In areas such as the suburbs outlying the city, is significant infrastructure built to support the driving habit? It would be interesting to see what the correlation is between GDP growth and vehicle "support industry."
This would further prove (if the correlation exists and is positive), that we are indeed borrowing GDP growth from Mother Nature in the form of cheap gas. The scales have to balance eventually.

How is the population density figure derived for each state?

If it is just the total population of the state divided by the total area of the state, then that may not tell you much. A large state may have most of it's population clustered in urban areas?

These days, I would say that the proportion of people living on the land is quite small and has gone down quite drastically in the last 50 years. Modern intensive farming techniques require fewer people to produce the same amount of produce (or even more).

I know that here in New Zealand, a lot of the population has migrated into big centres, while the rural areas have declined and sometimes even 'died'.

I agree with DuncaK. I find it too simplistic to take the population density as a measure of how urban or how rural a state is.

I think that a better measure would be to compare the population density of a state with the population density of its conties (the Census Bureau gives a very convenient xls file with all the different populations and densities). See what the variance is for each state. This will give you a measure to see if the distribution of the population is uniform (rural) or is non-uniform (urban). It looks like a fun exercise. It would be nice to complement your analysis with this information. But, unfortunately I am very busy today.

The Census Bureau also gives a nice picture:

I've always thought that transportation = business, but distance (VMT) might not be the best determinant (and it might not matter if it is mass transit or automobile). I think your trends would be much better if you were able to separate urban, suburban and rural areas. It's not just density too, how they are organized really matters. Some parts of New York City have the density suburbs and people drive a lot. On the other hand, many towns in upstate NY are decidedly rural, but they have a nearby town that can service all their needs instead of a Walmart 75 miles away.

Well, I for one am surprised that there is any sort of correlation at all.

First of all, we have the obvious fact that what we call 'states' are merely lines on a map, and there is no inherent reason why demographics, income, and driving patterns should conform to lines on a map. For example, the Chicago metropolitan area has far more in common with the New York metropolitan area than does Chicago with downstate Illinois or New York City with upstate New York. So when you deal with per capita mileage in such a state you are really dealing with an averaging of highly urban and highly rural areas.  What can that possibly tell you? My own state of Delaware is smaller than many counties in Texas but it has a highly surburanized northern part and a highly rural southern part. The income and driving patterns in each are very different.

 I really think it's more of an urban vs rural thing. (A person living in NYC can walk to a movie, theatre while someone in Wyoming might have to make a 70-mile round trip.) I strongly suspect that If you use data available in terms of Statistical Metropolitan Areas, you will see these per capita numbers to be fairly similar from one SMA to another SMA.   Then if you selected say 50 rural counties in different states at random and compared the data for them, you might see that they were not too far apart from each other.

By the way, on a day like this I'm all for your suggestion of towing Delaware down to the Carribean!

Am not sure if this affects your discussion of Delaware, but it has no sales tax.  As a consequence, anyone living in a border state, e.g., Maryland loves to drive to Delaware.  Delaware gets lots of visitors.

SS: Amazing to me, About ten years ago I manually loaded similar data into XL from the 95 world almanac, however I used per-capita personal income and selected motor vehicle stats, and came up with similar results. I remember manipulating the data and generating about 25 different columns of data & charts. At the time I was still working full time. I blew it off to the fact that across the US the value of the dollar is not equal. For instance, while traveling I never eat or stop along my route, I always exit, and find some small town, and the last surviving mom & pop restaurant or motel. There is always a kind of gebmutliehite that is not present elsewhere, corporate businesses can not provide that type of friendliness, and when they do it always seems to be not sincere. You can not buy the contentment and life of these small town folks. You have to live it, and it will never be fitted into a gated community. The value of their life style is worth much more than the value of their paycheck or earnings. After running countless service jobs in the oilfield you learn much about small town life and their values. Perhaps being a farm-boy I still have their common values.

Delaware also does not have state income tax.  It is our own home-grown tax haven.  

A couple of comments.

First, it seems to me that time spent driving is by and large wasted.  If we eliminate time spent with family or other pursuits as an alternative, there are two choices:  spend the time commuting or spend the time working.  Therefore, the greater the aggregate miles driven by a workforce, the lower the productivity (at least comparing similar workforces).   If you include personal time, greater time behind the wheel inevitably means more money spend on daycare, restaurants, etc.

