Search This Blog

Saturday, July 23, 2016

Why do Seats remain vacant?

Every year, a few weeks after the end of the admission season, there will be a report on vacant seats in IITs, NITs, etc. A fairly large number will be mentioned in the report. And we will be told that this happens because certain disciplines are not at all popular, and should perhaps be closed down, and may be those many seats should be increased in the more popular disciplines.

Indeed, there is no doubt that some disciplines are fancied by the herd of sheeps, assumed to bring in the riches. But is that the only reason. Are there really no students interested in studying those disciplines. In particular, given that a significant number of graduates don't go for core jobs, but instead go for MBA, finance jobs, IT jobs, and other such careers, how come such students are not taking admission in these "branded" colleges.

The reason for vacant seats, unfortunately, is not the lack of demand, but the stupidity of the admission process.

It is a common belief among the public as well as academic administrators that if we do a large number of rounds for admission, all or most seats will be filled. So we have seen our admission process go from 1 round to 2 to 3 and this year to 6 rounds for the admission to NITs and IITs. It is also believed that if we do joint counseling of larger and larger number of institutions, the number of vacant seats will go down substantially. Of course, both joint counseling and larger number of rounds would help a bit, but they can not solve the problem on their own.

If we look at the statistics of the 6 rounds, we notice the following: After the first round, there were 6490 vacancies. That means that so many people did not accept the admission offers. This is pretty reasonable given that there were almost 35,000 seats and many among those 35,000 would prefer places like BITS, IIIT-Delhi, and so on. The next round filled up these 6490 seats (well, barring a few that remained vacant. I am ignoring them since they are too small and don't change the main argument of this post). However, at the end of the second round, there were 2716 vacancies. Again, pretty reasonable given that they must have been offered unpopular programs and many would have preferred colleges outside JOSAA and not accepted admission. At the end of the 3rd round, the number of vacant seats come down to 1837, which is consistent with the argument made above, and is telling us that if we just keep doing a few more rounds, we would be able to find students for most vacant seats.

However, the story changes after this. At the end of the 4th round, the number of vacancies actually increase to 2021. Why did this happen. Well, some (or many) of those 1837 did not accept the admissions offered to them. But now, even those who had accepted admissions earlier, have started withdrawing. And this withdrawal becomes a serious business during the 5th round. At the end of 5th round, the number of vacancies is 4094.

A large number of students have withdrawn during the 5th round. Why did they not withdraw earlier if they were unhappy with their admission offer. Well, there was no incentive for them to withdraw, so might as well hope against hope and see what they can get in the 5th round (and what they can get in other colleges outside JOSAA). They withdrew during the 5th round since they were told that if they did not withdraw, their entire payment may be forfeited. (Some dd not withdraw even then, since they know that in the future MHRD will come to their aid and ask the institutes to return the money).

The government insists that there can be no financial penalty for withdrawal till the beginning of the semester. So we need to have all but the last round before the deadline of withdrawal, and only one round after the last date for withdrawal.

Unless people, who are not going to join, withdraw, we don't admit more students.
There is no incentive for people to take early decisions and withdraw as soon as they have multiple options.
But most importantly, and this is something that is often ignored by our academic administrators, even when someone withdraws, and there is a vacancy, we are filling that vacancy by someone who showed willingness to accept that admission more than a month ago, and who, in the last one month, has probably got many other admission offers. But since the counseling group does not know who is still interested one month later, they end up making offers to next in the queue who are not interested.

So, if consider the JOSAA admission process, the 6th round has filled up about 4000 vacancies, but if we take the survey of all institutes who have got admission through JOSAA, some time in August, I would not be surprised if there are still 5000 vacancies, which will not be filled.

Is there any way that we could have offered admission to more students. Of course, yes. But not by more rounds. We will have to solve problem by looking at the genesis of the problem.

