and their CPI, it is not correct to even loosely predict the cause of the effect based on statistical correlation.

ankur said...

Hi Dheeraj,

Glad to see the data. Quick question - How does your statistic look when we compare the performance at the end of the degree? I would be more interested in knowing the overall performance rather than the performance in 1st year as there could be other factors.


Manu Thambi said...

Hi Dheeraj,

Very interesting discussion.

Btw, I think @cipher has a point, we should be comparing the tail of 1,2 timers against the rest. If the data shows they do better, the interesting conclusion is that effort/persistence matters more in JEE compared to CPI.

I think that effort/persistence matters a lot for success later in life than pure natural ability. So imho, given two people with the same rank, IITs should prefer the person who took JEE multiple times.

A broader meta point: I think the admissions process should try to measure the qualities important for success directly (like mathematical ability, drive, language skills etc). Using membership in a unrelated class (like UPBoard, years since high school graduation, sex, caste) as a proxy for ability when you can directly measure ability is bad (both morally as well as for self interest of the IIT). For example, if we found that the average CPI of girls is lower than that of boys, would we ban girls from JEE? Even if the girls' average is lower, there are plenty of really smart girls who would outperform the guys. So isn't it smarter to just measure ability directly, instead of using sex as proxy for ability?

Finally, I think, there are valid reasons for excluding droppers, like: the expense for conducting JEE, and the "wasted" years for the droppers. Overall, if I were making the decision, I would allow people to take JEE any number of times.

Dheeraj Sanghi said...

@ankur, I did not check that (and don't ask me to open all the files again), because there is a significant difference in competition levels in various departments (for example, even top ranked JEE students will get poorer grades in the department courses, since they have a higher concentration of them in the department). There is also a difference in the graduation requirements across departments in terms of courses (small differences, only in a couple of semesters, but enough to feel uncomfortable with that comparison), and then there are some departments which have traditionally been tough grades versus easy graders. In IME department, for example, everyone seems to give only A and B grades. I won't give an example of the other extreme publicly.

Dheeraj Sanghi said...

@Manu, Your points are as always right on the mark. In the past I have made myself pretty unpopular amongst IIT alumni by my comments like, "JEE is one of the worst entrance exam as a selector of good students, its only merit is its fairness." And if you have followed our JEE saga over the last 3 months, we have been arguing that there are more structural issues in JEE that needs to be handled. The only thing about the Ramasami Committee report that we have liked and want government to do something about, was their recommendation that there be an aptitude test component in the admission process.

The problem, however, is that the government and government appointed directors would only want to do things which they can decided in a meeting and announce and implement immediately thereafter. Anything that requires hard work is ignored, rejected.

By the way, women perform quite well, once in IITs. It is just that there numbers are so small that one can hardly make statistical inference from data. But I wish based on limited data we have, we could make our classes a bit more gender neutral. (Indeed one of my first blog article three years ago was on gender imbalance in IITs. Check out this article.

Pallab said...


"Government statistics show that those who breathe, die".

The correlation between those of breathe and those who die is 100%, however from this if we were to infer that breathing causes people to die, it would sound absurd.

From your post are you suggesting that dropping years cause drop in CPI?

Manu Thambi said...

@Dheeraj, I wasn't implying that women who gets admitted do worse than men (I had read your blog article before). That was just a hypothetical, and choosing sex helps make the point while getting a rise out of someone ;)

Arguments on this topic aside, I really admire the work you have done pushing the admissions process in the right direction, with quite a bit of success!

Vijayant Singh said...

All points aside, IITs (not only Kanpur) should seriously think about their graduate students' output and quality, the dwindling numbers of MS/PhDs they churn out every year and so forth, here's why...

0 - Graduate studies open up specialized knowledge, while majority of undergraduate students are still learning to stem their (sometimes awfully bloated) sense of achievement in making it to the IIT system (for 2 years) and then next 2 in worrying about the campus placement/an MBA seat

1 - Undue attention on BS (evident via this article too) at a cost of monumental neglect (so far) towards the MS/PhD programs has resulted in sites like, this should not be the case ideally, people should instead be thronging to these places for a seat in PhD by now, it's been 50+ years already, honeymoon is over

2 - IIT-K was modeled after CalTech and (probably) MIT, but the scientific (not merely engineering) flair that you find in such campuses is missing simply due to step-motherly treatment to basic sciences (and graduate studies) all over India

3 - Comparing students on CPI/AIR etc. is arcane to me; it may sound rude, but quite a few professors generally derive great pleasures in classification of this sort whilst this time could have been better used else where

Bottom-line is, irrespective of prevailing pattern, JEE would keep attracting 15000 odd most hard-working and awakened kids (of that year i.e.) -- there's no dearth of kids in India by dint of its size -- even if you restrict the attempts to 1 and qualifying marks in +2 to 90%+

IITs now have to decide; whether they want to remain a public system where welfare of entire and inclusive society at large takes prominence, or a private university like system where the short-term performance of students and grants + aids etc. are be all and end all.

