Plan for the Lit Review

Deadline for the Lit Review is coming up soon. i’ve been thinking about how to attack it. i want to present a research program in the form of a lit review. Roughly (very), here is my view of my ‘research program’ thus far:

I. The Border Effect
I want to summarize the vast literature now on the border effect. Recent papers i’ve read/seen here are Lipsey and Swendenborg (NBER 13239, July 2007) “Explaining Product Price Differences Across Countries”
Gorodnichenko and Tesar “Border Effect or Country Effect” mimeo Aug 2006
Bergin and Glick “Global Price Dispersion: Are Prices Converging or Diverging” mimeo Sept 2006
Anderson and Van Wincoop “Trade Costs” Journal of Economic Literature Sept 2004

What i want to say: The Border Effect is confounded by the distribution of prices. From Bergin and Glick, there is a U-shaped pattern in price dispersion. they suggest oil, but i want to suggest industry dynamics instead.
Lipsey adn Swedenborg. They seek to explain price diversion as well. they find, similar to Gordornichenko and Tesar that the country fixed effects are very important (other than income per cap). so there is something happening at the country level. i think its industry dynamics, interacting w domestic demand side.

II. Multiproduct Firms
There are lots of stuff here. Mostly by Bernard and jensen, schott and kortum. Brambilla and weinstein too. Outline the empirical results

III. Trade and Firm Productivity
A. Begin with the theory, say melitz (2003)
B. Empirical — Bernard and jensen (1999), with the newest Trefler (2007).
C. Connect Multiproduct Firms and Firm productivity theory and empirics from A and B.

i think part I can be cut, if the choice is between a tight review of lit and an exploration of my ideas. i think i can write another document detailing what i want to do. But for the purposes of this exercise, i think i can come up w a decent review on II and III.

There is also a IV:
IV: Product Cycles and development
Industry structure has something to say about economic development and macro phenomena. the range of industries a firm can enter is wider is developed countries. this is an extension of the product cycles/quality ladder literature that is out there, but here the ladder is vertical AND horizontal. Macro phenomena, the business cycle and prices (linking back to part I, which is the border effect and price setting), can be linked with multiproduct firm, industry dynamics and reactions to shocks, national price setting etc.

yeah, in the midst of writing this, i’ll probably focus on II and III to come up with a more structured review of lit. develop my ideas in I and IV next week to show to some classmates and advisors. One baby step at a time…

Trade is… Good?

An interesting article called Robert Driskill’s “Deconstructing Free Trade” has come to my attention, and I simply had to write about it. The seed for the paper is wide-spread claim in the economics profession that free trade is always good.

On what is this claim based, i.e. the ‘good’-ness of trade? Any nuanced textbook on trade hits on the fact that trade good for the country, even as it may hurts some people within it. What does ‘good for the country’ mean? How can something be good for everyone? This is one of the fundamental issues in economics, and as a student of economics, I can say that it is very murky indeed.

What it is meant by ‘good for the country’ is that there exits a hypothetical redistribution of goods where, when implemented, each member of this society is better off according to her own preferences. The technical term here is that there are Potential Pareto Improvements.

Note that this criterion is (trivially) satisfied when everyone is already has an allocation in which each member is better off. The hypothetical redistribution is no redistribution.

Note as well that this is a hypothetical distribution. In many newer writings, this hypothetical distribution has been upgraded to US programs that seek to retrain workers or otherwise compensate them. However, for a long time, the mere existence of a distribution was sufficient.

Why is a hypothetical distribution good enough, if indeed it is? Is this evidence that all the mean-spirited digs about how dense economists are true?

Driskill counts three reasons, from a textbook by Krugman and Obstfeld, but can be distilled into two. The first reason is every economic change creates winners and losers. If we implement distributional plans everytime economic change happens, its likely no change would happen. This is a strange reason, because it implies that the costs of redistribution might well outweigh the benefits gained from the redistribution. In addition, another more forgiving reading of this argument is that trade liberalization and compensation schemes must always be incorporated as a policy package. Likewise, we have no reason why such schemes are feasible or on net beneficial.

The second reason is that the losers in trade liberalization are fewer in number and can organize a defense against the vast majority of consumers that are benefited. This seems convincing, because it has an air of plausibility. How many times have we seen lobbyists spending millions to protect their industries from competition? However, what many fail to point out is that this argument depends on the existence of over-all gains and the only gains we possess are promise of ‘potential’ gains only.

At its source is a fallacy of composition. What many trade models, such as the ricardian model, prove is that when individuals trade no one must be worse off. This is naturally a tautology – you will not trade you get wor. Hence, if we observe trade, at least one must be better off. This argument doesn’t carry to the level of society. At the level of society, some win and some lose – so we must consider transfers and re-distribution. That is where the problem lies.