Second, I did some very minimal research on obesity rates versus miles driven per year among various OECD countries.  Here in the U.S. we of course are the champions--greatest number of miles driven per year and highest obesity rate--and there appears to be a pretty consistent correlation.  The more miles you drive, the fatter you are.  It seems to me that this would also correlate to more health problems, more money spent on health care--and therefore lower productivity.  

The best advice that I can give Americans is the following: (1)  start trying to live on 50% of less of your current income and (2)  try to reduce the distance between your job and where you live to as close to zero as possible.  

The critical questions for "adaptation" are who drives what, where, and when.  My sense is that there is tremendous low-hanging fruit assoicated with each of those questions.

A suburbanite who commutes through heavy traffic every day to work may not be able to easily change where/when, but they can certainly change what they drive - perhaps slashing gas usage by 50%.

Country dwellers might car-pool to the big box store, increasing passenger-miles while decreasing VMT.

etc.

Of course, gas prices high enough to drive that kind of adaptation are certainly going to trigger some "restructuring" as some fuel uses are discarded.  Such economic turmoil could be called out as a GDP-VMT linkage, but it's really an after-affect.

Fixing your model...
Try removing the vehicle mile component from state GDP. Otherwise, you have vehicle miles on both sides of your model (imagine a state where the only economic activity was driving; the GDP and vehicle miles would perfectly correlate, violating assumptions of OLS regression).

This should make your model more robust, though the "fit" may decline (because you are reducing the identity component from the equation).

What you end up with is a relation between non-driving GDP (actual economic productivity?) and vehicle miles travelled (wasteful commuting?)

The way I would think of the puzzle is, imagine two states with the same population density. But one drives more than the other, and the one that drives more is poorer. Why is this?

Some have suggested that it could be that one state has a few big cities and little population elsewhere, while the other state has lots of small towns spread around. The first state will drive less and be richer than the second, due to the benefits of city living (which are enormous and often unrecognized (especially around here where we see some pretty romantic and unrealistic IMO "back to the land" visions)).

Another factor could be that poor people need to save money so they will drive farther to find better prices, going to Walmart instead of the corner store for example.

Or perhaps poor people tend to be less competent and efficient at economic planning (which is one reason they are poor) and so they inadvertantly waste more of their resources, including by driving unnecessarily.

let's not forget about the relationship between VMT and density.   I would be curious about the relationship between the two as well as a regression on GSP with both variables in the equation...

plus, the R2 is QUITE low for models of aggregated data on states...there's a lot of noise in there.  

Remember, also there's a lot of stuff going on at the lowest unit of analysis, the individual (but the data's not available, I know I know)...still, I would caution everyone not to pull an ecological fallacy and say anything about individual behavior based on this aggregate level data.  You can only compare state A to state B, not the people in those states or their behaviors.  

I don't think driving too much makes you poorer. I think poor people are forced to live way outside urban centers, and therefore drive more, because that's all they can afford. In these types of discussions everyone seems to overlook the fact that suburban housing is not only McMansions, but also McApartments that would cost three or four times more if they were located within walking distance of anything.

R^2 assumes normal distributions to be an accurate measure of variation explained (although it always indicates the amount of variance explained, in the technical sense - variance just makes most sense in the context of normal distributions). The two latter models are also very prone to outliers for the very same reason. Try to take one of the high-population-density samples out, and you get a clearly different result.

You can fix this by taking a logarithm of population density. If it does not make population density normal enough, then take the density to a small power instead of taking a logarithm (0.2 or so - find a power that makes it most closely normal - this is the so called Box-Cox transformation).

Your models will then not be linear in terms of the original variables, but that does not matter much, as the amount of variance explained is small anyway. Nonlinearities are buried into the noise.

Stuart,

I sit in awe at your statistical abilities here and am hesitant to point out a couple things.

One issue with Wyoming may be the combination of low population and high interstate traffic.  Interstates 80, 90, and 25 pass through the state.  You should not be counting this 'pass thru' traffic.  Coupled with the low population, this would significantly screw up the figures.  (Wyoming is 50 outta 50 in population)
http://www.infoplease.com/ipa/A0004986.html

As for Delaware, who knows?

I think the only true way to look at this is at a lower level using the Census bureau's Traffic Analysis Zones but that is probably a LOT of work.  It could be a good Master's thesis for someone though.

Wyoming is #1 state in per capita energy production.

If state GDP includes tax collections from energy severance taxes, this is likely to produce some skew.  It probably does, at least implicitly, since these revenues are generally spent to supportstate operating budgets, in lieu of income taxes, sales taxes, property taxes, etc. There are 10 states that are net energy exporters, some of which derive substantial income from well-head or mine mouth taxes on oil gas & coal.