I find this discussion very futile one (though interesting pass-time, hence writing here myself), none would like to give up their respective stands, JEE or no JEE, IITs will be fed bright kids as always at least for some more time, we'll keep gloating over this fact, and quite unfortunately, uncharitable remarks like "undergraduate factory to the US universities" will continuously be hurled by 'them' unless corrective measures are taken to better the research output.

Umesh said...

Enough has been said about CPI not being sole criterion of success so I will not comment on that again. I will bring another view to it. The real cost of a IIT education or anything in life is Opportunity cost. As 12th class student, we all are unable to understand it, but it becomes more predominant in later years of life. I support IITs for implementing the 1 year drop criterion and it comes from my personal experiences.
My story: Didn't appear for any engineering exam in 12th passing year. Deemed Engineering too expensive and no one in my circle (seniors/teachers) had cleared JEE. In 2000, appeared for JEE screening without much preparation. Started preparation and same year cleared Bihar Engineering but didn't get admission, because I want to go for JEE. Cleared JEE in 2001, but got my degree with 1 yr extension. Doing well now in Job.
The Cost : 2 yr lost, in which I could have easily done M.Tech from IIT.

My Best Friend, Saurav : Cleared Bihar Engineering in 2000 and went to NIT Trichy. Passed out in 2004 2 yrs before me with good grades. Went to IIM later.

Saurav's Elder Brother Gaurav: Brilliant chap. Was 3rd topper of his college with 80 plus % marks in 12th in Bihar board. Couldn't clear JEE in 2 attempts so went to HBTI Kanpur. Kept preparing for JEE in HBTI also and skipped his 1st year exam for JEE. Couldn't clear again. He was a topper throughout his B.Tech in HBTI but because of his mistake graduated in 5 yrs. (The longing for JEE).

My Another Batchmate, Shailesh : The most brilliant of our batch. Cleared JEE with 1 yr drop but about 1500 rank, took admission to IIT Kgp, and prepared for JEE again this time went to IITD with only slightly better rank. His CPI in Kgp also suffered. Lost a year because of preparation.

My Another Coaching Mate, Gopal : cleared JEE with me at rank 1300, went to IIT Kgp, Dropped out and went to FITJEE Delhi, prepared again and got rank of 85. Came in IITK CSE as my junior. Lost 1 yr because of JEE.

My Another Friend, Deepak : Cleared Bihar Engineering in 2001, couldn't clear JEE,so went to BIT Sindri in Civil, did M.Tech. from IIT Roorkee in 2007 and now employed with Airport Authority of India

My Younger Brother : Cleared Bihar Engineering after 1 yr drop. But I suggested him drop it and keep on preparing for JEE 2006. IIT implemented 1 yr drop rule. That plus some other issues and he got depressed and did really bad in AIEEE also. He vowed not to study and I had a hard time persuading him. Finally got admitted him to a Tier 3 college in BCA after 3 yr drop. Now he is pursuing MCA from NIT Trichy. He studies is delayed by 4 yrs because of all wrong choices.
In reality, I have seen more examples of my younger brother, Gaurav and Shailesh. In south I see more examples of Saurav and Deepak. The real cost of IIT is opportunity cost. The time lost in preparation which is irrecoverable. I sometimes think, was going to IIT the best decision I made. No doubt it taught me a lot and I have the IIT Brand. But the cost on family resources was really worth IIT at that time.
I was happy that IIT implemented the 1 yr drop rule. There was a time, when IITs were only way to get a sure shot job but that has changed now. My brother's classmate opted for a BCA from St Xaviers Ranchi and she got hired by Wipro in her final yr from campus. She is doing M.Tech from BITS Pilani now in 4 yrs. (while on job). The new JEE may miss some of the anecdotal examples and some diversity, but JEE is all about selecting the best based on statistics and statistics can't go on chasing the LONG TAIL of Data.

manav said...

@Dheeraj Sanghi.............i took 1 year drop in bansal classes,kota for iit 2009 but was not able to make it........i was deeply hurted and as an ill-effect of that i did my aieee 2009 very badly and got some 40,000 rank,,,,,,,,then i dropped 1 more year for NITs and got 6000 AIR(GEN),,,and im presently studying Electrical and electro. at NIT Surathkal(karnataka)...........but now 2.5 years have been passed away in the college and still im thinking that WHY DID I WASTE MY 1-2 EXTRA YEAR....and therefore was not able to perform well in my college and as a result my CGPA IS 6.0 NOW.....plz help me in getting rid of this stressful situation sir i have lost the STUDENT in MYSELF........plz sir..or any one plz.........

Umesh said...