Product Creation and Destruction, More Empirical Evidence from Market Research Data

More empirical evidence on product creation and destruction from Christian Broda and David Weinstein (BW).

They have an impressive dataset from AC Nielsen (which, my professor asked Prof Weinstein about, and she said it costs alot of money), which is a demographically balanced sample of households in 23 US cities. This data covers 40% of all expenditure in the CPI.

What are their findings?

1) Multi-product firms are the norm. The average firm sells 8 different UPCs (Universal Product Codes). The distribution of UPCs is skewed, with a large number of firms having small number of products.

2) The vast majority of product creation and destruction occurs within the firm. Over the last 4 years, 82% of product creation is done within firms, 87% for product destruction. ‘Creation’ is the ratio of value of new UPCs to Total Value. Destruction is the ratio of value of disappearing UPCs to total value.

Also,40% of household expenditures are in goods that were created in the last 4 years, 20% are in goods that disappear in the next 4. In BW, this is products in terms of UPCs, so it includes packaging and size as well. This is the most disaggregated data i have seen in an economics paper.

3) Product Creation is Strongly Pro-Cyclical — that is, it positively co-varies with aggregate measures of consumption. Consider the graph below:

In the recession of 2001, sales dropped and so did new product creation. In another graph, they document that destruction is  weakly counter-cyclical. During the same recession, destruction rose.

Why do Firms switch products? BRS (2006) Part IV

This is THE big question. BRS consider product specific, firm specific, and product-firm specific reasons.

PRODUCT SPECIFIC

Consider a demand shock that affects the demand for a good, and that every firm sees. Intuitively, this could be the ‘hot product’, so all firms climb on the hot product and all firms drop the ‘cold product’. This implies a negative correlation between adding and dropping products.

Note figure 1, which has data points on 5-digit SIC. It shows a positive correlation!

FIRM SPECIFIC

Lets say a firm shock that is common to all products makes it profitable to add products. However, this does not square with the fact that most firms add and drop products simultneously

Hence… we are left with product-firm specific reasons for product switching, or at the very least, separate firm and product shocks happening simultaneously. BRS check the later hypothesis, using separate firm and product fixed effects, and finds the R2 to be about 50%. Hence, there is room for firm-product specific interactions.

FIRM-PRODUCT SPECIFIC.

Examples of these are firm shocks (say managerial time) that affect products differentially. Then this would induce simultaneous adding and dropping. In the literature, we have the inelastic supply of managerial time.

Another way is for product shocks that affect firms differentially. An example would be changes in the products skill intensity, which would lead firms to (say) add more skill-intense products and drop relatively less skill-intensive products.

There is an existing literature on sunk costs that idiosyncratic firm productivity (i.e. firm heterogeneity). This literature involves firm dynamics — i.e. firm entry and exit. Clearly, the same modeling technique can be used for product adding and dropping. This modeling technique has equilibrium implication that the flow of firm’s adding the product must equal the flow dropping the product. This gives us a positive correlation between add and drop rates.

Another prediction: firms retaining a product should be more productive that those that choose the drop it. Dropped products should be products that have relatively low output (BRS calls this firm’s relative product size) and that haven’t been produced for as long (firm’s relative product tenure). They run a regression with dropped product rates (product-firm-time indexed) as the dependent variable, with tenure and product size as X variables (and other controls). Tenure and product size are negatively related to drop rates even after controlling for firm fixed effects, as expected.

Finally, consider that firm productivity in one market is positively correlated with productivity in other markets. This leads us to the conclusion that productivity leads to adding products. BRS investigates this for single product firms, and finds that more productive firms — labor productivity of TFP– add products and thus becomes multi-product firms.

They focus on single product firms because of measuring multiproduct TFP. Also, this highlights a selection effect — relatively more productive firms are selected to become multi-product firms, and its not the fact that they are multiproduct that causes firms to be more productive.

But not all is settled. One question is the breadth of the product mix. You can imagine that the probability of breaking into a sector is independently distributed among sectors, if we believe the firm heterogeneity story. But we find, in Table 15, that some sectors naturally go together (higher than what is predicted under the Null of independent distribution) , and other sectors don’t. Of course, this is the more natural prediction. But, how does this come about then?

BRS  mentions the relationship between firm growth and the extensive margin as future topics of interest in the theory of the firm. Also, the distribution of products within firms and the distribution of output for a firm’s products don’t obey Zipf’s Law.

Lastly, the source of why firms add and drop products simultaneously is unclear, hidden behind the ‘firm-product specific shocks’. How exactly are these shocks unleashed? BRS discounts shocks to human capital (i.e. managerial capacity) because most of the switching occurs within a fim’s existing plants. If it were managerial capacity, we’d expect more of the product switching to occur via acquisitions of other firms plants, as managerial skill is a firm specific factor. Hence, a new theory  of the firm should also explain within firm (re)allocation of resources that involves adding/dropping of products.