So the problems and potential solutions are:

1. We admit students only after some people withdraw. Why can't we admit more students then the so-called number of seats. We have data for many years and we know roughly how many people will not accept offers. Based on historical data, we can always admit more students. We can be a bit conservative not to get into a situation where we have more students than what we can handle. But let us face it. If in a particular year, we do get a few extra students, heavens are not going to fall.

2. There is no incentive for people to withdraw early. This is a huge problem and a lesson. When you try to be populist and do things which common people will consider as friendly to them, you will actually end up doing things which hurt common people. Allowing students withdrawal till the last minute without any penalty will result in unfilled seats. And thousands of students not getting admission hurts more than a few thousand rupees loss to a few people. The right thing to do will be to declare that free withdrawal can only be done till X days before the semester and after that deadline, every day, there will be an additional deduction of the money if a student withdraws. This will ensure that people withdraw early and we are able to conduct not just one round but multiple rounds after people have started withdrawing.

3. We don't know who all are still interested in the programs one month after they filled in the choices. This is really the biggest problem. Currently, the way we fill up seats is by way of a "Spot" round (which is not happening in JOSAA this year). In this round, people have to apply afresh. So those who are no longer interested are out of the system. And invariably they have to be at a location physically and give a significant amount of money within minutes of getting admission offer and since it is being done through physical attendance, if someone does not want admission, the next person is offered the same.

Spot round has its own problems, of course. Traveling on short notice is not easy and airlines make a lot of money in this season. (Here you go, the money that you saved through full refund in one college, you paid to the airline. So you really did not save much.) Everyone has spot round in the last week of July or 1st few days of August. So there isn't much option regarding traveling. Invariably, spot round happens after the semester has started. So the students are joining late, have missed out on the orientation program, have missed out on the first assignments, first lab, introductory lectures, etc.

Another way to solve this problem will be to have an incentive for students to withdraw from the counseling process. Again, you don't have any other handle on the students except a bit of financial handle. So if JOSAA (and other similar counseling processes) could ask for more money after the first couple of rounds, which will be refunded progressive less as the days pass by. Also, the student should be able to delete options that s/he has filled in. So you could ask them to deposit Rs. 10,000 after the second round, and then say that if the student is not offered admission in any of his/her choices, the entire Rs. 10,000 will be refunded. On the other hand, if the student is indeed offered admission, and s/he decides not to accept it, then the refund will be based on how delayed the withdrawal has been. This will ensure that as soon as I get an offer from another good college, I go to JOSAA website, and remove some of my lower preferences which I would not want to accept in comparison with the offer that I have received. This makes the counseling process much more efficient and reduces the stress levels in the system tremendously. Coupled with a few extra admissions, this can really revolutionize the admission process and make it absolutely smooth. But it will require a small financial penalty for late decision making by students and parents.

To conclude, the seats remain vacant not because there is no one to accept those seats, but because we have a brain dead (but populist) admission process which can not identify the students who may be interested in those seats.

Saturday, July 16, 2016

Improvement in Counseling Process

This is the admission season and counseling is the most common word that I hear in this season. Conceptually, the process is simple. A bunch of universities come together for joint admission process. Each one of them decides how they will rank the applicants. Each applicant submits a list of programs/institute pairs in the order of preference. The counseling software goes through all this data and matches applicants with a program/institute pair while ensuring that no one who is ranked higher as per that institute's ranking of candidates is given a lower preference than this combination.

Historically, the choice list is sacrosanct. You can't change anything in this once the allotment process starts. So you must fill up your order of preference very very carefully.

If we go back many years, the student had very limited options once the process of admission begins. If you are offered admission, you have only two choices - either pay the fee and you will then be considered in all subsequent rounds, and you finally join the program which was allotted to you in the last round. The other option was to opt out, and in that case you will be out from all future rounds as well. If weren't offered admission till a particular round, then of course your only option was to be considered in the next round.