Here is some per capita energy production data for CY 1999:
(sorry about formatting - should look OK if imported into spreadsheet)                           

state    PRODUCTION                       
abb.    TOTAL    PRIMARY    OIL    NG    COAL    TOTAL    MMBTU
    ELECT    ELECT                    PER
    QUADS    QUADS    QUADS    QUADS    QUADS    QUADS    CAPITA
WY    0.149    0.004    0.355    0.914    7.155    8.428    17573.6
AK    0.02    0.003    2.223    0.514    0.033    2.773    4476.1
WV    0.323    0.003    0.009    0    3.353    3.365    1862
LA    0.305    0.062    0.696    5.904    0.063    6.725    1538.1
NM    0.111    0.001    0.373    1.679    0.619    2.672    1536.1
ND    0.107    0.009    0.191    0.059    0.661    0.919    1450.6
MT    0.1    0.04    0.087    0.068    0.872    1.067    1208.8
KY    0.316    0.009    0.016    0    2.963    2.988    754.5
OK    0.187    0.011    0.409    1.745    0.035    2.201    655.5
TX    1.22    0.137    2.606    6.797    1.126    10.666    532.1
UT    0.125    0.005    0.094    0.292    0.56    0.951    446.3
CO    0.135    0.005    0.107    0.821    0.636    1.57    387.1
KS    0.144    0.031    0.168    0.615    0.009    0.823    310.2
AL    0.413    0.148    0.065    0.608    0.414    1.234    282.4
PN    0.664    0.257    0.009    0    1.621    1.887    157.3
IN    0.416    0.002    0.011    0    0.722    0.735    123.7
VA    0.255    0.106    0    0    0.685    0.791    115.1
IL    0.557    0.282    0.07    0    0.858    1.21    99.7
MS    0.12    0.035    0.104    0.123    0    0.263    95.1
AZ    0.286    0.138    0    0.001    0.25    0.389    81.5
WA    0.397    0.355    0    0    0.087    0.443    76.9
CA    0.63    0.328    1.584    0.425    0    2.336    70.5
OH    0.486    0.06    0.035    0    0.477    0.572    50.8
OR    0.193    0.157    0    0.001    0    0.159    47.9
SC    0.306    0.179    0    0    0    0.179    46
MI    0.354    0.062    0.045    0.308    0    0.415    42.1
AR    0.162    0.061    0.041    0    0    0.103    40.4
SD    0.036    0.023    0.006    0    0    0.029    39.8
ID    0.049    0.048    0    0    0    0.048    38.3
TN    0.319    0.12    0.002    0    0.064    0.187    34
NB    0.107    0.04    0.015    0    0    0.056    33.5
NH    0.056    0.039    0    0    0    0.039    32.3
VT    0.019    0.019    0    0    0    0.019    32
MD    0.178    0.054    0    0    0.081    0.135    26.2
NC    0.402    0.147    0    0    0    0.147    19.2
ME    0.041    0.023    0    0    0    0.023    18.1
GA    0.408    0.134    0    0    0    0.134    17.1
CT    0.095    0.052    0    0    0    0.052    15.8
NJ    0.194    0.103    0    0    0    0.103    12.6
MN    0.168    0.056    0    0    0    0.056    11.8
NY    0.495    0.21    0.001    0    0    0.211    11.6
FL    0.639    0.135    0.028    0.007    0    0.17    11.3
NV    0.105    0.015    0.004    0    0    0.019    10.4
WI    0.202    0.052    0    0    0    0.052    9.9
MO    0.252    0.035    0.001    0    0.008    0.044    8.1
IA    0.13    0.016    0    0    0    0.016    5.6
MA    0.135    0.024    0    0    0    0.024    3.9
HI    0.035    0.003    0    0    0    0.003    2.8
RI    0.023    0    0    0    0    0    0.4
DC    0.001    0    0    0    0    0    0
DE    0.023    0    0    0    0    0    0

    data from                       
    http://www.eia.doe.gov/emeu/states/_statequads.html                       
    1999 DATA PREPARED 6/29 2001               

I don't know if link shown is still valid, but that was my data source.

The scatter in your data might be reduced if you regress per capita GDP versus two variables, VM/capita and per capita energy production.  Unfortunately, this still does not explain Delaware.