Impact of Product Switching BRS (2006) Part III

This is a goldmine of a paper, empirically speaking. This data needs a theory to explain it. What are the rest of the empirical points that this paper is trying to bring out?

1. Most Firms add/drop products (68%), 47% enter and exit industries, 21% enter and exit sectors. This is just a simple percentage — if we weight it by output, the numbers becomes 93% of output weighted firms add/drop products over the 25 year period.
2. The average share of output of the average multi-product firm for recently added products (last 5 years) is 49% and for products that will eventually get dropped (next 5 years) is 46%.

To me, number 2 is not so surprising, considering the sheer volume of adding and dropping going on. But it does show that extensive margin activity accounts for alot of firm output.

3. Does output, output/worker, employment, wages, TFP move with adding and dropping of products? They run a 5 different regressions with the above as dependent variables, with time and industry fixed effects. BRS finds their Table 10:

BRS stresses that labor productivity is positive for product line adjustments, but TFP is positive only for adding, negative for dropping products. Note thatfor labor productivity and TFP, when firms both drop and add, the effect is not significant (typo for TFP!, that standard errer is 2x the coefficient). Is it just noisy?

4.How do firms change its product line? Does it use its own facilities, or does it acquire a plant from another firm? BRS finds that the vast majority (94% for product adding and product dropping) are from existing Plants.

However, those firms who do acquire new plants are more likely to expand their product lines. This makes sense.

6AM

It was dark and cold. I had gotten used to it, as much as anyone might get used to lack of light and feeling in one’s extremities. That is to say, not at all. I switch on the light. A solitary florescent bulb illuminates, but gives no warmth. Not a problem. I stumble to fill a tiny, rusted kettle with water and to breathe heat into the gas-fired oven. I take the matches from the shelf and with a deft gesture, i flick a stick onto the pad and light the oven. I’m amazed what 13 years of practice can do.

I sit in front of the stove, waiting expectantly for the water to come to a boil. Its not true, what they say about a watched pot. I wanted to wait — seven minutes and 48 seconds, give or take ten, to see the magic happen. But i think the better of it and start lighting up the other burner, and work on the three eggs and leftover longanisa with practiced ease.

Riding the Wave

In an earlier post on Rose Byrne, i said that deliberately taking a break from an acting ‘wave’ can be extremely costly. Looks like my armchair theorizing was right on the money:

RIDING THE WAVE: “Damages” star Rose Byrne reports the series’ company is down to four more episodes to shoot of the hit Glenn Close-Ted Danson legal thriller for the season, and “right now I’m so immersed I can’t quite see the light. I might spend some time in L.A. and try to get a film gig. I might take a break. If we go back for a second season, that wouldn’t be till January. I haven’t had a break since before we shot ‘28 Weeks Later’” last year, says the beautiful Australian actress. “This kind of a run is incredible for an actor. I don’t want to stop because I’d be terrified that would be the end. You have to go with momentum in acting careers.”

I’m pleased that Rose is having a great run. Also, a few words on fandom. Its strange yet comforting to find a strong international community of Rose Byrne fans on the web. We all think she’s beautiful, has great style, charming, and imbued with talent. Also, for most, we seem to have been introduced to her after having seen her in ‘Wicker Park’ and ‘Troy’ in 2004. Well, for me it was more like 2005 (the Philippines gets these movies a little later than the US and other countries).

I still remember the first time i saw her. It was her guest appearance in Late Night with Conan O’Brien to promote Wicker Park. I thought to myself, ‘Who is this girl?’ She was smart and charmed the pants off of Conan and myself. The very next day, i went to the mall to rent a copy of Wicker Park.

BRS W12293 Part 2

Continuing on this most meaty of papers, BRS (2006) discusses how product switching impacts US Manufacturing. Meaning, is this something that contributes meaningfully to Manufacturing growth?

The answer is yes. In Table 6, look at the second to the last row: Total Change over 1972-1997:

To interpret: US manufacturing grew 65% over this 25-year period. Twenty-nine percent (29%) is from firms Adding and Dropping Products — the Extensive margin– (189% and -170% respectively that sum to 19%, which is about thirty percent of the total change). The rest of it comes from changes in firms intensive margins, that is growth and decline in sales in existing/continuing products.

Looking over the 5 census periods, the importance of extensive margin is somewhat stable, especially over the more recent 15 year period, compared to the more volatile growth from the intensive margin.

