Wednesday, May 23, 2007

Do modern markets really harm the traditional ones?

Errata. I wrote "the treatment group consists of traditional markets opened between 2003-2006 which are located within 5 km from a modern market." The correct passage should be "sellers in traditional markets which are located within 5 km from a modern market(s) opened between 2003-2006." Sorry.

One one occasion, a student confronted my statement that market competition does not necessarily means the weak will have to face the strong in one battlefield. This student used the traditional vs. super/hypermarket as the evidence. To be honest, although I thought his statement (as well as others who shared this view) was still debatable, I could not provide any scientific evidence to refute his argument.

But a recent study of our colleagues at SMERU Research Institute now provides the empirical evidence. Their study shows that although it is true that sellers in traditional market has been loosing profits, it can not be attributed to direct competition with modern super/hypermarkets. The report is not yet available online (thanks to Dr. Sudarno Sumarto for letting me quote and discuss their work). But it has been quoted by a local news magazine.

They interviewed fresh food sellers in traditional markets in Greater Jakarta and Bandung area. The sample was then divided into two groups:
  • Treatment group consists of traditional markets opened between 2003-2006 which are located within 5 km from a modern market.
  • Control group consists of markets located in the same districts with that in the control group where no modern market exists within 5-km radius, but according to the regional site plan, there will be a super/hypermarket opening nearby in 2007. This last criteria is important to isolate the placement effect (the fact that a modern market is opening soon means that the location is equally attractive to modern markets).
In addition, the markets in both groups must not have been renovated after 2003 (the benchmark year in this study).

The difference-in-difference result (see below) showed that from 2003-2006, sellers in both group are losing profits. Interestingly, those in the control group (who faced no direct competition with modern markets) experienced bigger decline in percentage term. However, the difference is statistically not different from zero (but still provides no evidence that the modern markets harmed the traditional ones). The bottom row shows that in terms of earning, the control group experienced bigger decline. This may suggest that to cope with declining profits, traditional sellers tend to maximize sales rather than profit. Still, the difference is not statistically significant.
To put it in another words, this study shows that with or without modern markets, traditional markets are declining. If financially strong modern markets are not the main culprit, then what are? According to the qualitative part of the study, many sellers view street vendors (those selling similar products outside the marketplace) as more of the problem, as well as bad management.

Before we either believe or bash this study, there are some caveats needs to be taken into account:
  • First, the question of external validity: can the sample represent the population? (SMERU has acknowledged this in their study).
  • Second, as our friend Arya Gaduh pointed, although they had tried to correct the placement-endogeneity bias by only considering locations where a modern market is opening, the problem may still exist. The decision to open in 2007, not earlier, may perhaps reflect location preference.
  • Third, our co-blogger Sjamsu raised the issue of selection bias. Could it be that those who were interviewed only represents the 'winner,' while the 'losers' have already exit the business?
  • Fourth, the study was limited to fresh-food sellers. Sjamsu also referred me to a study conducted by a consulting firm that fresh food sellers are still the winner as they still have a comparative advantage: freshness. They start selling early in the morning (when modern markets are still closed). And, in the urban/suburban areas, even the most of the mid-high income still buy fresh food from the traditional markets (or ask their pembantu to do that). So perhaps the result was skewed towards the 'winners.'
Nevertheless, SMERU study has provided us a very valuable empirical evidence. More than just anecdotal or 'romanticizing' the David vs. Goliath epic in market competition.

Addendum. As presented, the numbers in parentheses from the table above are standard deviations calculated from the sample. The authors did not present the standard errors for the mean values and the DiD coefficients. However, in the appendix section of the report, the authors presented the standard error and t-statistics for the DiD coefficients. The t-statistics are 0.76 (changes in profit) and -0.32 (changes in earnings). So statistically speaking the DiD coefficients are not different from zero. The authors estimated several models, adding some controls in the regressions. The coefficients changed slightly, but the results are consistently not significant.

14 comments:

  1. Are the figures in parenthesis standard errors? It seems to me that they are very large.

    The standard errors of the difference-in-differences perhaps would be large too.

    ReplyDelete
  2. They are standard deviation, not standard errors. At least it was what appears in my version.

    ReplyDelete
  3. In here, I found 5 new traditional market places being opened operating every weekend around the city centre promoted by the councils (in 2 cities). And 3 big supermarkets being closed down in the suburbs due to the competition from other supermarkets. I don't know what the economic story behind this.

    The sellers are mainly Chinese migrants, Maori and Pacific Islanders and the products are mainly vegetables and fish. The buyers are mostly Asians (Indian, Chinese, Indonesian and Malaysian) and a few Europeans who want to get cheaper prices.

    As a migrant, I feel being offended with this as seemingly the councils are trying to create some kind of positioning of our market. But, it's true as we can save more to pay our bills, even though we lose our prestige.

    ReplyDelete
  4. In any case, it would be helpful to know the confidence interval of the difference-in-differences, I guess.

    ReplyDelete
  5. Anymatters, that's very interesting. I'd love to know more about the story. But it seems to confirm that everywhere, the authority really likes to step in. As this weird guy worries, market is not market anymore... (too bad that guy doesn't seem to know English)

    ReplyDelete
  6. "Treatment group" consists of traditional markets built between 2003-2006, and I assume, these are markets built AFTER the modern market competitors are built.

    This implies that when the markets were built, sellers who decided to sell in these markets had taken the modern-market competitor(s) into their calculation of whether to join (or not to join) the traditional markets. Only when participation ensures good enough return will they join the treatment-group traditional markets.

    (BTW, for the reason cited by Sjamsu on the "winner effect", the same problem remains even if they survey all traditional markets -- not just ones built between 2003-2006).

