28 July 2010

Michael Oppenheimer Responds

[Michael Oppenheimer of Princeton University has graciously offered this post in response to several comments made earlier this week on his recent PNAS paper on climate change and Mexico-US immigration. His contribution is much appreciated. -RP]

Our article in Proceedings of the National Academy of Sciences, “Linkages among climate change, crop yields and Mexico–US cross-border migration" by Shuaizhang Feng, Alan B. Krueger, and Michael Oppenheimer“, has drawn criticism in this blog. Before we respond in detail, let us point out that the problem of climate-driven migration has resisted quantitative analysis for decades. To contribute to filling this gap, we proposed a line of modeling drawing on empirical evidence about responses to past climate variations, which may be informative about responses to future climate trends. If others have better ideas for estimating how climate change affects migration, they should publish them.

Richard Tol argues that our paper “confuses decadal weather variability with climate change”. We were not confused and I doubt anyone who read the entire article would be confused either. Using the response to historical short term variability to estimate the response to a long term trend is a common first approximation in climate impact analysis when quantitative process-based information is lacking, such as for the processes underlying the immigration response. We state in the paper,
“Such a method implicitly assumes that the response to changes in the climate variables is linear and symmetric, and that migration responses to 5-y changes in climate conditions can be applied to longer run trends. We acknowledge that, in actuality, the response to a trend may differ from the response to periodic variability...”
... particularly because climatic variability during the study period has not been of a magnitude or duration comparable to what is projected in the future under various climate scenarios.

Richard further argues that the study
“fails to control for other determinants of migration that may well be correlated with weather during the sample.”
First, in terms of estimation, our approach (fixed effects IV) controls all time-unvarying state effects through state dummies, and addresses (controls for) other determinants of migration that are NOT correlated with climate through using climate as the instrument. If some of those determinants are correlated with changes in climate (other than fortuitously), then the estimated semi-elasticity would be affected, but we would be just attributing some of the impacts of climate on emigration through other channels to the crop yield channel (as the two cannot be separated anyway given the data).

Second, in term of prediction, we emphasize from the beginning that our study is a “ceteris paribus” one. In the language of economics, it is a “partial equilibrium”, not a “general equilibrium” study. The objective is to examine the “marginal” effect of climate change on migration, holding everything else constant, not to forecast the “total” number of emigrants in the future.

Finally, Tol states,
“they extrapolate beyond belief... their largest yield change is -48% between now and 2080. If technological progress would bring about a 1% yield increase per year, then the two effects cancel each other out. 1% may be too low, -48% is probably too high.”
We are careful to acknowledge the impact of other factors including technological progress in the paper:
“estimate of the elasticity of migration is conditional on many factors specific to Mexico for the period under study, such as the macroeconomic situation compared with that of the United States, the population share of youths (who are more likely to migrate), farming practices, the relative importance of the agricultural sector, and agricultural policies including responses to droughts and other climatic events that adversely affect crop yields.”
However, note that we are estimating the “marginal” effect of climate change. If technological change produces general increases in productivity and partially offsets effects of climate change, then predictions of “total” migrants would change, but the “marginal” effect of climate change would not.

With respect to the observation of Harrywr2 that at the upper end, our projections exceed the current size of the agricultural labor force: First, people not employed in the agricultural sector would also be affected by declines in crop yield, due to a well known and substantial “ripple effect”, and these would also be represented in our empirically determined semi-elasticity. Furthermore, the projection of between 1.4 to 6.7 million Mexicans migrating to the U.S. due to climate change applies to a 70-year time period. Thus, more than one generation would be affected.

Second, the projection is based on the assumption that Mexican adult population would stay at today’s level of 70 million. The projected numbers are significant but not necessarily “alarming” if they are put into the right context. The recent annual Mexican emigration flow to US is about 500,000. If in the future this is cut by half on average due to factors such as declines in fertility, to 250,000 emigrants, the total number emigrating due to all other causes would be 17.5 million in the 70 years.

...Michael Oppenheimer 28 July 2010

63 comments:

Roger Pielke, Jr. said...

Michael-

Thanks again for the reply. Let me ask a few questions.

First, your last sentence is rather remarkable. If a baseline expectation for future immigration is 35 million (500k * 70 yrs), then taking your analysis at face value suggests that climate is 5% to 20% of this value. If fertility alone can influence these numbers by 50%, and other factors are similarly as large (such as economics, demographics immigration policies, etc.), then it would seem that the "right context" for interpreting your results is that climate change effects on immigration are in the noise. Is this a fair conclusion?

Second, CBS news paraphrases your views as follows:

"Oppenheimer himself freely acknowledges:
the fudgy nature of predicting climate change’s effects, and that while the numbers make for a sexy headline, you shouldn’t take them too seriously."
http://www.cbsnews.com/stories/2010/07/28/tech/main6721031.shtml

Is this a fair characterization of your views?

CBS News also quotes you as saying:

"Our goal was not to project specific outcomes 80 years from now but to show the magnitude of problems that policymakers ought to pay more attention to."

Immigration gets an incredible amount of attention. What would it mean to "pay more attention to" the issue?

Any replies appreciated. Thanks!

Hector M. said...

Keeping other factors constant is a fair methodological procedure when the other factors CAN be kept constant while the variable of interest varies. But farmers behaviour CANNOT be kept constant if climate changes. It would be tantamount as keeping plant physiology constant when temperature and humidity vary. Farmer behavior is as essential to agriculture as plant behavior. This means that the very IPCC notion of "potential impacts" (impacts of climate change in the absence of adaptation) may be adequate for natural processes such as the growth of wild vegetation, but not for interactive human/nature processes such as agriculture (even in the absence of technical progress of the kind mentioned by Richard Tol, that in Latin American and Mexican agriculture has been proceeding at rates between 2% and 3% per year in recent decades).
Even marginal effects of climate change on migration (or other similar impacts) should take into account these interactions. Marginal effects with other factors constant might be adequate for evaluating the impact, say, in one year, but immediately the supposedly "constant" factors start responding, e.g. by farmers changing their choice of crops, their technology, or their own domicile and employment, such as migrating to cities (something they have been actively doing for decades).
Immigration to the US would undoubtedly continue, and even intensify in the coming decades. Some people leaving rural areas might cross the border instead of heading towards Mexican cities, as they usually do. But describing this process in terms of the "marginal" effect of climate change, operating along 80 years with all other factors constant, is not a defensible methodology. Marginal effects are only valid for, well, marginal changes of an infinitesimal nature and in the short term. Marginal changes in the long term are hardly definable.

