If you want to bash up Nehru, do it properly.

(I am taking this back. See here)
I was browsing through the Blog Mela posts and one of the entries over there is by Shyam Nair who is trying to bash up Nehru.

Now that is an entirely commendable activity and I can only wish that more and more young men take up the hobby, but you should keep up some standards. For example, you should not approvingly link to Rajeev Srinivasan’s article on what he calls the “Nehurvian Penalty” that is, I am afraid, an embarrassment to anyone who understands numbers, as I have explained.

13 thoughts on “If you want to bash up Nehru, do it properly.”

  3. If I understand you argument correctly:
    The model proposes a certain GDP growth (7%) and compares this with the actual growth (3%) and deems the difference to be the Nehruvian penalty.
    This is incorrect because the difference could be an error in the model itself; and that one can’t use a hypothetical number for comparison.

    But hey – the model when applied to other nations (which followed the policies-it-advocates) seems to predict properly. So, given that empirical evidence for the model, the number of 7% isnt so “picked-from-a-hat” is it? The comparison, while not writable-on-stone, at the very least, does not seem like an an embarrassment to anyone who understands numbers

  4. Come on. You’ve not understood my argument correctly. That’s Rajeev Srinivasan’s misunderstanding that you are “understanding”. The article itself says that they were *testing* their model by “running it backwards for the period 1960-2000”. Please note: Testing the model.

    I’ve explained the problem in my post, but I’ll try again. You don’t test a model by comparing the results of your model with an unknown number i.e. the GDP growth that would have occurred if Nehru had followed certain policies. You test a model by comparing the results of your model with the known number, i.e. the GDP growth rate that actually took place. The model is supposed to predict the output (i.e. GDP growth rate), given certain inputs (such as macroeconomic policies literacy rate, corruption index, etc.)

    Now if you want to test the model you have to plug in the values for the inputs as they existed in 1960 and see if the model gives as its results, the GDP growth rate that actually occurred. That’s what they did and it turns out that at least for India, the model overpredicts the GDP growth rate. 

    What you are saying is a perfectly legitimate analysis to do. You can take the non-policy variables as they actually were in 1960 and the policy variables that you are recommending, and figure out the growth rate of GDP that would have occurred if India had followed good policies. If this is higher than what actually occurred, then the difference could be legitimately called the Nehru Penalty. But that is no way to test  the model.

  5. RR: you seem fixated on Gsach’s “testing” the model on India. That was not Rajeev’s thrust and not mine either. One can “test” on other nations that did followw good policies, and predict what would happen for nations that didn’t.

    Let me state it more concretely:

    Model definition:
    1. Model, let’s call it M, predicts the growth given macro-economic variables, and assuming that the govt follows “good” policies.

    Empirical Validation:
    2. When back-applied from 1960-2000, M’s predictions were pretty close to the actual growth rates for nations which followed “good” policies i.e. which satsified the assumptions of the model.

    3. Given 1 and 2, it can be said that if India had followed “good” policies, it’s actual growth rate would be close to that predicted by the model. The “difference” could be called a penalty for bad policies.

  6. Ravi, I think you need to put out a plain English apples and bananas type example for 7*6 to get what you’re saying.

  7. In my defense the blog wasnt abt Rajeev’s article but on the state of The Planning Commision, but anyways before saying anything,what else would you deem as the reason for difference?
    schmo_OR_Pro
    Shyam

  8. Ravi, I think you need to put out a plain English apples and bananas type example for 7*6 to get what you’re saying.

    *sigh*
    My opinion of the cartel is slipping.

    Let me clarify in “plain English apples and bananas”.

    GoldManSachs applied their model to 1960-2000,
    for “testing”.
    Now, one can “view” this as:
    “testing” the model for countries which followed favorable policies like US, France, etc.
    Using the “validated” model to predict for countries like India if they had also followed favorable policies.

  9. Frankly 7*6, I don’t know how I can make it clearer. What exactly do you mean I am “fixated” on the idea of testing it? It was Rajeev who reported that they were testing the model. I’ve used his exact words. By creatively reinterpreting words you can do anything, even make Nehru seem like a supporter of liberalisation.  But if you take the plain meaning of words and use some common sense, the only interpretation that is plausible is mine.

    They said that they were “testing” it and the table column headers say quite clearly “Predicted GDP growth rate” and “Actual GDP growth rate” All the countries are given in the same table and no distinction is made between those who followed its recommendations and those who didn’t. How much clearer can it be? If they were actually doing what you imagine they are doing, they’d need another column saying “Index of closeness to recommended policies” so that they can prove that following the policies they recommend is positively correlated with high GDP growth rate.  But in that case they wouldn’t need a “Predicted GDP growth rate” because the correlation between the “Index of closeness to recommended policies” and “Actual GDP growth rate” would speak for itself.

    Try applying the Occam’s razor once in a while. The simplest explanation is that Rajeev made a thoughtless error.

  10. RR: Let me outline where we are disagreeing succinctly.

    See, the thing is that, as my earlier comments said and as even GoldmanSach’s report itself said, the conclusion [GDP gap due to flawed policies] is logical.
    What you are arguing is that one way of arriving at this conclusion [using a test-point] is illogical. This I agree with. But then you go ahead and postulate that Rajeev also followed the flawed deductive path you outlined.
    This I do not agree with – it fails the Occam’z razor. And is also immaterial, since the conclusion is logical anyway.

    To expand on the above para, and to answer your prev. comment:
    GoldmanSach’s point wasn’t to show a positive correlation between “Index of closeness to recommended policies” and “Actual GDP growth rate”.

    GSach’s intent was to predict the growth rate.

    This much I agree with you. So does Rajeev as you point out.

    GSach’s report itself then comments on this table that their model is validated since countries like US, etc. which followed adequate policies matched their predicted growth rate. And they postulated that the discrepancy for countries like India was due to their unfavorable policies.

    So at least we should agree that the conclusion is logical and that Nehruvian Penalty is a reasonable hypothesis.

  12. The report says:

    “For those countries where policy settings were not particularly growth-supportive — India, Brazil and Argentina — actual growth fell below what we would have projected.”

    So, going by Ravikiran’s argument, the authors of the report themselves didn’t get it right! Rajiv Srinivasan’s fault then only is that he accepted their claims uncritically. 🙂

    However, the authors’ claims do remain consistent either if

    a) the said “growth-supportive policy-settings” are not parameteres of the model, or,

    b) back-projections are made using current data, not the ’60s data; and while “growth-supportive policy-settings” are indeed used as inputs to the model, they are perhaps assumed to remain constant throughout the period for which projections are made.

Comments are closed.