The first improvement in this process happened when people started realizing that if someone has got admission in an IIT, and has blocked a seat in an NIT, inability to withdraw in the middle is ensuring that that seat goes waste. So a withdrawal anytime till a last date was introduced early on, which essentially meant that you are out of the system and you will not be considered for any future admission round either.

Often, the last round of admission would happen after the classes had already started. A large number of institutes and universities in India believe that their teachers are anyway useless and hence if admissions happen even one month after the semester has started, no one loses anything. However, the students had other issues. If I have joined NIT X, and after a week of classes, you tell me to move to NIT Y within a day or two, it will be a huge problem for me to pack up, find train reservation, or travel in unreserved class to reach the other corner of the country, settle their without the benefit of an orientation program, quickly make new friends, copy notes of the classes that I have missed and so on, only to be told to get out and join NIT Z after a few days. So there was a demand that once I have joined a particular institute, I do not wish to change my location even if I am getting my higher preference in a subsequent admission round. And this introduced the concept of "sliding." That I am only willing to be considered for my higher preference in the same campus but nothing else. Of course, this option was made available not only for later rounds, but even earlier rounds, so that one can plan one's travel and get reservations done. Notice that the order of preferences remained sacrosanct. Sliding only allowed that some of my options could be removed from the list, but those that remain were strictly in the same order that I filled up in the beginning.

People extended this further and said that once I have got admission in a particular program, I don't want to go through a phase of anxiety regarding what I might get in the subsequent rounds, and I just want to freeze my admission. This resulted in creation of "freeze" option for the students. This option is often used not to avoid anxiety, but when someone wants to rethink the order of preference, and feels that the current admission offer is better than what one might get in the subsequent rounds. Which ever way you look at it, essentially what has happened is that we have allowed the student to delete a large number of options, without giving an option to change the order of preference of the remaining list.

What next?

The biggest problem in today's counseling process is the following. When I was filling up the choice list, I did not have any admission offers. I was interested in 50 different programs, and I listed all of them in the order of my preference. However, a week later, I received an admission offer from another college which I would rank higher than the last 10 options out of 50 that I have filled. If out of these 50, I receive an admission offer in my choice number 41st, I really don't want to accept it, since I have a "better" offer outside this system, but if I don't accept it, I would not be considered for my higher preferences in later rounds. So I accept it knowing fully well that if I don't get higher preferences I won't join this program. This causes inefficiency in the system and hurts everyone - this student, other students, and the institute. We must solve this problem. And the interesting thing is that it is absolutely trivial to solve this problem. Most counseling softwares would probably need a few lines of code to change to allow this. So when a student is offered a particular choice, the student should be able to say that I don't want to accept this, but I still want to be considered for my higher preferences. To ensure seriousness of this choice, one may take a bit of money also from the student. And programming wise, this essentially means that the current offer and any lower options are deleted for this student. Notice that this is exactly what the software was doing in "sliding" - delete certain options without changing the order of preference for the remaining options.

Actually, allowing students to delete certain options at any time without changing the order of remaining options will solve other problems as well, and this does not create any problem for the counseling process. For example, we went through the historical need for "sliding" where we said that it was introduced because it was difficult to move from one institute to another at very short notice. Consider this. I have been allotted a program in NIT Allahabad. I certainly don't want to travel 700 KM to go to NIT Jaipur in the next round. However, I have no problems in moving to IIIT Allahabad, the college next door, in the next round. The "sliding" option does not help me there.

And hence what is needed is an option to delete some program/institute pairs from my original list at any time. And as long as the order of remaining options remain the same as in the original list, it won't create any problems for the allotment process. Having this option allows a more efficient allocation, which benefits all stakeholders.

Why is original order of preference sacrosanct?

There are many students who want to change their order of preference after the allotment process has started. This happens because they normally do not have adequate time to do a proper evaluation of all options and do fill up some of the options that they later regret. While we may not be sympathetic to such students, if allowing change would cause problems for our process and create confusion, but if nothing like that would happen, could we not allow some change of order of preferences?

