9.30.2011

Geography of stream names

Right at the intersection of geography and text analysis is this map by Derek Watkins showing what local  waterways are called across the lower 48 states. If this isn't a latent variable for a major axis of cultural variation I don't know what is.

via The Atlantic.

9.29.2011

The Agony of Somalia

It can be easy to let the abstract nature of research obscure the relevant aspects of your work. When you work in areas like development that abstraction can come as a cost, and sometimes reminders are in order. To that end, I point you to musician K'Naan's NY Times account of returning to his homeland, Somalia, during the middle of a major famine (previously covered here):
...And then come the makeshift camps set up for the many hungering displaced Somalis. They are the reason I am here. If my voice was an instrument, then I needed it to be an amplifier this time. If my light was true, then I needed it to show its face here, where it counts. Nothing I have ever sung will matter much if I can’t be the mouth of the silenced. But will the world have ears for them, too?
I find the homeless Somalis’ arms open, waiting for the outside world and hoping for a second chance into its fenced heart. I meet a young woman watching over her dying mother, who has been struck by the bullet of famine. The daughter tells me about the journey to Mogadishu — a 200-mile trek across arid, parched land, with adults huddling around children to protect them first. This mother refused to eat her own food in order to feed abandoned children they had picked up along the way. And now she was dying because of that.
I highly suggest reading the rest, and keeping it in mind when you think about the extended social impacts of climate.

Different perspectives on climate and conflict

Nature Climate Change had a special feature article on climate and conflict in the current issue.

Perhaps more interesting than the coverage of our paper (which received more than its fair share already) are the various perspectives, stories and interpretations that Ms. Jones, the writer of the article, obtained from different researchers in the field.  Read it here.

A map of previous conflicts (1980–2005) attributed
to environmental issues, their sizes and causes.
Credit: Nature Climate Change

The only substantive comment I'll add is that the correlation between ENSO and conflict doesn't seem to be weakening in our data.  Jones writes
This 'recent-resilience' effect can also be seen in Africa: include very recent data and the link between heat and conflict starts to evaporate, says Buhaug. Even Miguel agrees that the link is disappearing, thanks in large part to recent economic recovery of some African nations and an overall decline in conflicts on the continent over the past 15 years. “If the trend in Africa continues, and this relationship continues to weaken, that would be a wonderful thing,” he says. Homer-Dixon, however, is sceptical. “We're seeing resurgent conflict in many places now. When the current food crisis hit, it knocked a lot of countries back.”
When our paper was being reviewed, my coauthors and I checked if this weakening was visible in our data since one of our referees pointed out that the relationship in Burke et al. weakened in the last decade. Unfortunately, if there is any trend in the strength of the link between ENSO and conflict, it's a strengthening of the relationship in the last two decades.  Because we weren't certain if this trend was meaningful (we have a very small sample to begin with) we didn't include this result in the paper.  But it is somewhat concerning, from a social perspective, that we don't find the weakening trend that one might expect.  We're not sure why such weakening isn't visible, but my coauthor Kyle Meng is using new datasets and analysis to try and figure it out (so stay tuned).  One hypothesis is that in our rapidly globalizing world, planetary-scale spatial correlations in environmental factors play an increasingly important role in commodity markets or geopolitics, but this is still only a hypothesis.

9.27.2011

Tide prediction and D-Day

The September issue of Physics Today has an excellent article on the use of analog (i.e., physical as opposed to electronic) computers to predict tides during World War II, and in particular during the planning of D-Day:

Farquharson became responsible for finding a way to provide Doodson with the harmonic constants—or the data to calculate them—he would need for making accurate tide predictions for the landing beaches. But because of extremely tight security, Farquharson could not tell Doodson the biggest secret of the war—that the landings were to take place in Normandy. So he labeled the coastal location to which a particular set of harmonic constants or tidal data belonged with a code name. And he, in Bath, and Doodson, in Liverpool 200 km away, used those code names in their exchange of letters and telegrams.