Raven said...

Such analyses are no better than astrology. It is impossible disentangle all of different factors and attribute any portion to climate change. In fact, there is no credible evidence that can link changes in the local climate in Mexico to CO2. It could be regional variations or even land use driven.

We should face facts: if Dr. Oppenheimer had produced a paper using the same methodology that suggested climate change would reduce migration there is no chance that it would have been published or even discussed. They only reason this paper saw the light of day is because it supported the political agenda of the people pushing for CO2 mitigation policies.

The reputation of science will continue to decline in the public consciousness as long as the scientific establishment refuses deal honestly with its own biases when it comes to this topic.

Roddy said...

Presumably if agricultural yields in Mexico are changing (falling) to the extent that they cause economic-driven migration, there will be changes in yields in the USA too in relevant areas such as Florida and California, with an economic effect there.

What is the impact of the climate change posited on US agricultural production? Up/down? Does it vary by region (higher yields in, dare I say it, Minnesota)?

In general, what is the impact of decadal climate change on agricultural yields compared to other impacts (as in mosquitos)?

Wikipedia tells me that agriculture represents 3.9% of Mexican GDP - a rather small number - although it employs 18% of a 38m workforce, or 7m people.

So a decline in agricultural production would have a rather limited 'well-known and substantial ripple effect' surely? The 'multiplier' effect of agriculture (using terms very casually) would need to be very significant?

And so on.

I am no clearer following Michael Oppenheimer's response, for which thank you, on what appear such major prima facie limitations on the paper's ability to provide useful food for thought for policymakers.

Malcolm said...

A study like this could have concentrated on modelling the movement of Americans from the southern states to the northern states in the event of significant climate change over a similar period. In doing so many socio-economic-political factors would have been minimised in the modelling process leading to a more meaningful and robust result.

But to link Global Warming to the politically charged issue of Mexican emigration is delibrately playing the racist card.

The 30 million plus Mexican-Americans must feel sick to their stomachs that scientists in acting as advocates have delibrated fostered resentment to their presence.

This is a new low for warmists.

jgdes said...

"Using the response to historical short term variability to estimate the response to a long term trend is a common first approximation in climate impact analysis "

That phrase neatly summarises what is wrong with the entire field and why the vast majority of impacts papers can be safely ignored. In reality much real data are available should people bother to look, so there is no need to make wild extrapolations based on a totally unrealistic assumption of linearity. However such data invariably show zero trends over the long term.

Theres the rub! To use long term data is to prove that vast swathes of researchers are just wasting everyones time and money solely in trying to justify their reason for existence. As turkeys don't vote for Xmas, the funding bodies really need to be more rigourous.

eric144 said...

Dr Oppenheimer

You have certainly highlighted the American horror of a Mexican invasion. However, I live in the south of Scotland and was wondering whether climate models predict a larger invasion of French or Germans into Scotland.

To Americans, Europeans probably all appear the same, but there are differences. Scotland has a traditional 'auld alliance' with France (actually a mutual hatred of the English if we are being honest) and a couple of million French working in our vineyards in 2080 shouldn't be a problem. The Germans are historically a different matter. If they lose 90% of their manufacturing to India and China, we probably couldn't cope.

thank you

Sean said...

This is nothing more than political lobbying wrapped up in science themed wrapping paper. It's gotten much more common in the last 5 years. It's the same junk as predicting that "climate models say that the midwest and central part of the country will see more temperature rise" a few weeks before a vote is scheduled for in the senate regarding cap and trade. Or a report on sea level rise from last fall (just before Copenhagen) that sea level rise will be highest in the mid-Atlantic states (including DC). It's amazing that these scientists turned lobbyists have not caught on to the fact that politicians and their constituents are not as stupid as they think they are. In fact, they know the game all to well, roll their eyes when the read these things, and wonder what science has come to.

Roger Pielke, Jr. said...

Michael Oppenheimer responds to my questions in -1- above by email:

""Roger, briefly:

1. Nothing that is statistically significant is "in the noise". Rather, it would be one of several factors influencing immigration.

2. No, this is not a fair characterization of my views.

3. We would like this aspect of migration and immigration to be considered among the several factors driving such changes, as policy makers consider their responses to climate change, and to immigration.""

Roger Pielke, Jr. said...

-9-Michael

Some brief rejoinders to your replies:

1. From a policy perspective this reply doesn't really work. For instance, the level of unemployment in Colorado is statistically significantly related to the state's higher education budget. Should Congress consider this factor when deciding whether to extend unemployment benefits? No, it is in the noise. Another way to say this is that unemployment decision making is insensitive to many relationships that are "statistically significant."

2. OK. Please do feel free to correct the statement.

3. You have simply restated my question to you. What does it mean to "consider" such changes in the context of climate or immigration policies? What should policy makers be doing differently than they are at present?

Thanks.

Frontiers of Faith and Science said...

Dr. Oppenheimer,
Thank you for taking the time to respond.
Compared to recent rates of Mexican immigration to the US your projections do not seem very significant.
Perhaps, since using decade length time periods means you are talking about weather patterns and not climate, it would be good to look at historical movements of people, like those that occurred in the Roman era for example and see how weather patterns drove those.
I am wondering this: does it take much work to tie nearly every study back to climate, or is the paradigm of climate simply the lens through which any academic study has to be projected in order to be credible or to secure funding?

jstults said...

