20 October 2009

Water-related ag job losses much lower

About nine months ago, Richard Howitt (my dissertation adviser at UC Davis) and his coauthors estimated the impact of reduced water exports from the Sac-SJ Delta on agricultural jobs and revenues in the San Joaquin Valley. In that study,* they estimated the impact to be a $1.6-2.2 billion drop in revenues and 60,000-80,000 job losses.

Although Howitt thought that their numbers on revenues were of greater interest, it turned out that the numbers on jobs attracted much more attention. The media, politicians and farmers all seized on the higher figure** (naturally) in making their arguments that export cutbacks were resulting in terrible human suffering. This pressure was at least partially responsible for Judge Wanger's decision to take "human impacts" into consideration when enforcing the Endangered Species Act, i.e., to consider larger exports.***

In my opinion, Howitt et al.'s estimate of impacts was ripe for misunderstanding:
  • As they mentioned, markets could reduce the impacts.
  • Their estimates were confounded with high unemployment in places like Firebaugh and Mendota, where 30-40 percent unemployment is the norm, year after year, high water or low.
  • They were the result of a simulation (the impact of water on crops) fed into a simulation (the impact of crops on revenue) fed into a simulation (the impact of revenue on jobs). Read my critique of this method here.
Although I have expressed my concerns on this blog, Jeff Michael's criticism of Howitt et al.'s result has got more attention (and deservedly so). Michael's forte is regional economics, and he was not seeing any impacts in the employment data for the area. (Here's our water chat.)

Anyway, what's upset me about this whole business is not the well-meaning debate between two academics, but the way that various partisan groups have used the original 80,000 figure** to rant about "blood on the arid soil" on the one side (Hannity and Stewart on Hannity), or claims of "nothing happening here, please move along" on the other. Although I have been very critical of farmers billionaires looking for special privileges or fronting astro-turfing groups that claim to represent workers but really represent businesses, I do recognize that reduced water deliveries have reduced wealth and income in these communities.

(I am really sad that the fight has turned into "pumps on" vs. "pumps off." First, the pumps ARE on. Second, water markets could have reduced the economic and social cost of shortages.)

Right -- so that's some background...

In this revision [PDF] to the revision**** to their original study, Howitt et al. admit to using some flawed data in the first study and estimate 21,000 total job losses, with 16,000 from drought and 5,000 from pumping restrictions.

Although their note is pretty geeky (IMPLAN vs. REMI vs. SWAP simulation methods and detailed comparisons with Michael's estimate of 6,000 lost jobs, of which 2,000 are from pumping restrictions), and I still dislike simulations (the world is too complex), I am happy to see Howitt et al.'s efforts to improve accuracy.

And yet, I have the following gripes:
  • They claim that county data are inaccurate because "impacts on the west side of Fresno county might be greater than on the east side." After several emails with Howitt, I do not understand this problem. I do understand that their model resolution is better than county-wide, and that their statistics are more "granular" than county EDD statistics on employment, but the numbers should match (in the medium term). If we want to talk about data problems, we can go on for days!
  • It does seem as if Howitt et al. are defending assumptions when calibrating and interpreting their model that contribute to higher job losses. (It's obvious that they have an incentive to do so, since they -- like everyone -- prefer that their new conclusions support their old ones.) Of course, Michael appears to be doing the same, in the opposite direction. He averages his estimated impact of 12,000 losses and a "no impact" lower bound of zero losses to arrive at a final figure of 6,000 losses.
Going forward, I have a few things to say:
  • Partisans will use the numbers that serve their interest, i.e., we will still see the inaccurate and disowned 80,000 figure.**
  • 21,000 lost jobs is bad but remember that those are only about one percent of the region's 1.7 million jobs (in a population of 3.8 million).
  • Howitt and Michael are at least converging in agreement on the strengths, weaknesses and differences between their methods, even if their results vary.
  • I had lunch with Howitt and Michael last week. Both of them were visibly tired of the controversy over these numbers. Although everyone likes some attention to their work, they've been getting perhaps too much -- especially of the negative variety.
  • At this lunch, I proposed that I write a short note comparing and contrasting their methods and results (similar to my attempt here). They think that's a good idea (getting everything in one place, side-by-side) -- as long as they don't have to devote too much MORE time to my effort. I will pay attention to the response to this post before I proceed.
Bottom Lines: (1) Politicians will use whatever number suits them, even when it's discredited. (2) It's hard to estimate macroeconomic impacts from either simulations or aggregated data. Academics need to take (1) into account when doing (2), otherwise, their work is liable to be misused in ways that do NOT serve the public good.
* The study used January 2009 data on groundwater pumping and ZERO surface water exports from the Delta.

** Howitt says he has no idea how the 60,000-80,000 figure was inflated to 90,000 lost jobs, but it's sadly obvious why politicians, pundits and partisans would keep using a number he discarded five months ago.

*** Exports were reduced in compliance with the "biological opinion" that they were harming the Delta Smelt, an endangered species.

**** In the earlier revision, Howitt et al. considered updated data on groundwater pumping and the updated estimates of surface water deliveries (from zero to 10-15 percent of contracted allocations) and estimated 35,000 job losses and $980 million in lost revenue; the recent update used better data to estimate 21,000 job losses; economic losses are the same.