Les Lambert

I'm not sure your basic hypothesis holds up.  Or that it will hold up post-peak, anyway.  Yes, urban areas produce more wealth now.  But that doesn't mean they'll continue to do so in the future.  Post-peak, will Wall St. continue to pull millions of dollars into New York City?  Stock brokers and financial analysts may be generating wealth, but they aren't really producing anything.  

However, I think you may be right that a rural homestead is not necessarily the best way to ride out the peak.  Tainter points out that often, people huddle closer to the cities as collapse approaches.  One, the government may provide jobs, food, etc., in the cities.  (Possibly confiscated from rural farmers, who have to go to the city and get on the dole if they want to eat the food they grew.)  Two, it becomes too dangerous.  Isolated farms and small villages cannot defend themselves from war parties, bands of looters, etc.  

Though Kunstler thinks the suburbs are the worst place to weather the peak, someone who survived the Argentinian economic collapse recommends them as the best:

After all these years I learned that even though the person that lives out in the country is safer when it comes to small time robberies, that same person is more exposed to extremely violent home robberies. Criminals know that they are isolated and their feeling of invulnerability is boosted. When they assault a country home or farm, they will usually stay there for hours or days torturing the owners. I heard it all: women and children getting raped, people tied to the beds and tortured with electricity, beatings, burned with acetylene torches.

Big cities aren't much safer for the survivalist that decides to stay in the city. He will have to face express kidnappings, robberies, and pretty much risking getting shot for what's in his pockets or even his clothes.

So, where to go? The concrete jungle is dangerous and so is living away from it all, on your own.

The solution is to stay away from the cities but in groups, either by living in a small town-community or sub division, or if you have friends or family that think as you do, form your own small community.

Some may think that having neighbors within "shouting" distance means loosing your privacy and freedom, but it's a price that you have to pay if you want to have someone to help you if you ever need it. To those that believe that they will never need help from anyone because they will always have their rifle at hand, checking the horizon with their scope every five minutes and a first aid kit on their back packs at all times... Grow up.    

A few parting comments re this particular thread:

While I am aware that a lot of statistical data is collected on a state-by-state basis, the reason for doing so is purely historical and bureaucratic,  and there is no valid statistical reason for collecting data in this manner. This is particularly true in the environmental field, where air pollution, and to a lesser extent water pollution and solid waste movements, have absolutely nothing to do with the location of political boundaries.

For example, it is next to ludicrous to try to correlate air quality in my home state of tiny Delaware to say Delaware's GDP or miles traveled, etc. The simple and obvious reason is that the wind does not come to a screaching halt at the borders between Delaware and PA, NJ, and MD.  We contribute to the air quality in the neighboring areas of PA, NJ,and MD, and they contribute to the air quality in Delaware.  Furthermore, the major corridor of I-95, one of the most heavily travelled arteries in the country, runs smack through the northern part of the state.

So, the valid statistical unit for the analysis of air quality should be a region consisting a circle of perhaps 50 miles radius encompassing northern DE, parts of southeastern PA, a chunk of NJ including the Camden area and suburbs, plus perhaps a tiny peice of northeastern MD. The EPA has recognized this problem and that is why they often divide regions up into
'air quality districts' for the purpose analyzing air quality.  

(By the way, Delaware has one of the highest per capita cancer rates in the country, but I believe this is a statistical artifact stemming from the fact that the heavily populated northern part of the state is epidemiologically indistinguishable from the greater Philadelphia area.)  

Regarding the driving habits of poor vs affluent people: it's been my observation that i) poor people don't think much in terms of gas mileage and tend to drive the cheapest car they can find, even if it's an old beater of a gas-guzzler,  ii) poor people will drive long distances to find whatever work they can, and iii) the further you are from urban areas and the near suburbs, the cheaper the housing. Convenience can be very  expensive - something that someone like Kunstler doesn't seem to fully appreciate.  

On a lighter note, I would just like to point out that the character Miller in the movie "Repo Man" made the following keen observation:

"The more you drive, the less intelligent you are."

Carry on with your real discussions!

The role of commuting-related gas use can be broken down through Census data; Matt Kures has estimated that $.50 increase in gas price relates to over $300 million in increased gas spending annually for Wisconsin commuters. Detailed report with maps of "stretch commuters" can be found here:
http://www.uwex.edu/ces/cced/mwe/

Stuart,
"50 of the 52 US states."
when did the US acquire 52 states? The CIA in their fact book still think it is 50 plus 1 district.

I'm afraid that the problem with this analysis is the geographic scale you have chosen.  There is simply too much variation with the urban and rural landscapes within states for a meaningful comparison across states.  I understand the data issues, but that is part of the research challenge.  