Another way to say this is that unemployment decision making is insensitive to many relationships that are "statistically significant."
Or that there's a difference between practical significance and statistical significance.

Richard Tol said...

1. Weather and climate. In an unpredicted drought, farmers have no choice but to take the hit. If exceptionally dry weather becomes the norm, crop management, crop choice, and land use adapt to the new circumstances. Therefore, a regression of crop yields on weather variables provides information on the impact of climate change without adaptation -- that is, the worst case.

2. Omitted variables. There are two problems: efficiency and bias. Econometricians tend not to worry about the efficiency of their estimates (as significance levels are biased downwards), but with few observations (as in this paper) they should as there is little reason to assume that the parameter estimates are close to their "true" value. The issue I raised, however, is bias. If an omitted variable is correlated with an included variable, parameter estimates are biased. This paper uses panel data, with two periods and a number of states. Over time, all sorts of things have changed -- not just the weather -- including international trade, US migration policy, and so on. The estimation procedure cannot distinguish these changes from changes in the weather. Over space, it may well be that a range of development variables are correlated with climate as there is a strong North-South gradient in climate and development in Mexico. The authors argue that the panel (time x space) overcomes both problems, but that is an untested leap of faith.

3. Extrapolation. I have no problems with ceteris paribus in an academic paper. I would hesitate to get it into a newspaper, though. More importantly, even if we focus on crop yields only, then climate change is still not the dominant signal. I would think that crop yields in the less developed parts of Mexico are far below potential. So, if you worry about migration, you should go help farmers reach their potential.

Hector M. said...

Roddy said:
"if agricultural yields in Mexico are changing (falling)..."

Well, in fact they are not falling by rising. "Maize yields in Mexico have greatly increased since the 1960s to the 2000s, from a mean of 1027 kg/ha in 1961-64 to 2715 kg/ha in 2001-04 (FAO, Faostat, 1961 to 2004 archive series), increasing at about 2.46% per year or an accumulated 164% in 40 years, equivalent to 1036% in a century.
At half the rate observed in 1961-2004 (i.e. with yields increasing at 1.23% per year instead of 2.46%), yields in 2055, even reduced by 8% due to climate change, would be 4774 kg/Ha, reaching 8276 kg/Ha in 2100, still low compared with potential maize yields at the current state of technology (where irrigated and fertilised hybrids yield more than 15 MT/Ha), but quite higher than those of the 2000s. That would be undoubtedly not top-notch by the standards of the year 2100, but still much higher than current yields, in spite of intervening climate change (for a Mexican population by the year 2100 that is expected to be declining and not much higher than today in absolute numbers)." (H.Maletta and E.Maletta, "Climate change, agriculture and food security in Latin America and the Caribbean", http://ssrn.com/abstract=1619395, section 8.3.2).

Mexican small farmers are by now (as most small farmers elsewhere in Latin America) mostly obtaining their sustenance in the market, and not mainly from their farms. [...] total on-farm consumption represented only 1.7% of the income of all "rural" households (i.e. in this case those located in places with population below 2500), according to INEGI 2007 [the National Household Survey]. [...] The number of vulnerable ("low status") smallholders keeps decreasing in Mexico (it was estimated at 1.69 million in 2005, decreasing to 1.59 million in 2007, which is a 3% decrease per year), along with a slowly decreasing rural population and a rapidly expanding capitalist agriculture. This process is unrelated to climate change: it is chiefly due to a long-term process of peasant proletarianisation and penetration of market relations in the countryside, accelerated lately by NAFTA. The process is hardly novel: it has happened before in all capitalist countries since the Industrial Revolution and even before, and keeps happening everywhere.[...] Smallholders are located mainly in Southern and Central Mexico, where water resources are relatively abundant, whilst capitalist irrigated agriculture prevails in the arid and semi-arid North, the region more directly threatened by increasing water scarcity" (idem, section 10.4.1). The share of rural population in Mexico (mostly composed of other groups, not subsistence farmers) is expected to fall from current 22% to about 12% in 2050 y about 6% in 2100. Growing at less than half the rate observed in 1950/2010, GDP per capita in Mexico would reach a level higher than today's Germany or France by 2100. Even if income distribution, contrary to world experience, worsens along that period of economic growth (with the lowest decile receiving 0.5% of total income, instead of the current 1.2%), poverty (relative to a $2.50 PPP dollars a day per capita) would be quite small (below 5%). Undernourishment (measured by FAO) is now at about 4%, deemed not significant, and would be almost nil by 2050/2100. Migration there would be, but not triggered by mass hunger caused by climate change.

Roddy said...

Hector, I meant if they are falling in the future world of global warming that is the presumption of Michael Oppenheimer, and falling enough to trigger emigration, then ...... what is happening the other side of the border? Presumably, ceteris paribus, falling there too, until we get up North to Minnesota.

My presumption, like yours and Ricahrd Tol's, would be that yields have a long way to rise on a pan-Mexican basis.

I liked the comment about Scotland being invaded by hordes of French and German migrants.

Stan said...

When you don't have enough information to say anything meaningful, don't say anything. This paper is garbage because, in the end, it doesn't mean anything.

Oppenheimer says that policymakers should take his paper into account when making policy about climate and immigration. How? His paper says if it gets hot Mexicans might come across the border over the next 70 years. Wow. I'm impressed. So that means policy should change in what way?

Geez, this is pathetic. If this represents what scientists do, we are clearly not getting any value for our investment in "science".

models-methods-software.com said...

What is the actual information content in any of these what-if studies? Firstly, they are connected to, in very fundamental ways, the IPCC what-if scenarios; frequently the most pessimistic of those are chosen. Secondly, the evolution trajectory of all of humankind's interactions with the environment cannot be known for even 50 years. As noted in the comments above, the evolution in agriculture practices alone has been tremendous over a few decades. This evolution was not ever predicted, it happened because someone was committed to the concept. Thirdly, I do not see that the uncertainties associated with each critical aspect have been propagated through the analyses. And finally, causality seems to be hanging on by a very thin thread.

