Tuesday, May 30, 2006

Putnam confirmed

In "Making Democracy Works" (1994), Robert Putnam shows how civic community matters in governance and democracy. He made Northern vs. Southern Italy has his case. Then, in "Bowling Alone" (2001) he wrote that civic participation has been declining in the U.S. TV and radio were to blame for that.

Scholars have been disagreeing over his work. I will leave the disagreement for another discussion. One interesting question to ask would be "how true is the claim that TV and radio contributed to declining civic participation?"

An interesting paper by Ben Olken tried to answer that using survey data from more than 600 villages in Centarl and East Java. The methodology is very interesting. He looked at the number of TV channels can be reveived in each village. Based on Putnam's work, he hypothesized that the more channels can be received, the lower the level of civic participation. Since Putnam also argued that level of civic participation correlate with governance, more channels should also negatively correlated with quality of governance.

But a simple linear regression suffers from reverse causality and omitted variable bias. There may be other factors correlated with channels reception and participation as well as governance. To deal with this problem, Olken exploited the exogenous factor that affects channel reception: geography of each village. Villages surrounded or near the mountain will receive less channels (perhaps only TVRI and RCTI). So he uses the information on several determinants of channel reception in each village (geography, topography, relative position from nearest transmitter etc.) as the instrument.

The IV regression results were fascinating. Number of channels is negatively correlated with participation in social groups (community meeting, gotong-royong, arisan or religious groups), trust (other measure of social capital), and "missing expenditure" (as proxy for corruption and governance).

So, Putnam's theory is confirmed then?



  1. It's one thing to claim that TV-watching is related to social participation. And, it's really not surprising that the relationship is negative: the more you spent watching TV the less time you have for gossipping around with neighbours. The day is only 24 hours, never more, as far as I remember.

    But putting "social capital" into growth estimation is another thing. Claiming that social participation has positive impact on economic growth is a little too much. I think Ben Olken realizes this and that's why he rightly doesn't touch on that issue.

    I wouldn't be surprised if any econometric exercise trying to put the so-called "social capital" into growth estimation turns out insignificant. At least if the variable used is "the number of organizations you are in", "the number of social meetings you attend" or the likes.

    I am skeptical with growth estimation with a "social capital" variable in it mainly due to this issue of measurement and definition.

    Because I think any trade-related variable has done a good job in explaining social capital. That is, social capital is embedded in trade. (Again this is a matter of definition, you may not agree with me if you define social capital as in my anecdotal examples above).

    Using participation rate in social association as a measure of social capital is meaningless. I can register to hundreds of associations or attend 50 meetings a day. But so what?

    I think the word "social" always refers to the context where more than one person is involved. Think about trade. Trade is and always is an act of more than one person. If you and I trade, we are engaged in "social activity". That is, to my own(?) definition, "social capital" -- if I really have to use that phrase. Therefore we don't need any new variable (and headache of finding a good proxy).

    How to increase "social capital" (my "social capital") then? Free trade. The more you restrict trade, the less interaction among economic agents you would expect. And the less growth you would get.

  2. I agree it's meaningless if we use (cross-country) growth rate as the RHS variable. even though it's significant, it's till meaningless. In fact, most of the cross-country growth regressions do not actually have real meaning apart from explaining correlation, aren't they (but that's a different issue).

    To be fair to Putnam (I'm not really a fan of him either) and all his critics as well as followers -- the debate goes beyond whether the rel'ship b/w TV and social participation is negative. The questions are:
    1) is it true? [perhaps not, when you're watching news channel, you may be politically more engaged]
    2) is there a causality? is TV really causing the decline?
    3) regardless of the cause, if civic participation IS declining, what does it mean for governance (corruption, monitoring), democracy etc.?

    (yes, yes, at UI we call this 'non-economic factors in development').

    as for participation as measure of proxy -- again depends on what RHS variable we are talking. National GDP growth is meaningless. but "prob of getting govt help," "prob of getting job," "success rate in adopting new variety" may be relevant. not saying that the relationship will be significant, but it is a relevant issue.

    and sure the concept of diminishing returns still apply. enrolling in 100 organizations may not be more helpful than just 2. But participating in 2 community meetings may make difference than none.

    btw, surely we also have problems in measuring 'human capital.' and how to measure 'physical capital' has long been a mystery in economics... :-)

  3. Well, it's been almost 40 years since Mincer's article on human capital appeared in the JPE and it's already a little over 30 years since Becker's Human Capital was published. I am pretty comfortable with what economists have come to define as human capital and how we try measure it. How long will it take for social capital?