There are a few journal articles that take on urban form and energy consumption that you may find interesting.  It's a slightly different research question, but closely related to what you are investigating.  Below are references for a journal article that reviews this literature and a book that updates a classic study on energy and urban form.      

A good lit review (I can email this to you if you can't access it through a university library):

Anderson, W. P., Kanaroglou, P. S., & Miller, E. J. (1996). Urban Form, Energy and the Environment: A Review of Issues, Evidence and Policy. Urban Studies, 33(1), 7 - 36.

Summary. The spatial configuration of cities and its relationship to the urban environment has recently been the subject of empirical, theoretical and policy research. Because of the disciplines involved, relevant articles are scattered over a large number of journals. The objective of this paper is to put the issues in perspective by reviewing the basic concepts and relationships involved, and to evaluate critically the current state of knowledge about urban form, energy utilization and the environment. The scope of the paper is limited to urban transport energy use and the associated emissions. Suggestions for further progress in the field are offered , with emphasis placed on integrated urban models as useful and policy-sensitive analytical tools.

The following book contains an update of a classic study on energy consumption and urban form that you may also find interesting:

http://books.google.com/books?ie=UTF-8&hl=en&vid=ISBN1559636602&id=pjatbiavDZYC&dq=S ustainability+and+Cities:+Overcoming+Automobile+Dependence&prev=http://books.google.com/books%3F q%3DSustainability%2Band%2BCities:%2BOvercoming%2BAutomobile%2BDependence%26lr%3D%26client%3Dfirefox -a&pg=PP1&printsec=0&lpg=PP1&sig=AFv85BESXwVZD8rdHLa9hoQvLGk

I'm sure I remember reading a study somewhere that the strongest factor in surviving adversity is a strong community.  Living in the back of beyond is no good if you don't have people around you who can pull together.

Personally I have moved out of the city into a small town (pop 10000) which has very strong sense of community, the lowest number of chain stores in the UK, meaning most things can be bought locally.  Very high interest in sustainability / organic farming. Good public transport links etc...

Society can more easily create value, or gdp, when large concentrations of people work together. At the same time, cities separate workers from close proximity to their jobs. On the farm, you walk outside and you are on the job site. Or, in a small town, everybody can walk to work.  In cities, you may live several miles from work, and need to go there five days/week. So cities are high energy consumers that lure workers with both entertainment and high salaries, or at least high enough to provide the energy necessary to get to and from work.

The larger the city, the greater the average separation from work, and the more energy/capita required to maintain production. Most cities have grown to large size during the age of cheap energy, and some may be larger than what would be ideal in a high-cost energy environment.  At the same time, the small-town model would not be a good replacement because it requires too much land area.

As the cost of energy rises cities (and all other consumers) will become more energy efficient. People will move closer to work, smaller companies will move to the burbs, and/or people will tellecommute.

I believe the reason that driving results in less GDP/statistical area is that a high percentage of the cost of the gasoline is not counted in the GDP of the particular statistical area. It is profit that is creamed off the top and returned to the oil producing areas or oil company headquarter locations (Canada, NY, Alaska and perhaps Texas and even Oklahoma and we can probably add Deleware) and operating division areas in the form of cost of production and profits. Thus it is my theory that if you do the same analysis on Canada, OPEC and the states of Texas, Oklahoma, Alaska and Louisiana, they will be seen to have unusually high "GDP" in relation to per-capita miles driven.

The states with lower population densities are mostly states with right to work for less laws. Wages are lower and absentee ownership of businesses and even large agricultural operations exports GSP to urban states. Many if not most small family farms have at least one family member who must commute long distances to some low wage retail or health care job in order to keep that farm in the family.

Fantastic work, great site.

A couple of loose ideas...

1.Transportation is often of goods, especially retail. Service industries are less intensive I would assume, in fuel terms.
Retail has largely been on a dowturn for five years or so (depending on region) and margins have been squeezed, profits are poor, wages are poor etc...

2.People with assets are often - ipso facto - richer. Assets are homes. Mant are  self employed. The richer self employed person doesn't travel so far/owns thier own home/has seen it's value rise...

3. Driving jobs are poorly paid very often. Taxis, van drivers, removals etc...

Just a few late evening thoughts...

    I understand that for many years there was a very strong correlation between the amout of bananas imported into Greta Britain and the growth of the Royal Navy. Perhaps the correlation being discussed here is a similar one.