I do know with absolute certainty that ceteris paribus will not obtain. Not even over one decade.

On a side note, are we back to the olde what-if vs. estimate vs. projection vs. prediction vs. EWAG issue? Most of these have been used in the discussions. I will say that the paper does not address a prediction. I will also say that the paper addresses at most an EWAG, and very likely a plain WAG.

Tomás Enrique said...

So... the climate change affects only developing people?... what's about the crop fields on USA, what about the unemployment of american people?
This models assumptions and limitations make reality from imagination, Let's see:
1. Farming in Mexico its mainly seasonal on the south, depending of the rain season, and irrigation driven on the north, there are some differences on both.
2. Inmigration of Mexicans to the US it's not driven by falling on farming, it's driven by poverty, lack of job opportunities and poor education, the real problem it's not the absence of crop fields or farming, it's the absence of new and additional sources of jobs besides farming.
3. The regional view (mexican farming) fits on global view (climate change)... and that's an lie... it's supossed that climate change it's global, so, wherever mexicans go, there will be no crop fields, globally speaking.
4. Mexican border inmigration grows because ALL mexican economy (not just farming) falls, why not include on the modeling the drug trafficking, the negative impact on tourism, the effect of the AH1N1, the impact of corruption on the mexican system?

Numbers on models tells the wrong story.

Richard Tol said...

-17- @model-methods-software
Ceteris paribus is the model/scenario equivalent of a controlled experiment: Change one variable at a time, and see what happens.

This is the preferred academic strategy as it maximizes insight into causality.

It is the wrong way to go about giving policy advice or predicting the future.

hardycross said...

Would we expect US citizens to move to Canada and the Canadians to move up to the Arctic by 2080? The Mexicans would have to move up to the US since the Columbians would be moving up to Mexico.

Josh said...

This story is just the kind of daft alarmism that confuses rather than informs and sadly Michael Oppenhheimers defense of it looks too fussy to be helpful.

As usual Raven hits the nail on the head and Eric's comment at 7 made me laugh out loud. I come from Scotland too so the idea of climate change bringing hordes over the border is very comic - just think if they were all English, Eric, that would even make the front page of the Stornoway Gazette.

Josh said...

Hardycross- hilarious. I guess the knock on effect will be the penguins invading Chile. This has to be cartooned.

Hector M. said...

Roddy,
an assessment of impact on agriculture cannot rely on an evaluation of what the yields would be now (with today's crops and technology) is climate suddently changes, but what would happen in 50-100 years if climate gradually changes along with existing farming technology. All points to greatly increased output in 2100, total and per capita, even taking account of the impacts of climate change.

There are several adequate methods for this kind of assessment. The best by far is Integrated Assessment as in the works directed by Günther Fischer at the Internac. Inst. for Adv. Systems Analysis (IIASA, working closely with FAO). See several Fischer et al papers in recent years. Too complicated to explain here, please see my work referenced in previous comments of mine (http://ssrn.com/abstract=1619395).

Harrywr2 said...

Dr Oppenheimer,

12% of the Saudi Workforce is employed in agriculture which constitutes a little over 3% of Saudi GDP. About 30% of the Saudi population is made up of 'guest workers', effectively creating an inward migration.


http://www.ameinfo.com/100371.html

13% of the Mexican workforce is employed in agriculture representing a little less then 4% of GDP. Mexico has an outward migration.


It doesn't appear that climate is the primary driver of migration or agricultural economic opportunity.

The percentage of Mexican GDP attributable to agriculture has been trending downward for 40 years. From 14% to currently just below 4%.
http://www.google.com/publicdata?ds=wb-wdi&met=nv_agr_totl_zs&idim=country:MEX&dl=en&hl=en&q=Mexico+Agriculture+GDP

vincewerber said...

I look at the immigration situation a different way...
I am told that upwards of 30% of those leaving Mexico are, in fact, 'Natives'.. as in Native AMericans from the south lands... OK that's my bias in this but...
What consideration has been given to social and political matters? One assumption here is that all of these persons are 'farmers'... remember ass-out-of-u-and-me...

I have several Pesos in my wallet.. yes I do... right now... from Mexico and several other Countries to our south and I can tell you that single one dollar bill American IS worth so much more on the rate of exchange than all of my Pesos...

Now tell me... would you work for pennies (Pesos) or a few dollars...?

Follow the money...

Wesa (cat in Cherokee)

baltwo said...

See Willis Eschenbach's take

Sharon F. said...

Hector M.-- I can't tell you how refreshing it is to have someone else on the blog who knows about plants.

You said "This means that the very IPCC notion of "potential impacts" (impacts of climate change in the absence of adaptation) may be adequate for natural processes such as the growth of wild vegetation,..."

You are right that it doesn't work for agriculture, but it also doesn't work for wild plants. Here is what we don't know:
1) how much of a change in temperature will kill them (pure physiology)
2) How much genetic variation there is in their offspring enabling those offspring to adapt
3) The impacts of climate change on diseases, insects, herbivores, and soil microflora.. all of which impact plants.. and the totally unpredictable interactions of those factors in the future influenced by unknown levels of climate change, unknown microclimate at the level the plant perceives it, and a variety of other unknowns.

It seems to me there are probably more useful things for people to do than attempt to predict combinations of factors when the factors themselves are unknown.

Sorry, Roger, this didn't have to do with Oppenheimer's article, more of a pet peeve.

Roger Pielke, Jr. said...

Thanks to those of you who have emailed. The Blogger recent comments widget is indeed broken, and I assume will be fixed by Google in due course.

Roger Pielke, Jr. said...