Changing the original order of preference could lead to small problems. Let me give an example. Suppose JEE Rank 1 has given CSE at IIT Kanpur as the first option initially and received that admission offer. Now, she wants to change it to CSE at IIT Bombay after knowing that everyone prefers that IIT. If we allow this, in the next round, this girl will get CSE at IIT Bombay, and someone who had been offered CSE at IITB in the first round would be told, sorry, we are withdrawing our offer. A naive implementation would do this. So we need to be careful in implementation. No one should get a lower preference in any subsequent round. Once an offer of admission has been made, in future, the student must either remain with that admission offer or get an offer which was a higher preference. Under no circumstances can one be given a lower preference in a later round.

But does it mean that there can be no change in original order at all, and all changes will cause such problems. Certainly not. First of all, the logic of allotment can be changed to ensure that the student in our example above gets CSE at IITB only if a seat gets vacated under normal circumstances, and not otherwise. The logic will be a bit complicated but not something that can not be coded.

However, somethings can still be allowed without getting into such problems. If we consider a student who has received an admission offer to a program which was 10th in the preference list. At this time, any re-ordering of preferences within the top 9 preferences would not cause any difficulty or confusion in the subsequent rounds. The problem of what we describe above in the example of JEE 1 student happens when we want to shift any of the lower preferences to higher preferences. So if we are shifting what was earlier at #11 in this example to #9, we could potentially get into the problem.

As I state above, we could rewrite the logic of allotment to handle this as well, but if those managing counseling feel uncomfortable with that, at the very least we can do two thing that I have stated above:

1. Allow deletion of any option at any time, including when it is being offered.
2. Allow reordering of options among those which are higher than the currently allotted option.

The first one would enhance the efficiency of the process tremendously which is beneficial to all stakeholders. The second one would allow rethink on part of students.

Of course, this has security implications. If we allow such changes, what stops a hacker to change someone's options to benefit some other student. Notice that in the current system, for sliding or freezing, the student is required to be physically present in some location. We could do the same thing for any change in the list. You have to be physically present. Of course, that would mean increased costs. A person has to receive a form, verify identity, and then allow. To ensure that such changes are not done routinely and very frequently, one may keep some price for such a change. (And of course, the security implications are exaggerated. When we can allow banking transactions based on OTP and other second factor authentication, we could implement something similar for these changes as well.)

Saturday, July 9, 2016

Admissions in Delhi University

While I normally write about admissions to Engineering colleges, this summer, being in Delhi, and reading all the Delhi newspapers, I realize that an average person in Delhi cares much less about IITs or Engineering admissions in general, and the real excitement is about admissions to Delhi University. So I tried to read up and find out about the issues and it appears that there are some serious issues there too, and the debate is unfortunately, not based on facts.

There is a fear in Delhi that next year, good colleges affiliated to Delhi University would not admit students who have done their 12th class from Delhi. Some people feel this is an exaggerated fear, that the Delhi Chief Minister is unnecessarily creating a big issue out of nothing, purely for political reasons. So let us understand where is this fear coming from.

In a recent blog article "The incredible unconstitutional admissions process in Delhi University" the author has pointed out that the TN board has more than 20,000 students who have got more than 97% marks. The total number of seats in Elite colleges affiliated to DU are about 2,000. So assuming that many students in TN are interested in seeking admission to elite colleges in Delhi, there would be no seat left for Delhi students, since the number of students in Delhi with above 97% is tiny. And we have not yet looked at number of students from other "liberal" boards like AP and Telangana.

Why is this problem being raised this year. Well, the story is that there is this one school in Erode, whose principal has been encouraging his students to apply to elite colleges affiliated to DU, particularly SRCC in the recent past and many students from that school were actually studying in SRCC. This year, his secret of success was leaked out and students from many schools have applied to SRCC. Here is one news article from Economic Times which says 80% of the students in SRCC are from Tamilnadu and some more from other states, with hardly any admissions of Delhi students.