To produce accurate tide predictions for the Normandy beaches, Doodson ideally needed accurate harmonic constants calculated from tide data measured at or very near each landing beach. But the British had no such data. The only harmonic constants they had were for the two French ports that bracketed the beaches: Le Havre to the east and Cherbourg to the west. Simple interpolation would not work, because shallow-water conditions varied from place to place. Shallow-water distortion of the tide can speed up the rise from low to high water, giving demolition teams less time to do their work. Very accurate shallow-water tidal data from the beaches themselves were therefore essential.

The 1943 Admiralty Tide Tables did include three locations near the Normandy beaches. But all three were secondary stations, and the tables contained warnings about the probable inaccuracy of predictions for those locations. Now Farquharson desperately needed tide data direct from the invasion beaches. So British teams in small boats and midget submarines carried out several secret midnight reconnaissance missions on the enemy beaches. Those dangerous missions did yield a few tide and current measurements, but much less than is normally required for tidal analysis.

On 9 October 1943, Farquharson sent Doodson a three-page handwritten letter marked “MOST URGENT” (see figure 3). It included 11 pairs of harmonic constants for the location code-named “Position Z.” He asked Doodson to produce hourly height predictions for four months commencing 1 April 1944. “The place is nameless and the constants inferred,” he wrote. “There is in fact very little data for it. I am gambling on the inferred shallow-water constants giving something like the right answer.”
The next time I'm stressing out about slow computation / missing data / coding difficulties, I'm going to stop, breathe, and say "well, at least I'm not trying to figure out how to land 160,000 soldiers on a beach at a specific point in the tidal cycle at a specific time of day so that we can defeat the Nazis."

hat tip: Maggie Koerth-Baker of boingboing

9.26.2011

Credit markets, unique identifiers and biometrics

Once in a Starbucks I met a guy who worked in the US credit-rating industry and he told me that one of the biggest under-stated issues in the credit market was the miss-assignment of credit penalties because individuals have the same name (sorry, John Smith). He instructed me to go and check my credit scores to make sure the rating agencies weren't assigning me penalties that were due to someone else's failure to repay a loan (but given my name, I wasn't terribly concerned).  However, I was completely stunned, given the state of technology, that we still rely on names to uniquely identify people.

The same issue hold in poor countries, except it's exacerbated by weaker credit-rating institutions and greater difficulty tracking individual people.  This recent NBER working paper illustrates the importance of unique identification for credit markets and demonstrates how biometric technologies can help.

Credit Market Consequences of Improved Personal Identification: Field Experimental Evidence from Malawi
by Xavier Gine, Jessica Goldberg, Dean Yang

Abstract: We report the results of a randomized field experiment that examines the credit market impacts of improvements in a lender's ability to determine borrowers' identities.  Improved personal identification enhances the credibility of a lender's dynamic repayment incentives by allowing it to withhold future loans from past defaulters and expand credit for good borrowers.  The experimental context, rural Malawi, is characterized by an imperfect identification system. Consistent with a simple model of borrower heterogeneity and information asymmetries, fingerprinting led to substantially higher repayment rates for borrowers with the highest ex ante default risk, but had no effect for the rest of the borrowers.  The change in repayment rates is driven by reductions in adverse selection (smaller loan sizes) and lower moral hazard (for example, less diversion of loan-financed fertilizer from its intended use on the cash crop).

9.22.2011

Income, Emotion and Life Evaluation

Another interesting 2010 PNAS article I missed:

Daniel Kahneman and Angus Deaton

Abstract: Recent research has begun to distinguish two aspects of subjective well-being. Emotional well-being refers to the emotional quality of an individual's everyday experience—the frequency and intensity of experiences of joy, stress, sadness, anger, and affection that make one's life pleasant or unpleasant. Life evaluation refers to the thoughts that people have about their life when they think about it. We raise the question of whether money buys happiness, separately for these two aspects of well-being. We report an analysis of more than 450,000 responses to the Gallup-Healthways Well-Being Index, a daily survey of 1,000 US residents conducted by the Gallup Organization. We find that emotional well-being (measured by questions about emotional experiences yesterday) and life evaluation (measured by Cantril's Self-Anchoring Scale) have different correlates. Income and education are more closely related to life evaluation, but health, care giving, loneliness, and smoking are relatively stronger predictors of daily emotions. When plotted against log income, life evaluation rises steadily. Emotional well-being also rises with log income, but there is no further progress beyond an annual income of ~$75,000. Low income exacerbates the emotional pain associated with such misfortunes as divorce, ill health, and being alone. We conclude that high income buys life satisfaction but not happiness, and that low income is associated both with low life evaluation and low emotional well-being.


("Ladder" is a 10 point scale about how satisfied
an individual is with their life.)

[The findings on emotions remind me of this comedy routine.]

9.21.2011

Hail cannons?

My sister just emailed me from France to tell me that all the vineyards they were visiting talked about weather modification via "hail cannon."  I had never heard of this and did a quick search to see what it was and if there's any evidence that the technology works.

hail cannons in 1901
modern hail cannon installation
(from boingboing)
Apparently, it's a 100 year old technology with [perfect] properties that make it impossible for individuals on the ground to evaluate whether it works or not.  The manufacturer's claim is that directed explosions on the ground can generate shock waves that interfere with hail stone formation.  More "support" for this claim is here, but the latest scientific paper cited is a theory paper from 1966 (be sure to check out this entertaining "explanatory" animation).

It doesn't seem like anyone has seriously tried to evaluate whether this approach works. In 1981, a review of the evidence suggested that our understanding had changed little between 1902 and 1975:

From: History Repeated: The Forgotten Hail Cannons of Europe
Stanley A. Changnon Jr. and J. Loreena Ivens (BAMS, 1981)

A more recent review contains additional interesting history:
[H]ail cannons are the best known apparatus to fight hail by force.  They direct the sound of an acetylene explosion, more than 120 dB, upward by way of a conical vertical muzzle. Originally these cannons were developed in 1896 by Albert Stiger, an Austrian winegrower.  When in Stiger's valley no hail had fallen for two  years, employing these cannons became an immense craze in Austria and in northern Italy.  A cannon industry  developed, and over ten thousand cannons were employed in the region around 1900.  At a hail conference in Lyon in 1901, many different makes of cannon were offered for sale (Fig.5).  However, when regularly locations with plenty of cannons were heavily damaged by hail, public confidence decreased and the authorities organized a systematic experiment  of several years in some Austrian and Italian regions.  By the time that this experiment was concluded in 1906, most farmers already had sold their useless cannons as scrap iron (PERNTER, 1907; ODDIE, 1965; CHANGNON and IVENS, 1981)
RCT anyone?

9.19.2011

Global Food, Environment and Economic Dynamics

Some colleagues and I put together a new site that focuses on global-scale empirical research on links between environmental conditions, food production and economic responses:

g-feed.org

It's currently pretty simple, aggregating some work and datasets related to these issues, but hopefully it will continue to grow.

If you have any ideas/suggestions for the site, let me know in comments.

9.15.2011

Heisenberg ruins social science too

Well, not really "ruins," but it seems that the uncertainty principle might be alive and well in social sciences too. I'm a bit late on this, since it came out in PNAS a few months ago. But obviously still worth it.