By email from Shakeb Afsah:

"Policy Tabloidism built on Econometric Models

Prof. Oppenheimer’s original paper is getting a public peer review that highlights the weakness of the formal review process which can be sometimes soft, and consequently clear a paper for publication on the basis of its policy appeal. Such papers often rely on econometric models that create an illusion of accuracy and entice researchers into making bold but questionable predictions and forecasts. Prof. Oppenheimer’s estimate of climate change driven emigration from Mexico ranges from 1.4-6.7 million is a good example. He offers various caveats like the assumption of linearity for the impact of climate variables (simply cannot be justified), 5-yr change mirrors the long term trend (this is a huge leap of faith) and infrastructure of poor farmers can’t afford investment (ignores the possibility of public investments). The instrumental variable methodology itself has many restrictive conditions and additionally, their estimate of semi-elasticity then feeds into the estimates of food supply impact of climate change derived from a SEDAC paper that itself is loaded with data caveats, and is not a properly peer reviewed paper (http://sedac.ciesin.columbia.edu/giss_crop_study/CCMdatalimits.html ). So the estimates of quantification by Prof. Oppenheimer is essentially based on a model from gray literature. Now if Prof. Oppenheimer’s paper was meant for a classroom discussion as a teaching tool, it would have been ok, however it is now placed in the public domain with a serious policy message. And this is where Prof. Oppenheimer has made a big mistake—he has peddled a rather half-baked piece work as policy intelligence.

I will would also like to bring attention to another economics paper that evaluates the correlation between temperature and underweight new born babies—which made its way to the prestigious American Economic Review. And here again on the basis of statistical correlation the authors show that high temperature leads to underweight babies, but offer no medical/biological explanation for such a link. This kind of paper made its way to Joe Romm’s blog who further sensationalized it by adding an image which looked more like a pro-life campaign than a blog on climate science (see the links below) .

The point here is simple—economists have to do some serious soul searching about how they extrapolate the results of their regression analysis into policy messages. If the current trend is to create some internet chatter, they are doing a great disservice to their profession; and if they are truly serious they have to work harder to get better data, be upfront about caveats, and given the data and methodological caveats they should ask if it is even worthwhile to do a research. If they don’t learn to take a hard look at their research, they risk getting completely discredited in the new world of public peer review.

http://www.econ.ucsb.edu/~olivier/DGG_AER_PP_2009.pdf

http://climateprogress.org/2009/06/10/energy-and-global-warming-news-for-june-10th-hotter-planet-means-more-underweight-babies-china-aims-for-20-renewables-by-2020/"

Roger Pielke, Jr. said...

By email from Shakeb Afsah, part I:

"Policy Tabloidism built on Econometric Models



Prof. Oppenheimer’s original paper is getting a public peer review that highlights the weakness of the formal review process which can be sometimes soft, and consequently clear a paper for publication on the basis of its policy appeal. Such papers often rely on econometric models that create an illusion of accuracy and entice researchers into making bold but questionable predictions and forecasts. Prof. Oppenheimer’s estimate of climate change driven emigration from Mexico ranges from 1.4-6.7 million is a good example. He offers various caveats like the assumption of linearity for the impact of climate variables (simply cannot be justified), 5-yr change mirrors the long term trend (this is a huge leap of faith) and infrastructure of poor farmers can’t afford investment (ignores the possibility of public investments). The instrumental variable methodology itself has many restrictive conditions and additionally, their estimate of semi-elasticity then feeds into the estimates of food supply impact of climate change derived from a SEDAC paper that itself is loaded with data caveats, and is not a properly peer reviewed paper (http://sedac.ciesin.columbia.edu/giss_crop_study/CCMdatalimits.html ). So the estimates of quantification by Prof. Oppenheimer is essentially based on a model from gray literature. Now if Prof. Oppenheimer’s paper was meant for a classroom discussion as a teaching tool, it would have been ok, however it is now placed in the public domain with a serious policy message. And this is where Prof. Oppenheimer has made a big mistake—he has peddled a rather half-baked piece work as policy intelligence. . ."

Roger Pielke, Jr. said...

Shakeb Afash, part 2:

"
I will would also like to bring attention to another economics paper that evaluates the correlation between temperature and underweight new born babies—which made its way to the prestigious American Economic Review. And here again on the basis of statistical correlation the authors show that high temperature leads to underweight babies, but offer no medical/biological explanation for such a link. This kind of paper made its way to Joe Romm’s blog who further sensationalized it by adding an image which looked more like a pro-life campaign than a blog on climate science (see the links below) .



The point here is simple—economists have to do some serious soul searching about how they extrapolate the results of their regression analysis into policy messages. If the current trend is to create some internet chatter, they are doing a great disservice to their profession; and if they are truly serious they have to work harder to get better data, be upfront about caveats, and given the data and methodological caveats they should ask if it is even worthwhile to do a research. If they don’t learn to take a hard look at their research, they risk getting completely discredited in the new world of public peer review.



http://www.econ.ucsb.edu/~olivier/DGG_AER_PP_2009.pdf



http://climateprogress.org/2009/06/10/energy-and-global-warming-news-for-june-10th-hotter-planet-means-more-underweight-babies-china-aims-for-20-renewables-by-2020/"

Roddy said...

I never saw that global warming means underweight babies, and really, really, underweight too: 'Due to the effects of hot temperatures, mean birth weights will decrease, on average, by 0.22 percent among whites and 0.36 percent among blacks.'

Roger, Shakeb Afsah gives one reason for 'soft' peer-review, being 'clear a paper for publication on the basis of its policy appeal' - I'm not in academia, but how much is the simple sausage machine of being published, being cited, raise your score, and your institution's score, everyone connives to game the system?

I had dinner with the Warden of my Oxford College the other day, he's American, and he explained the extreme drawbacks of the system, and the deleterious effects on quality of work (negative) and quantity (too much).

jgdes said...

