Kids with tuition fare worse. REALLY???

Another week, another anti-tuition article to debunk. It’s a love-hate relationship with these kinds of articles because whilst it is tantalizing picking apart the ‘analysis’, I wonder how many will be miseducated by that article.

So the author, a lecturer at the department of economics at NUS compared retrospective data collected in 2012 of Singaporean student performance in PISA (Program for International Student Assessment) and determined that students with tuition fared worse.

Readers have already commented to the inappropriate statistical analysis at the end of the article. So I will comment about the methodology from a clinical trials perspective and make it relevant to biology (since I am a H2 BIO tutor haha).

Clinical trials are always conducted prospectively. Meaning identify a novel treatment, design a trial, start enrolling patients and collect data off them. Do an analysis, back it up with statistics and then come to a conclusion. Is it useful or not?

Nobody will do a clinical trial backwards (retrospective). That is go back to the medical records analyze their situation and come to a conclusion the usefulness of a new drug or existing treatment.

The main reason for that is because the more the same set of data is analyzed the more likely one will find a relationship between an independent and dependent variable (in this case student have take tuition or not and their score in PISA). This may not be because the relationship is true but rather a chance finding. Just like if a person keep buying TOTO or 4D or lottery tickets, over a period of time there will be a chance that person will make a correct guess. Is it because the person is lucky or is it because the person has a deep insight and made the correct guess?

Also, retrospective studies are very biased and cannot be controlled. Nobody can make causative conclusions from such studies (like what the lecturer did) and can only point out possible correlations. This in turn needs to be confirmed with prospective studies. Why are retrospective studies biased? Let me give you one example.

PISA is not a universal test ALL students in Singapore take. Therefore the sample collected is not representative of the students in Singapore and their performance relative to whether they go tuition or not. Take a top tier Independent Singaporean school for example, between PISA and MENSA or the Olympiads, the latter 2 are more prestigious and time is spent preparing for it. PISA is then left to government schools. Can you see how that will skew the results?

If the author is serious about the comparison, then design one prospective study, randomized and double-blinded. Then see what the results conclude. By the way, it is not uncommon for tuition centers to brag about their students achieving gold for Olympiads (scroll to the bottom of the page).

At the same time, may I also point the author to the article mentioning the top performance of PISA in Singapore due to tuition! COMMENT.


  • On December 12, 2016