Being surveyed can change later behavior and related parameter estimates
Alix Peterson Zwane, Jonathan Zinman, Eric Van Dusen, William Pariente, Clair Null, Edward Miguel, Michael Kremer, Dean S. Karlan, Richard Hornbeck, Xavier Giné, Esther Duflo, Florencia Devoto, Bruno Crepon, and Abhijit Banerjee
(Edited by Eric S. Maskin, Institute for Advanced Study, Princeton, NJ)
Does completing a household survey change the later behavior of those surveyed? In three field studies of health and two of microlending, we randomly assigned subjects to be surveyed about health and/or household finances and then measured subsequent use of a related product with data that does not rely on subjects' self-reports. In the three health experiments, we find that being surveyed increases use of water treatment products and take-up of medical insurance. Frequent surveys on reported diarrhea also led to biased estimates of the impact of improved source water quality. In two microlending studies, we do not find an effect of being surveyed on borrowing behavior. The results suggest that limited attention could play an important but context-dependent role in consumer choice, with the implication that researchers should reconsider whether, how, and how much to survey their subjects.
Perhaps the IRB should tax RCT's for polluting each other's experiments?

9.14.2011

Volunteer data work and a data visualization contest

Two quick notes for code and stats -inclined Fight Entropy readers:
  1. Data Without Borders is holding their first event in New York next month. If you're an empiricist looking to do some good or an NGO in need of some data help, check them out here
  2. The team over at Information is Beautiful have announced a (recurring?) data visualization contest, The Information is Beautiful Awards
Both look potentially fun and rewarding.

AEA session for the winter meeting

Sponsored by Fight-Entropy! (not really...) The full preliminary program is here.


Jan 07, 2012 2:30 pm, Swissotel, Vevey 1 
Association of Environmental & Resource Economists
Environmental Constraints and Land-Use Decisions(Q1)
PresidingMAXIMILIAN AUFFHAMMER (University of California-Berkeley)
When the Levee Breaks: Land, Labor, and Capital in the Deep South
SURESH NAIDU (Columbia University)
RICHARD HORNBECK (Havard University)
The Impact of Climate Change on Crop Choice in the United States
ROBERT MENDELSOHN (Yale University)
ZHIMIN LI (Yale University)
NAMRATA KALA (Yale University)
Climate and the Locations of Crops
SOLOMON HSIANG (Columbia University)
DAVID LOBELL (Stanford University)
MICHAEL J. ROBERTS (North Carolina State University)
WOLFRAM SCHLENKER (Columbia University)
JARROD WELCH (University of California San Diego)
Economic Impacts of Climate Variability and Climate Change: Evidence from a Quasi-Experiment with Great Lakes Water Levels
MICHAEL MOORE (University of Michigan-Ann Arbor)
HSING-HSIANG HUANG (University of Michigan-Ann Arbor)
Discussants:
JEFFREY VINCENT (Duke University)
JARROD WELCH (University of California-San Diego)
MAXIMILIAN AUFFHAMMER (University of California-Berkeley)
OLIVIER DESCHENES (University of California-Santa Barbara)

9.13.2011

Natural disaster management game

I know some people really love designing simulation games to educate people. I have no strong feelings on this and also have no idea if this kind of training works (RCT anyone?), but here's one I found today: www.stopdisastersgame.org

It's sponsored by the UN International Strategy for Disaster Reduction.

If you hate fun as much as I do, you can read this instead.


9.08.2011

El Niño and Future Climate Change

Over the last two weeks, my coauthors and I have answered a lot of questions about the relationship between El Niño and future climate changes, so I figured it was worth a quick post with helpful links.

The last report from the International Panel on Climate Change says in its executive summary
Multi-model averages show a weak shift towards average background conditions which may be described as ‘El Niño-like’, with sea surface temperatures in the central and east equatorial Pacific warming more than those in the west, weakened tropical circulations and an eastward shift in mean precipitation. 
All models show continued El Niño-Southern Oscillation (ENSO) interannual variability in the future no matter what the change in average background conditions, but changes in ENSO interannual variability differ from model to model. Based on various assessments of the current multi-model data set, in which present-day El Niño events are now much better simulated than in the TAR, there is no consistent indication at this time of discernible changes in projected ENSO amplitude or frequency in the 21st century.
with more detail provided here.