Sharon
Actually your point does have to do with what should be considered in these impacts papers so it is not off topic. I'd point out too that it has always been well established that the medieval warm period in Europe and the Roman warm period before it coincided with better crop yields, more available money to spend on fancy buildings, and even taller people. Conversely cold period have always been historically bad for everyone. Hence despite probable increases in insects, diseases and invasive species, it has long been taken as read, and indeed even blindingly obvious, by most scientists in most fields, that warmer is usually always better for plants and animals. And if that wasn't enough then the fact that the planet has been greening up despite all the deforestation surely is enough. All of which contextual analysis is missing in most scary impacts papers of this genre.

Does this blatantly shoddy work not call into question everything else ever done by this scientist?

jgdes said...

In other news, the 2009 Bordeaux vintage is confirmed as the best ever. Respected critic Robert M Parker declares it as "the finest vintage I have tasted in 32 years". Thereby edging out the "perfect year" of 2005.

Hans Erren said...

You can't have it both ways:

Either poor countries stay poor. Then they don't burn fossil fuels, the climate doesn't warm, and there can't be climate refugees.

Or the poor countries will get very rich, but then there won't be poor people to migrate to the US.

Matt said...

I may have missed it, but I don't see where Oppenhiemer discussed the huge impact of NAFTA on Mexico's ability to compete with significantly cheaper and heavily subsidized US grown ag products....

Josh said...

Hi
cartoon for Michael.

http://www.cartoonsbyjosh.com/Climate_refugees_scr.jpg

eric144 said...

Another brilliant cartoon Josh.

" just think if they were all English, Eric, that would even make the front page of the Stornoway Gazette"

I didn't want to scare anyone with the ultimate nightmare scenario.

I was forced (through work) to live in the English ghetto that surrounds Glasgow University for five years. Despite being given free housing and education for their children, they had a chip on their shoulder and refused to learn the language. They spoke a very base and corrupt version of English known as RP. Also, they weren't the sophisticated metropolitans you would expect. In the main, they came from provincial backwaters like Oxford and Cambridge.

Rebuilding Hadrian's wall would only be a first step.

Cheryl said...

I agree with Hector but I will take it further.

Oppenhiemer ignored the impact of WTO and NAFTA free trade agreements that caused a reduction of farming households. According to a study by Jose Romero and Alicia Puyana carried out for the federal government of Mexico, between 1992 and 2002, the number of agricultural households fell an astounding 75% - from 2.3 million to 575, 000.

85% of Mexico's farmers are small and marginal and grow largely grains and oilseeds. They normally use farmer saved seeds. However that seed stock has now become contaminated by banned GMO corn genetics.

GMO crops normally require more fertilizer, water and care than native corn. Oppenhiemer did not determine if the crop failure was due to the introduction of nonnative seed genetics, the switch from the tough white corn to the more delicate yellow corn of interest to big AG or competition from foreign grain that bankrupted the farmer and left his fields untended.

As far as I am concern Oppenhiemer is busy hiding the human devastation caused by NAFTA and the WTO Agreement on agriculture. Both agreements are designed to "open up" new areas for international Ag giants by flooding the market with cheap US grain. This bankrupted the small farmers so the land can then be consolidated into monoculture farms owned by the Agricultural giants such as Smithfield, home of the swine flu epidemic.

Once most of the native farmers were driven off the farms (2002) US Biofuel legislation - again driven by big Ag - was implemented in the USA wiping out surplus reserves and driving up the prices on grain worldwide. (My cost of grain doubled and has since tripled.) This culminated in the 2008 food riots.

Good Grief, Even ex-president Clinton understood it was world politics (Woldbank/IMF SAPs, NAFTA, WTO) that caused the food riots and starvation and not the climate!

Clinton: "“Today’s global food crisis shows we all blew it, including me when I was president, by treating food crops as commodities instead of as a vital right of the world’s poor, Bill Clinton has told a UN gathering."

Harrywr2 said...

"And here again on the basis of statistical correlation the authors show that high temperature leads to underweight babies, but offer no medical/biological explanation for such a link."

Having 3 daughters I don't need a study to explain to me the causation of hotter weather to low birth weight.

It's vanity. A few extra pounds hides easily in sweater weather. A few extra pounds doesn't hide at all in swimsuit weather.

The warmer the weather, the more self conscious some portion(at least my 3 daughters) of the US female population will become about their weight.

Stephen Pruett said...

If this was an epidemiological study, there would be much concern about confounders and whether all the possibly relevant ones has been identified and controlled for or eliminated as a cause of previous changes of migration associated with changes in climate. I am unfamiliar with this line of work, but in my field, establishing cause-effect relationships is hard, and only rarely done decisively without experimental interventions. The case would be bolstered if there were quantitative data demonstrating the "mechanism of action". In other words, why did climate change cause people to move? I would think this could be done using anonymous surveys of recent migrants and actually asking them why they moved and determining what portion of them moved because of anything that could be related to climate. I think that type of evidence would be preferable.

Richard Tol said...

@Harry2wr
The study is about birth weight.

The problem is that it does not properly control for other factors that may affect birth weight. Particularly, in the US, there are pockets of deep poverty in places that are hot and humid.

willis said...

In my article on this study, I have shown that Oppenheimer et al. did not calculate the crop yields correctly, which renders the study meaningless.

For unknown reasons, Oppenheimer et al. calculated their so-called "yield" figure as production divided by area planted, rather than using the correct yield figure.

Perhaps Dr. Oppenheimer might comment on that error, and let us know whether he plans to issue a correction or a retraction to the paper because of the error.

w.

NewYork said...

Pielke and Tol's argument are a series of illogical red herrings.

The study's abstract alone clearly indicates that they are considering the emigration effect of the decline in agricultural productivity due to climate change independent of other factors. I find this useful. Pielke doesn't. His basic philosophy is that since there are other factors involved, it's impossible to make a prediction, and therefore the study is "silly". Using this logic, I cannot possibly determine the impact that taking a long vacation in Hawaii will have on my bank account, since there are other things affecting my bank account, and the vacation may have unforseen costs or mitigating factors. Nothing to worry about. I might as well vacation all year. I find that notion silly. Pielke (and others) consistently use such logic to downplay global warming concerns.