The fear is that now that the cat is out of the bag, next year we can expect a large number of students from all the southern states to apply to all the elite colleges affiliated to DU and perhaps even the next best colleges, removing any chance of Delhi students to get admission in these colleges. And this will happen not because suddenly the south Indian students are academically better prepared than Delhi students, but because their boards are totally messed up.

I certainly believe that the fears are not exaggerated, and what is suggested in the previous paragraph is almost certain to happen, unless the admission process is changed.

In another related development, the media has detected that CBSE has been raising the marks of 12th class students by large amount in a completely arbitrary fashion. Here is a report in Economic Times. And here is my blog on the topic. Because of enhanced media scrutiny of CBSE, and the arbitrary way in which these marks have been increased so far, there is possibility that CBSE may go on backfoot, and don't increase so many marks next year, which will only increase the gap between Delhi students' marks and students of South Indian boards. And that is a perfect recipe for disaster.

Of course, one can argue that there is poetic justice in all this. For so many years, students from UP board, and Bihar board, and Rajasthan board and most eastern boards were not getting admissions in these elite colleges because CBSE was more liberal than many of these boards, and certainly one couldn't argue that in the past only Delhi students were smarter and students from other boards were not smart enough. But still, the kind of dominance that students of South Indian boards can unleash is unprecedented.

But this is where the issue becomes political. Delhi Government has demanded that either there be an admission test, or normalization of marks, or a preference/quota for Delhi students at least in colleges funded by Delhi taxpayers. Let us look at these options one by one.

Let us first look at the demand for admission test. Delhi University admits students to all the program it runs through its constituent departments through an admission test. However, only a very tiny fraction of programs run in affiliated colleges are allowed to admit students through an entrance exam. This is strange. If an entrance exam is considered good way of admitting students for DU's own programs, why is it not a good way of admitting students for programs run in DU's affiliated colleges. You can't have something good for you but the definition of good changes when it is for others.

Normalization of marks is an extremely difficult process as we have seen during the process of normalization in engineering colleges. However, the difficulty in engineering admissions was also to somehow have normalization which not just normalizes the marks but also the boards. So the same performance in a better board should be considered higher than the same performance in weaker board. And that makes it impossible to come up with a good normalization scheme. If we forget about normalizing boards, and only try normalizing marks, there can be simpler methods that can be adopted very quickly, and would be difficult to game by different boards.

Many suggestions have come in. Some have said that we can use percentiles instead of raw marks in percent terms currently used. Another suggestion that has come from both sides of the political divide is to give an additional 5 percent weight to Delhi students. (5 percent is roughly the difference between average marks in CBSE and average marks in Tamilnadu board. We could actually have a more elaborate system where we try to match the averages of all boards or some other percentile of all boards.). I am not a statistician and wouldn't know which is the best way to normalize but at least as a layman, it appears that any such system would be better than no normalization.

The most controversial suggestion has been to reserve seats on domicile basis. Please note that the reservation for Delhi students is being demanded in only those colleges which are funded by Delhi taxpayers.

Of course, since the issue is politicized, many are quick to point out that such a thing is unheard of, not possible, never happened before, and calling Chief Minister all sorts of names for even suggesting this.

However, the detractors of this demand are completely wrong. Quota for host state is actually very common in central universities. Not only that, quota for host state (that is, Delhi) exists even in Delhi University in at least one affiliated college that I know of (and I am not including minority institutions who have more autonomy in their admissions).