10 comments:

Anonymous said...

I have followed this debate carefully and I have a few comments, which I'll keep brief.

1. I Absolutely agree that politicians, interest groups, and "news people" need to have some accountability when reporting skewed numbers.

2. It appears there were no data problems in the analysis by Howitt et al. The problem, as they describe it, was using the REMI model to map revenue losses into employment losses. As I read it, the REMI model they initially used placed all the ag industry into the ag services sector which has much higher employment per $ revenue. As such, employment losses were overstated.

3. They used two simulations. SWAP to generate water impacts and revenue losses and IMPLAN to generate job losses.

4. I read that differences in impacts within counties are an argument for using regional models instead of aggregate statistics.

5. I don't see any assumptions in the Howitt et al piece that are being used to justify higher losses. In fact, they use strictly data whereas Micheal uses assumptions (ag labor shortage??) to obtain "desired results."

6. Hannity is outrageous, thank god for Jon Stewart.

Anonymous said...

I think such a post will likely get wide circulation.

I think it would be helpful if you also provide at least a sliver of more detail about how the drought is causing less pumping. Just like partisan talking heads, a lot of people make assumptions of what drought means. For example, the salinity standards attached to the projects' water rights.

btw: you should also, hit the tv circuit. Tape a brief explanation as a video response to the Hannity video on YouTube. Call him out. :-)

KE said...

A friend of mine commented that we are living in a post-fact world. I think that is sadly true.

David Zetland said...

@Anon -- look at p6 of this ppt presentation - it shows drought reduced pumping by 1.6maf and smelt by 500taf.

Andrew said...
This comment has been removed by the author.
Andrew said...

To the first anonymous poster above:

Point 4 is incorrect.
Michael doesn't make an assumption about an ag labor shortage - the ag sector has been complaining about not having enough workers for years. He is simply supposing a shortage because that's what farmers are telling us. He then uses this assumption to draw conclusions about farm employment - if you read his report, he actually concludes there is evidence of an ag labor surplus, which is in contrast to what farmers have been complaining about (a shortage).

Josh said...

Great post. I see a problem in getting true prices for water, down the road. Ag. has such a stranglehold on the Ca. legislature that they usually get exempted from water conservation measures, and I believe they will get exempted from paying for water. For example, what does the highly touted 20 by 2020 say about ag. use? Diddley, that's what.

Anonymous said...

@ Andrew.

I have read that report in plenty of detail. He is assuming a labor shortage, period. I quote directly from his report:

"However, this estimate of 11,700 jobs makes no adjustment to account for labor market shortages common before the drought, so they should be viewed as an upper bound on potential employment effects. As discussed earlier, a large labor shortage could mean that a drought that reduces the amount of land under cultivation could theoretically have zero impact on total farm employment. Thus, the total impact of current water shortages on employment ranges between 0 and 11,700 jobs."

To clarify, he notes that relevant research indicates that there is a surplus but he throws this aside and assumes a shortage. He uses this ASSUMPTION to impose a lower bound of ZERO on his job loss estimate of about 12,000 and CHOOSES a midpoint of 6,000. That is not empirically justified and is definitely a big assumption. There are no such occurrences in the Howitt et al piece, which was my point.

Jeff Michael said...

@Anon. Howitt is not using data, he is using simulations from models (with huge assumptions).
The data for farm jobs in the SJ Valley this year (and last year) show increased farm employment. That seems impossible due to water, but makes sense from the perspective of the regional labor market. From 1999 to 2006 farm jobs crashed and farm output did not. Labor shortages provide a theoretical explanation that is consistent with the last decade of farm employment data and what farmers have been saying.
We use 2 approaches. 1. The demand driven, input-output simulation gives 12,000 lost jobs with realistic multipliers. 2. Looking at the current data gives 0 lost jobs (the purpose of the shortage discussion is to give a theoretical explanation). The result is a range of 0 to 12,000 lost jobs, and we place the point value estimate in the middle of the range. You can try to make that sound arbitrary, but I would argue averaging 2 approaches is better than just using the approach that best supports your position.
If you can explain the last decade of farm jobs data consistent with the assumptions of input-output modeling, I would like to hear it.

Jeff Michael said...

@anon What is Howitt’s huge assumption? Mainly that input-output model is right (even before he tweaks it to spit out a massive 30 jobs per $1 million).
Input-output models assume perfectly elastic labor supply, and production functions that do not allow input substitution. Those are tenuous assumptions in any market, but particularly bad for agriculture. Academic studies show agriculture labor supply is very inelastic, and that agriculture labor demand is very elastic (implying lots of substitution possibilities). The implication is that input-output models will overestimate employment changes from marginal changes in farm output.
One final point about Howitt’s geographical claims. He is modeling land not people, and landowners are absorbing the largest economic losses. He needs to map the owners of the land before saying anything about spatial impacts on human welfare. Some of the largest landowners in the area live in Beverly Hills, Pasadena, Silicon Valley, etc. Even the “family farms” of Westlands have many out-of-town owners and use contract labor from other areas.