Craig 1st said...

willis --43

Thank you for getting to the nub of the issue and keeping it simple. Not bad for a cowboy scientist.

Christopher said...

"GMO crops normally require more fertilizer, water and care than native corn."

Why on Earth would GMO crops uniformly require more resources to grow? Are you comparing GMO crops derived from local variants, or imported variants? If the latter, it's not because they're genetically modified.

jstults said...

NewYork:His basic philosophy is that since there are other factors involved, it's impossible to make a prediction, and therefore the study is "silly".

I think you are confused. Pretty clearly, the point of their argument is that, when you're making policy, the relative magnitude of the effect matters (practical significance), as opposed to focusing on whether you can measure it or not (statistical significance). Tol indicated that he thought this sort of study was a useful academic exercise, but wasn't very useful for informing policy decisions. It was spun in the press as having policy implications, and the author seems to be making that claim as well.

Mike said...

One simple question to ask is how Michael Oppenheimer and his colleagues know what the effect of climate change will be on Mexico?

Current climate models are known to be poor at making projections for the change in regional climates, and different models can project different changes in climate for the same region. So which projection(s) did they use and why?

Harrywr2 said...

Richard Tol said... 42

@Harry2wr
"The study is about birth weight."

From the CDC
http://www.cdc.gov/nchs/data/nvsr/nvsr57/nvsr57_07.pdf
"These data show that in 2006, 13 percent of all mothers gained less than 16 pounds, which is considered inadequate for most women....For mothers of at least term (37 or more weeks gestation), singleton births, the percentage who gained less than 16 pounds increased nearly 50 percent....over this 16-year period."

If the mother doesn't put on enough weight then the baby is going to be 'under weight'.

US Poverty levels decreased between 1990 and 2006.

Lacking grant money and an obvious economic reason why mothers aren't gaining enough weight during pregnancy I'm going to hypothesize 'vanity' as the cause and blame it all on Twiggy rather then Big Oil.
Of course in order to get grant money I'll probably need to get a PHd and hypothesize 'climate change'. :)

I agree your broader point, a large number of factors needs to be controlled for to determine a 'causality' for low birth rate.

Stephen Pruett said...

Poor analogy, New York. You know how much effect spending $2000 will have on your checking account. You can get unequivocal direct data on it. No one knows the degree to which climate change contributes to migration. It is not only likely that there are other causes that are quantitatively more important, there is no direct, "mechanistic" cause-effect evidence in the whole chain. Climate change is correlated with crop production and crop production is correlated with migration. The problem is that no one has actually gone to the migrants and asked them why they moved. I think that is the only direct, definitive data in this situation. Doing it would generate its own set of problems, but direct data trumps correlation, and a quick search indicates that many surveys of latinos in the U.S. have been published.

Richard Tol said...

-44 NewYork
I do not dispute that climate change would affect crop yields, nor that crop yields would affect migration.

My problems with the paper are (1) that it overestimates the impact of climate change on migration and (2) that it presents a controlled experiment as a prediction.

Would you care to be specific in what sense my critique is illogical?

Harrywr2 said...

jstults said... 47

"I think you are confused. Pretty clearly, the point of their argument is that, when you're making policy, the relative magnitude of the effect matters (practical significance)"

The study extrapolated a emigration effect caused by a drop in productivity due to corn.


USDA report on NAFTA
http://www.fas.usda.gov/itp/Mexico/030607StrengtheningNAFTA.pdf

If one looks at what is happening in Mexico under NAFTA, mexican corn production is already under extreme price pressure from US corn exports, however Mexican fruit and vegetable exports to the US are doing nicely.

One would expect a 'crop substitution' to occur, I.E. more labor and acreage put into fruit and vegetable production and less into corn production.

A study of what will happen in 2070 if nothing but the climate changes in the next 60 years is not particularly useful to anyone.

We can say with a high degree of certainty that the Mexican crop mix will be different, the level of Mexican agricultural employment will be different and the differential between US and Mexican wages will be different.

One can not determine how 'climate change' will effect a crop mix without knowing the crop mix even if we had the ability to predict regional climate 60 years in advance.

Sharon F. said...

Cheryl and Christopher:
Cheryl, you said:
"However that seed stock has now become contaminated by banned GMO corn genetics.

GMO crops normally require more fertilizer, water and care than native corn."

Christopher said that all GMOs can't possible require more resources.

The answers to your points need to be couched in terms of what are the crops Mexicans are producing. I don't know how much corn comes from Mexico; I do know that NAFTA had Mexico outcompeting Florida for fresh fruit and vegetables. That is what Harry2wr is referring to. Few of these are GM.

GMOs can be designed to use fewer or more resources or to do better under drought or other climate-changed conditions in the future. GMO's in the future can be anything. GMO corn varieties in the present have certain attributes.

What Mexican farmers will be growing in the future is anyone's guess.

Sharon F. said...

I am detecting a general trend here..when two economists and an astrophysicist write about agriculture there could be a problem. I think if this article had been reviewed by someone with a background in agricultural sciences, these problems may have been caught.

It seems as if we are seeing a bit of disciplinary territorial aggrandizement into the terrain of the applied biological sciences by physicists and economists.

Perhaps we need a moratorium on calling something "science" when the most knowledgeable disciplines are not involved in a journal article on a given topic (agriculture, epidemiology, and medicine in these examples).

Harrywr2 said...

Sharon F. said... 53

"I don't know how much corn comes from Mexico;"

US Agricultural Exports to Mexico - 1991/93 vs 2007/09
http://www.ers.usda.gov/Briefing/NAFTA/Data/usxmx.2007-09.xls

US Agricultural imports from Mexico -1991/93 vs 2007/09
http://www.ers.usda.gov/Briefing/NAFTA/Data/usmmx.2007-09.xls

As one would expect, US crops with a relatively low labor content benefited from NAFTA. Mexican crops with a relatively high labor content benefited from NAFTA.