So, a large number of central universities have a quota for host state even in the programs run by them (and not in affiliated colleges), and where the state government does not give a paisa of support. If that is considered alright, then a demand that there be a quota for host state not in the programs run by Delhi University, but only in programs run by affiliated colleges seems pretty reasonable. If a quota for host state is alright when the state government does not pay a single paisa, it appears that the demand for quota for host state is pretty reasonable when it is for only those colleges which are actually receiving support from the state government. And let us remember that Delhi Government has traditionally had the most liberal definition of domicile of all states. Anyone doing 12th class from a school located in Delhi is considered local, while in many states, one has to spend several years in that state, should have residence in that state, etc., for local domicile.

So what is being demanded is that we should have an admission test with no favor to Delhi students. If admission test can not be organized within a year, then some sort of normalization of marks across different boards. If experts can not agree on this also, then at least a quota for host state in a very limited way. So no quota in programs conducted by Delhi University through its constituent departments. No quota in majority of affiliated colleges which include most of the elite colleges. Only quota in those colleges which receive support from Delhi taxpayers.

I think this is a very reasonable demand particularly considering that host state quota exists in a large number of central universities even where the state governments do not support it at all.

I would go a step further and suggest that if Delhi University does not agree with this, Delhi Government should close down some colleges and convert them into campuses of Ambedkar University.

Tuesday, July 5, 2016

Marks inflation in CBSE

You are a Delhi student who did 12th class exam in PCM and English. Received 92% and are happy with it. Sorry to deflate your happiness, but you actually received 82% and CBSE arbitrarily increased your marks to 92% just so that you can compete in Delhi University admissions with boards like TN, AP, etc., where the marks distribution do not follow anything close to normal distribution that you would expect in any large exam.

It follows a story by News18 in early June where they had pointed out several anomalies in the marks distribution in CBSE.

For the last 10 years, every year, it was being pointed out that a very large fraction of CBSE students were getting 95% marks in most of the subjects. CBSE kept saying that it was only doing some minor normalization. But finally, it is out in the open. The minutes of the results committee clearly say that everyone who received marks between 80 and 95 in Maths this year was actually awarded 95 marks. And similar increases were done in a large number of papers and that is why 95 was a magical number in CBSE results. What is interesting is that despite this being pointed out every year by people who were able to download CBSE results and analyse them, it has taken us 10 years to finally get evidence from CBSE that this is indeed happening.
Consider the consequences of this. A student works very hard and gets 95 marks in mathematics. Another student is not so bright and receives 80 marks in Mathematics.  Both of them are being awarded 95 marks. In all other subjects, both perform similarly, but the not so bright student gets one mark extra in English. Now in all admissions, the not so bright student is ahead of our bright student, despite getting a total of 14 marks less. This is when the admissions and careers are decided by one mark here and there. And remember, this is arguably the best managed board in the country.

I can understand that CBSE students have to compete with boards who have no qualms about giving very liberal marks to all their students. I can understand that increasing raw marks is one way to do that, and it is suspected that all boards are doing that. But shouldn't there be a better method to increase raw marks which maintains the relative positions of the students to some extent. Any such "moderation" would distort marks, but could we not ensure that people lost advantage of only a few marks in the process and not such a huge number of marks.
And this is what shocks me. That the Results Committee of as prestigious a board as CBSE does not have anyone who understands basic statistics, even school level statistics. And such a bunch is deciding the future of our kids.

So, it is not just Bihar board that is to be blamed. Most of the boards in the country are doing things which are academically stupid. Look at Telangana board. 20% of the students have received 95% marks or higher in class 12th this year. And of course, AP board will have to learn to compete with Telangana. They used to be the most liberal in earlier days. In 2016, they are the second most liberal graders with 80th percentile being 94% marks.

The school education in the country has collapsed. Out of 3 crore babies born in a year, no more than 3 lakh students would have received quality education and would be well prepared to take the challenges of a globalized world. May be a few lakhs more will learn on their own and manage decent careers. By doing what our boards are doing, we are not just risking the future of our kids, but risking the future of our nation, as the two are one and the same.

The blog that talks about CBSE marks inflation over the last decade and more is linked below.