Also interesting to note is what the authors call “Excess Reallocation”, which basically means that there is a LOT of movement in the extensive margin. To make that 19% contribution, added products grew output by 189% and dropping products shrank output by 170%. Thats alot of shaking to get to 19%.

Looking at this excess reallocation problem, we find that its somewhat ameliorated over several 5 year periods, but still quite significant overall.

The stability of the contribution speaks to the possible existence of a neat equilibrium.

From a modelling perspective, i wonder if this can still be explained by completely rational actors? Or do i need to incorporate some information problem? Or, do i need to incorporate some kind of trembling hand perfect equilibrium where firms branch out in other sectors/industries/products via a trial and error/rule of thumb method?

Multi-Product Firms and Product Switching

More on empirical facts on firms, industry structure, and dynamics. This is a paper by Andrew Bernard (Dartmouth), Stephen Redding (LSE) and Peter Schott (Yale). These three guys are at the forefront of the new generation working in the theory of the firm. The interesting thing about this paper is their use of a confidential dataset from the US Census Bureau. They look at quinquennial data (every 5 years) from the Manufacturing Census Longitudinal Research Database between ’72 to ’97. There are 1,848 5-digit Standard Industrial Classification (SIC) codes, which BRS refer to as products. This data isn’t completely disaggregated, but it is as disaggregated as anything available from the US government.

This is plant-level data. The sample includes plants that are ‘big enough’, small plants are excluded from their analysis. They can observe the whole range of products (5-digit) of each plant. They aggregate to the firm level because product-mix decisions are done at the firm level. About 140,000 firms survive from each of the 5 census dates. There is a lot to appreciate here, so I’ll take it slow.

They define sector as the 2-digit SIC (20 sectors), industry as 4-digit SIC (459 industries), using categories defined in 1987 SIC classification system.

Looking at Table 2, there are two things to notice.
1) the number of industries and products produced in a sector vary widely.
2) Products vary widely in how they are made both BETWEEN industries, which we can see by comparing the means, and WITHIN industries, by comparing Standard Deviations. Capital intensity is highest in Petroleum, Skill Intensity is highest with Printing. I haven’t checked, but am willing to bet, that the covariance between capital intensity and skill intensity is high accross sectors.

The next interesting thing that multiproduct firms dominate US manufacturing. Firms producing in multiple industries and sectors produce 87% and 76% of output.

The average multi-industry firm operates in 3.1 industries. The average multi-sector firm operates in 2.5 sectors, and the average multiproduct firm produces 4.0 goods. While this is interesting, in addition i would like to see how whether multi-industry firm branch out to ’similar’ industries — similar either in capital intensity or skill intensity.

Lastly, we note the huge differences between multi-product/industry/sector firms and single product firms in terms of output, employment, etc. Interesting to note is that the advantage of ‘multi’ firms practically fades when it comes to TFP. Is this just because of the way they computed TFP (multi-factor superlative index number — something which i know nothing about)

MNCs and the Introduction of New Product Varieties

In Irene Brambilla’s “Multinationals, Technology, and the Introduction of Varieties of Goods”, we find interesting stylized facts about Multinationals in China.

The World Bank’s 2001 Investment Climate Survey in China has data on 1,500 firms for 5 cities (Beijing, Tianjin, Shanghai, Guangzhou, Chengdu) over 1998-2001. One thousand of these firms match with 27 3-4 digit manufacturing industries, while the rest correspond to services. These industries were chosen because they had high growth and innovation rates (Apparel, Household Appliances, Vehicles and Vehicle parts, Electronic Equipment, Electronic components).

They were asked about their introduction of new varieties over the three years. What is a new variety in this paper? They decide that a new variety is one where:
a) different from all varieties produced by the same firm, regardless of whether it is already available in the market
b) Since this is subjective to the firm, they limit this by considering only varieties that are at least 10% difference in price.

Unfortunately, the paper does not elaborate further on this classification process. I would like to know, via an example, how new varieties are determined.

The table above doesn’t include ‘outliers’ which eliminate 40% of the sample. These outliers are firms formed after 1998 and those that have more than 100 new varieties. The latter are true outliers because she reports that 97% of firms have less then 20 new varieties.

Some things to note about the table above. There seems to be a marked difference between foreign firms and domestic firms. For the table, there are 605 domestic firms adn 273 foreign firms

a) state owned private firms are large (employment), while cooperatives and fully owned foreign firms are the smallest
b) in output per worker, foreign firms are much more productive.
c) foreign firms produce higher than average number of products.

In addition, most of the foreign firms (~58%) are from Japan and Hong Kong, 17% from US and Canada, 14% from other asian countries (Singapore, Malaysia, Thailand, S Korea), and the rest from Europe. The bulk of HK firms have <50% foreign ownership, which suggests that chinese firms partner with foreign firms in HK to gain some of the investment incentives.