USDA has done in depth studies Mexican agriculture.
http://www.ers.usda.gov/publications/FDS/may04/fds04D01/fds04D01.pdf

NewYork said...

NewYork: His basic philosophy is that since there are other factors involved, it's impossible to make a prediction, and therefore the study is "silly".

jstults: I think you are confused. Pretty clearly, the point of their argument is that, when you're making policy, the relative magnitude of the effect matters (practical significance), as opposed to focusing on whether you can measure it or not (statistical significance).

Pielke's argument is certainly not limited to relative magnitude:

Pielke: Even with the voluminous caveats in the paper, to conclude that "climate change is estimated to induce 1.4 to 6.7 million adult Mexicans (or 2% to 10% of the current population aged 15–65 y) to emigrate as a result of declines in agricultural productivity alone" is just not credible.

As for the very broad relative magnitude argument, the existence of potentially more important factors does not automatically imply the lesser factors are of none or negligible importance. That hasn't been demonstrated, and I believe Oppenheimer has addressed that argument here.

On a somewhat different note, I've read one of Pielke's papers on U.S. hurricane damage. It contains critical caveats such as not attempting to account for the development of mitigating factors over the 20th century (such as improved buildings, better warning systems). By itself, this factor would reduce hurricane damage, yet it's not included in the normalized costs, and therefore the trend is biased low. By Pielke logic, such important caveats should preclude saying anything about policy. However, that does not stop him from implying policy should account only for increased development in disaster-prone regions, and ignore the climate impact:

"However, it should be clear from the normalized estimates that while 2004 and 2005 were exceptional from the standpoint of the number of very damaging storms, there is no long-term trend of increasing damage over the time period covered by this analysis. Even Hurricane Katrina is not outside the range of normalized estimates for past storms. The analysis here should provide a cautionary warning for hurricane policy makers."

To quote Pielke "It is almost as if the paper is written to be misinterpreted."

Landsea follows up with "There is nothing in the U.S. hurricane damage record that indicates global warming has caused a significant increase in destruction along our coasts."

When Pielke and Landsea account for 20th century technological advances in their normalization process (data manipulation in skeptic-speak), I'll take such a statement more seriously.

Roger Pielke, Jr. said...

-56-New York

There have been no trends in frequency or intensity of hurricanes at landfall in the US, why would you expect there to be trend sin damage?

darcymeyers said...

So stuff is likely to happen over the next 80 years which may have an effect on immigration? One of those factors is likely climate change, but it is hard to accurately quantify? This is a highly deterministic analysis that dresses up a crystal ball. Perhaps climate change increases food production comparatively with the US and other factors like effective governance or economic prosperity reverse the migration trends over the next 80 years. Hogwash...

Tamara said...

I wonder how this model would work if applied to other instances of immigration by farmers? Based on a trend from 1840 to 1850, how much of the Irish population would be expected to have moved to America by 1920? Based on the 1930-1940 trend, would there be anyone living in the Great Plains today? We can certainly link those immigration events to effects of the climate, but we couldn't extrapolate that into the future.

S. said...

Willis Eschenbach did an has an article about mr. Oppenheimers paper on WUWT (the link is http://wattsupwiththat.com/2010/07/28/border-transgressions/)

According to him the papers author made a serious error in their calculation ,which might possibly be because the data they used came from an official Mexican website that publishes the country's agricultural data,and therefore is in Spanish , and was possibly misread because of lack language knowledge or they invented a new definition of harvest yield measurement. Turns out that to calculate their climate change and migration correlations to harvest yield they divided the acreage planted each year with the weight of harvest same year for each of the areas under scrutiny. But the planted area is almost never the same as the harvested area so for obvious reason the the usual procedure is to measure yield per harvested area, and not planted area, and as far I can see the purported correlation breaks down when things are recalculated with that in mind, and thereby invalidates the conclusion. But enough of that

I quote the following from mr.Eschenbach's article

"Because the correlations of the yield are central to their analysis, this error invalidates the paper and requires the recalculation of all the relationships. Remember that their thesis is:

Climate Change —> Reduced Mexican Crop Yields —> Migration to US

Note that there are two separate mathematical relationships in their claim. One relates climate change (temperature and rainfall) to changes in yield. The other relates changes in yield to migration rates. An error in the yield, therefore, requires a recalculation of both relationships, with new error bounds, etc."


And I think that even in a "kindergarten climate science class" this would be enough earn the paper a big "F" grade and a suggestion that the authors to go and redo their homework.

Anyway I suggest you read the whole of Eschenbach's article, I at least find illuminating.
And I have not noticed any response to it from mr. Oppenheimer or his coauthors yet. Perhaps he does not have a defendable position here , or has not become aware of the posting, but I sure would love to see the what his response on it would say.

willis said...

Roger, perhaps you could ask Dr. Oppenheimer about the use of production / area planted, a metric I have never seen used for the purpose before ... I discuss this in my article here.

Pasteur01 said...

Roger -

Referring to the image of the billboard on your original post, this paper is indeed nationalistic drivel neatly wrapped in environmental defense. Not that they intended to foment nationalism or racism, but to exploit angry feelings for the purpose of bringing attention to the authors' pet cause.

In the context of a very real crisis where bodies are piling up in the streets and deserts along the border, the authors and the publisher were at best insensitive. I might suggest an excursion down from the ivory tower to visit El Paso, Juarez, Tijuana, Nogales, etc.

As for Dr. Oppenheimer's response and the ensuing discussion, debating Silly Science is, well, silly.

EliRabett said...

Three names: Anasazi, Okies, Sahel (ok that was drought, but all signify climate changes imposed on then current conditions)

Post a Comment