    As such, the result is probably not surprising: In a sense, participants in the treatment-group markets are facing the same "equilibrium" as the control group, and therefore, the Diff-in-Diff results should not be significantly different.

    The result of this study can say something about the effect of modern markets on traditional markets iff the DiD was done on data surveyed BEFORE and AFTER a modern market is erected.

    Having said this, I am not even sure that the study is asking the right question. Suppose, for instance, the study did find that (delta) profits are significantly lower in the treatment group -- so what?

    It is, after all, to be expected: More competition (whether from modern markets or from another traditional/mom-and-pop stores) will cut into incumbents' (read: existing traditional markets') profits. Should we worry? Well, only if we sell in these traditional markets. Otherwise, we should be happy that we're gaining consumer surplus.

    Which is why I think this study is much ado about nothing.

    ReplyDelete
  7. Some of the study design described by a.p. are wrong. Arya, read the whole paper carefully before commenting. Don't base your comments on the review put here.

    Furthermore, next time, perhaps it would be wise for one to wait until the paper is available online before reviewing it in a blog. Or, when one is already itching to put the review on public space, make sure one is quoting correctly.

    In the interest of complete information: I am closely related to the study.

    ReplyDelete
  8. Daniel,
    Unfortunately, this post is the only version available to me (or, for that matter, to the public). I can only comment based on this review alone, which -- given my lack of access to the paper--I can only assume to be accurate.

    At any case, can you tell us what is the evaluation design really like?

    ReplyDelete
  9. I guess the modern markets do harm the traditional markets. They probably are still facing each other in the one battlefield at this moment.

    But I suppose sooner or later each market will find its own battlefield to fight with others from its kind. There will be market for modern market and another market for traditional market.

    Just like what anymatters described. Or like Wal-Mart did to other traditional-modern supermarkets.

    ReplyDelete
  10. I have not read the paper yet. However, regarding Sjamsu’s point, I think sort of procedure which may overcome this problem. So we need not care very much about selection bias in situation where data suggests that relative margin between prices of goods sold and their marginal costs in traditional markets do not vary within a district (this is not assumption but supposedly based on data). In the end, profit is only a function of goods sold.

    So, if modern markets had decreased profit and goods sold in traditional markets, we could (still) predict whether there are losers or not by replacing observation from those markets which are free from modern markets and imposing the magnitude of the impact on these samples-negative profit means there are the losers out of samples in treatment group. One implicit assumption we need, that is distribution of sellers (big sellers and small sellers within traditional markets) are normally distributed in each traditional market.

    Surely, this is not the best, but this is at least what can be done. Hence, at this point I disagree with Arya- surveying all traditional markets will make a significant difference, particularly in concluding whether modern markets really create the losers or not in traditional markets.

    But, I predict that this margin seems to vary since there are many kinds of traditional markets within a district. For example, price margin in primary traditional markets (pasar induk) is lower than that of secondary traditional markets. Then the magnitude of the impact of supermarket on each type of traditional markets will also vary significantly—which makes this study more complicated.

    I agree with Arya’s point. The issue should be emphasized on consumer surplus. But it also worth noting that rent-seeking practices by local government in traditional markets make modern market in favorable situation.

    ReplyDelete
  11. Daniel,
    I take the responsibility of reviewing the study here. Yes, I am "already itching to put the review on public space," because I found this is a great study, in terms topic and methodology

    Yes, I am aware that it hasn't been made online. But given that a version of it has been circulated (and even discussed in local media) without any disclaimer ("draft-do not quote" thing), I considered it as a public goods.

    I sincerely apologize for doing such thing, or if I violate any codes of conducts.

    Of course, there may be 'lost-in-translation' problems in translating a method-intensive study into popular language. Any errors are not intentional. And I hope this will be a chance to learn more about "some study designs I described" that was wrong.

    Or did I misquote some of the findings?

    I also take the responsibility of quoting some comments for the study. Again, the comments may not be correct. But I found them still valid, at least to raise and be discussed academically. And certainly it would be very great to here your reply to the comments.

    ReplyDelete
  12. I'll comment on the sampling frame, since it's the animal most people are interested on. The usual caveat applies, since this paper is still unpublished.

    The treatment group consists of long-established traditional markets, near which a supermarket had been opened between 2003 and 2006. The respondents are established traders, who opened their business before 2003. They were sampled on a PPS basis, which basically means they represent the markets.

    Oh, traders selling non-food or ready-to-eat items were excluded, but all other food sellers are included. So, not only fresh food sellers.

    For the control, a.p.'s description is already correct.

    To circumvent the problem of excluding the losers, the survey also asked the respondents about those leaving the market. I think their responses allowed us to dispel this problem.

    For more detailed results and sampling frame, the paper will definitely be put on SMERU's website for free download. Tunggu tanggal mainnya!

    In the mean time, there are other interesting papers on SMERU's website. One on unemployment, the other on private sector growth. For those ones, I don't mind entertaining criticisms more openly.

    ReplyDelete
  13. Thanks for the clarification, Daniel!

    I agree that you have had a proper sampling frame. I raised the issue of external validity because it's a standard question we should keep on asking (and debating) in any randomization-based study.

    I admit I used the term 'fresh food sellers' incorrectly. This is an example of 'lost-in-translation.'

    For the winners/loser inclusion/exclusion issue -- Sjamsu and I actually were thinking in the 'Heckman selection bias' framework. Although I don't know whether applying it will add any significance that vs. what you've applied in the study.

    And of course, I will review more SMERU papers!

    ReplyDelete
  14. Who wants to bet on the diminishing of traditional market? I do.

    ReplyDelete