Please re-read my earlier comment about opportunity cost of IITs. None of my friends cleared IIT JEE and they are all doing good. My younger Brother's friend didn't clear even AIEEE and all of them still have a job and earning decent.
Examples :
1. His School Friend : low AIEEE 2005 rank. Went to a private college JITM, newly started, had to struggle for a job a bit due to recession in 2009, but is now working in July Systems and doing decent.
2. His School Friend : A girl: BCA from St. Xaviers Ranchi or was it Women's College ? Campus placed in Wipro in final year. Doing Distance education M.Tech from BITS Pilani sponsored by Wipro.

3. His school Friend : A guy : B.Tech from private college HIT, Haldia in WB. M.Tech from NIFT (national institute of Foundry & Forge Technology), Hatia, Ranchi (Jharkhand). Passed out in 2012. Cleared BHEL Engineering Exam in current year. Now working with BHEL, Hyderabad.

4. My Wife : No JEE or Kota Preparation. No Year drop. BE in IT from Asansol Engineering College (2004-08) (a private college), Campus placement in Accenture and Cognizant. Joining delayed by 1 year due to recession in 2008. Became lecturer in local college for BCA @Rs. 7000/pm. Joined Accenture in 2009. Two promotions in 3 years and can go onsite, but we both prefer to stay in India. Most of her batchmates are placed and working in services based companies and doing good.

5. My wife's Best Friend : Same eduction. Placed in Satyam but Satyam was involved in Fraud in 2008 so offer cancelled. Recession meant no other opportunity also. Worked as lecturer in local college for 1 yr. Joined M.Tech from Hindustan College of Arts and Science, Chennai (2010-12). Worked as part time lecturer in same college for BCA in 2nd year. Now works as lecturer in DMI College of Engineering in as Lecturer (2012 onwards)

Umesh said...

6. My Younger Brother : Mentioned in my previous comment. Cleared Bihar Engineering in 2005 with 1 yr drop. He couldn't give his best to studies because of lot of family issues. Dropped for IIT JEE 2006 on my advice and I didn't allow him to take admission in BIT Sindri Mining saying it would affect his prepartion. Couldn't appear for IIT JEE 2006 because of 1 yr rule. I didn't teach him anything, no oaching and an AIR of 70000 in AIEEE. Missed Bihar Engineering because exam got pre-poned. I judged and scolded him and he lost the student in him. After 1 year of serious conseling and persuasion, he agreed to appear for entrance exams after May 2006. Only 3 exams left, Allahabad Agriculture univ, BCA Bharati Vidyapeeth Pune, IIPS Indore BCA +MCA 6 yrs. Cleared all three and took admission in IIPS Indore for BCA + MCA in 2007 with a 3 yr break. Did below average in 1st yr, average in 2nd yr. I encouraged and supported him, visited his hostel in 2nd yr during end sem for help in exam. 3rd year started preparing for MCA entrance from a local famous institute run by a single man. (institute produce half of top rankers in MCA entrance). Gradually started doing good inside coaching, but wasn't the best in institute. I supported him. Got real nervous and had little preparation for BCA final year exam because of coaching commitments (took 7 am till 10 pm) of his time with breaks for classes only to get attendance complete. I flew to him on a one day notice, without informing my boss. Was able to teach him only for 6 days. because for 1st 4 days he went to coaching and didn't dare talk to teacher about letting him go. So I had to talk to his teacher and asked him to let him study. Taught for 6 days and he appeared for 3 papers. I came back when he still had 2 papers.
Result : Top Ranks in all MCA entrance exams. Chose NIT Trichy. Didn't get good ranks in MCA Roorkee because I pulled him out.
Passed : BCA final year papers.

NIT Trichy : No support in course work from me. I wanted him to study with batchmates.
1st year: Got supple (I pointed out angrily that I can't come every time leaving my job to help him with his course work)
2nd Year : Decent (encouraged but just over phone and gave study material/video lectures)
3rd Year : Better and on going.

Conclusion :
1. Cracking entrance exam using highly specialized coaching doesn't mean that you will do good once inside.
2. Support and encouragement matters a lot, but won't be needed if you don't make the mistakes in first place and don't advise wrong.
3. Mistakes are correctable.
4. Drops are not advised, definitely not more than 1 yr.
5. India now offers a lot of opportunities but become employable. Learn rather than get grades. If college is not good, you can get study materials from NPTEL/MIT Open Course ware/Berkley and Stanford/ A good source is I learned my CS fundamentals from there, even after I am a B.Tech from IITK

Harsh Kumar Narula said...

I wrote JEE in Hindi,dropped twice and passed out with 8.8 CGPA from IITB. I think you are missing some parameters in this analysis. May be category/quota etc. No offense to anyone.

Dheeraj Sanghi said...

@Harsh, no predictor is perfect. If someone could predict future with certainty, then there will be several other exciting career options for him/her. And hence one can only talk about statistical measures to improve prediction. A single example, or even several examples of wrong prediction are meaningless, since in any system of prediction, you will have such examples. The point is to find a system where there will be less number of such examples.


What if a guy passed 10th nd coz of some reasons took a 2 year gap in b/w 10th nd 11th?????will IIT allow such a guy to appear for exam?????I mean will they accept him????