  4. Ujang -- I meant it in practical terms. I believe, and accept, the definition and measurement of HC. But still we can always question whether it should be measured on years of education, years of experience (and the squares), or aptitude test score, or something else?

    (and we know some data sets don't have exact years of educ, experience, so we will have to construct them).

    in the macro level, the variable of a country's HC usually measured in % enrolled in elementary, high school, % engineers etc. but there are puzzles where education matters privately, but can't explain well cross-country growth or income differentials.

    and enter the Spence model: does education really increase your human capital? or is it just a signaling mechanism?

    but I got your point. and I personally feel that 'social capital' is treated as a black box. I don't think we should really find the overall proxy of it. rather, we can use some parts of it to analyze some (un)interesting things like:
    - does attending pengajian increase our happiness?
    - does attending arisan provide us mechanism to smooth consumption?
    - does being active in Posyandu really improve children's health?
    - does high village meeting attendance correlate with low corruption?
    - etc.

  5. I agree that the variables we usually use to measure education have some limitations, but by now we are already familiar with those limitations and the ways to circumvent them. We've also come a long way in terms of using various objective health indicators as measures of human capital in health. From theoretical perspectives, we now know enough about education and health and how to incorporate them into household production function, etc. We learn about Spence's signalling model and some ways to test it. And so on. That's okay. That's what decades of accumulated research bought us.

    I'm arguing that we know much less (much being the operative word) about the mechanisms through which social capital works, what are the channels, where does social capital come from (predetermined? endogenous?), etc. As you said it yourself, it's still a black box. Aco provides a good example in his comment above, if trade may already have social capital embedded in it, how useful is the concept of social capital in that case? The fact that we can do better by focusing on smaller parts like you are suggesting - arisan, pengajian, village meeting - makes you wonder: maybe these things do operate very differently. But then why do we need such broad concept as social capital in the first place?

    As Durlauf (2002) and some others have argued, so many phenomena can be linked to trust and cooperation , and so many times the ill-defined concept of social capital may lead to functional and nonfalsifiable claims. For example, whenever cooperative behavior occured (such as trade) it must be because there is social capital. See also this paper where Durlauf lays out the daunting econometric problems faced by anyone who wants to do the empirical work properly.

    That is far from saying that the empirical research on social capital so far have been useless. On the contrary, I agree that there are a lot of new insights that we can learn from. But in our profession these kind of things tend to progress slowly, and there's always some danger in overselling statistically significant results. In my opinion, a question mark at the end of this posting's title may be more appropriate ;-)

    Newsflash: Hold on! After reading Olken's paper, I suggest that we all end this discussion and call it a day. I was looking in particular, at the acknowledgement on the very first page:

    "...Adam Smith and Zejd Muhammad provided outstanding research assistance..."

    With a research assitant of that calibre....

  6. LOL, i was going to say something then i reached the end to discover that the shop's closed. Ha!

    I do find this very interesting to most of my references come from soc studies and not economic. It's an interesting subject... wish you guys come up with this stuff more often and not so much of the other things i can't comprehend (puny brain, fault's mine LOL).

    In any case, back to TV and social capital, there're lots of people also arguing that TV/Radio/Internet creates soc capital in an entirely new way and thus measuring participation in the immediate vicinity (traditional society) is less relevant. Examples would be international fund raising, cross border activism, etc. new tech and improved communication simply allow society to network and related in different way. To do this empirically, i think, is probably still just as difficult...

  7. just a disclaimer and clarification: I wasn't advocating that we adopt the term social capital, and see it as having a similar role as human and physical capital. I just used the term because de facto, it is what is used now.

    as Becker once said, economics has powerful tools and methodology. and other social sciences have interesting questions. that says, the debate is not whether we should or should not use the term social capital. but whether researches like Olken's are valuable contribution for scholarly debates, at least in terms of methodology.

    I agree, I should have put the question mark in the title (i did put it in the end, though...).

  8. i never heard putnam until just now (hehe), but at a glance is hard to convince me that TV and radio create decline in civic participation. (what does he means by TV and radio by the way? is it means: technology?)

    extrapolating to the current lifestyle, a new medium is here: the Internet, but from what I experience, instead of declining there are many examples of a blossoming participation and new activism on it.