The single figure that tells the main story:


The figure summarizes the output of the different climate models according to two measures. On the horizontal axis, it shows how much the spatial pattern of future warming looks like the spatial pattern we observe during El Nino events.  The fact that most of the points are on the right hand side indicates that future warming probably will look "El Nino-like" based only on its spatial distribution.  On the vertical axis, the figure shows what happens to "inter-annual variability," which is similar to a normalized standard deviation of a NINO3-like index.  The fact that the points are evenly distributed above and below one indicates that we really don't know if El Nino events will become more or less frequent or extreme.

9.07.2011

More on climate, economics and conflict


Tom Vogl points us to a recent article of the NBER Reporter where Ted Miguel thoughtfully lays out a brief review of links between weather/climate changes, economic impacts and civil conflict.  He argues that the broader issue of civil conflict represents an important and under-studied topic in economic development:
Civil war is clearly central in the study of international economic development, yet leading development economists have often overlooked it, and some undergraduate textbooks do not even mention the issue. Over the past decade, however, many economists and other social scientists have worked to better understand the causes and the economic legacies of internal warfare, often in collaboration with political scientists and other scholars. The main goal of this research summary is to describe some of this progress, with a particular focus on the role of economic shocks, weather, and climate in driving the patterns laid out above...
and puts forward a few suggestions
While deriving policy implications is not the main goal of this summary, the literature does have certain implications. For example, the empirical relationship between violence and low and falling incomes suggests that implementing insurance schemes to protect poor societies from negative income shocks might reduce future rounds of bloodshed. One possibility is expanded regional drought insurance for farmers. Another is foreign aid contingent on objective conflict risk indicators (for example, weather or commodity price shocks, or a coming El Nino year) – what I have elsewhere called “rapid conflict prevention support” – to bolster local economic conditions when the risk of particularly political violence is high.

I particularly appreciated Ted pointing me towards this recent article in ReStat:

Economic determinants of land invasions
F. Daniel Hidalgo, Suresh Naidu, Simeon Nichter, and Neal Richardson

Abstract: This study estimates the effect of economic conditions on redistributive conflict. We examine land invasions in Brazil using a panel data set with over 50,000 municipality-year observations. Adverse economic shocks, instrumented by rainfall, cause the rural poor to invade and occupy large landholdings. This effect exhibits substantial heterogeneity by land inequality and land tenure systems, but not by other observable variables. In highly unequal municipalities, negative income shocks cause twice as many land invasions as in municipalities with average land inequality. Cross-sectional estimates using fine within-region variation also suggest the importance of land inequality in explaining redistributive conflict.

9.05.2011

"Adaptation and the envelope theorem"


Here's a simple, elegant and important point about the economics of climate change, but it applies to other environmental changes equally well. (I was recently at an entire conference dedicated to the economics of adaptation, and nobody mentioned this idea.)

William Nordhaus writes in a 2010 paper published in the journal Climate Change Economics (emphasis added):
Adaptation and the envelope theorem 
Including potential adaptation is beyond the scope of the current study. However, if changes in the means and higher moments of environmental parameters are small or gradual, and if agents make decisions on the basis of appropriate expectations, then omitting adaptation will have, to a first approximation, no effect on correctly measured damages. The reason is due to the “envelope theorem” of decision making. Under this result, the first-order cost of changing environmental conditions is equal to the first-order cost of adapting to those conditions. Of course, if environmental conditions change very rapidly, expectations are wildly inaccurate, or the cost of adapting is very non-linear, then second-order effects come into play. We would then need to consider adaptation costs explicitly.
What he's saying is that in the current equilibrium, individual's investment in adaptation to the current climate should be optimal (or close, given constraints/distortions).  And if it's optimal, this means the marginal benefits of additional adaptation are equal to the marginal costs.  So if we introduce a small change to the current climate such that it becomes optimal to adapt a little more, we will adapt slightly more at the current marginal cost and only reap exactly the same amount in marginal benefits (since the two are equal).  This means, we don't "win" by adapting. Instead, we just adapt slightly more at zero net benefit, so the overall social cost of the climatic shift remains unchanged.

Now, if only I could remember where I left my copy of MWG...