Flazatron

Better Image Capture

2010-02-21T16:57:00.000-08:00

I tried to figure out what webcams would work from Linux and allow me to control the shutter speed. The existence of setpwc focused me on Logitech webcams. The marketing copy didn't mention shutter speed but did mention something called "RightLight2" technology which I figured was related (parenthetically, it seems like as a particular technology improves, it becomes less intelligible to technologists; the reason being of course is that it becomes more intelligible to the non-technologists who outnumber us). Anyway I bought a Logitech C905 which was like $70. Sadly, it was not compatible with setpwc because it is too new: it uses the uvcvideo driver. However whatever the default settigs are on the C905, the image capture looked good in broad daylight. Plug and play for once ...

The real problem was mounting the camera. My cheap $10 webcam with horrible shutter speed had a perfect suction cup mount. The C905 comes with a clip that I'd hoped would clip to my rear-view mirror but the mirror was too thick on account of fancy auto-dimming logic. So I dismantled the cheap $10 webcam down to the suction cup and then tied the C905 to it. Hey, I paid $10 for a suction cup ...

Now I have some real data: we had to go to a birthday in Culver City today so I did a capture for the trip, which included a police car.

The next step will be some tools to manage monster collections of image training data because I had to manually binary search through the pile of today's capture to find this police car.

Feature Extraction

2010-02-04T12:49:00.001-08:00

I'm still trying to figure out a webcam which is Linux compatible with a shutter speed that admits capturing motion in direct sunlight. Until then, more armchair computer vision 101 ...

Winn et. al. use a standard set of filters for extracting features from images: Gaussian, derivatives of Gaussian (aka DoG), and Laplacian of Gaussian (aka LoG). Here's what they look like:


Gaussian
Laplacian of Gaussian
Y-derivative of Gaussian
X-derivative of Gaussian

OpenCV has a method for computing GaussianBlur which is sufficiently flexible to reproduce Winn et. al., but to reproduce their usage of DoG and LoG I wrote my own OpenCV kernels (although maybe a combination of the GaussianBlur method with the Laplacian method and Sobel method would work?). It turns out separable filters are much faster and everything above except the LoG is separable; in addition the LoG can be written as the sum of two separable filters, the second derivative of Gaussian in x and y. Thus with something like


#include <string>
#include "cv.h"
#include "highgui.h"

#include "cyclopsutil.hh"

using namespace cv;
using namespace cyclops;

int main (int, char** argv)
{
  string file (argv[1]);
  Mat img = imread (file);
  vector<Mat> color (3);

  assert (! img.empty ());
  assert (img.channels () == 3);

  cvtColor (img, img, CV_BGR2Lab);

  split (img, color);

  imwrite (file + ".L.png", color[0]);
  imwrite (file + ".a.png", color[1]);
  imwrite (file + ".b.png", color[2]);

  for (int sigma = 1; sigma < 16; sigma *= 2)
    {
      string labels[3] = { string ("L"), string ("a"), string ("b") };
      Mat conv = color[0].clone ();
      Mat convtwo = color[0].clone ();
      std::stringstream out;
      out << sigma;

      GaussianBlur (color[0], conv, Size (6 * sigma + 1, 6 * sigma + 1), sigma, sigma, BORDER_REPLICATE);
      imwrite (file + "." + labels[0] + ".gauss." + out.str () + ".png", conv);

      sepFilter2D (color[0], conv, -1, dygauss_kernelx (sigma), dygauss_kernely (sigma), Point (-1, -1), 0, BORDER_REPLICATE);
      imwrite (file + "." + labels[0] + ".dygauss." + out.str () + ".png", conv);

      sepFilter2D (color[0], conv, -1, dxgauss_kernelx (sigma), dxgauss_kernely (sigma), Point (-1, -1), 0, BORDER_REPLICATE);
      imwrite (file + "." + labels[0] + ".dxgauss." + out.str () + ".png", conv);

      sepFilter2D (color[0], conv, -1, dyygauss_kernelx (sigma), dyygauss_kernely (sigma), Point (-1, -1), 0, BORDER_REPLICATE);
      imwrite (file + "." + labels[0] + ".dyygauss." + out.str () + ".png", conv);

      sepFilter2D (color[0], convtwo, -1, dxxgauss_kernelx (sigma), dxxgauss_kernely (sigma), Point (-1, -1), 0, BORDER_REPLICATE);
      imwrite (file + "." + labels[0] + ".dxxgauss." + out.str () + ".png", convtwo);

      imwrite (file + "." + labels[0] + ".lapgauss." + out.str () + ".png", conv + convtwo);
    }

  return 0;
}

applied to our awesome cop car picture's luminosity component we have:


Gaussian (sigma = 1)
Gaussian (sigma = 8)
DxGaussian (sigma = 1)
DxGaussian (sigma = 8)
DyGaussian (sigma = 1)
DyGaussian (sigma = 8)
Laplacian of Gaussian (sigma = 1)
Laplacian of Gaussian (sigma = 8)

Initial Data Collection Attempt

2010-01-31T21:18:00.001-08:00

Well I finally got the webcam and the netbook into the car and starting doing video capture. The good news is, the suction cup mount on my $10 webcam is perfect for the windshield.

The bad news is, it turns out $10 webcams do not automatically adjust gain. Driving around during the day yields a time-lapse movie of what appears to be the afterlife. Here's a daytime driving still:

I see dead people! Twilight looks a little better:

But many of the stills have alot of blur in them, perhaps the low quality CCD has a slow shutter speed.

Maybe if I can find a software shutter speed control then I can both operate during the day and eliminate most of the blur. Otherwise, I'm going to have to pony up for a better camera.

Computer Vision 101

2010-01-26T20:50:00.000-08:00

Since it's been over a decade since I've done anything computer vision related, I need to acquaint myself with progress in the space. I figure reproducing a current paper should get me in the zone; furthermore, I'm familiar with latent Dirichlet allocation (LDA) as applied to text and I'm aware that is has been applied to images, so I figure that's a good place to start.

Wang and Grimson describe an extension to LDA which incorporates spatial correlation; but there is enough to confuse me already, so I need something involving vanilla LDA (which I'm familiar with). Going back a bit further, Sivic et. al. describe an application of vanilla LDA to image object recognition. Since LDA applies to discrete documents (that is, something consisting of "words"), the images need to be decomposed into sets of tokens. The approach taken is

Extract feature vectors from local patches of the image.
Use vector quantization to reduce the feature vectors to a codebook.

Sivic et. al. extract features from interest regions, whereas Wang and Grimson follow the approach of Winn et. al. and extract features densely. Interest regions are just another way for to get confused and screw something up, so I'll defer that for now.

Step #1 for Winn et. al. is to decompose the image from RGB to CIELAB. CIELAB is a cool encoding of color space which is supposed to mimic the response of the human visual system, so that makes sense. Poking around for an implementation of a converter, I came across OpenCV, which has tons of routines in it including color conversion. Writing a program to split an image into CIELAB components is pretty straightforward:

#include <cassert>
#include <string>
#include "cv.h"
#include "highgui.h"

using namespace cv;

int main (int, char** argv)
{
  string file (argv[1]);
  Mat img = imread (file);
  vector<Mat> color (3);
  assert (! img.empty ());
  assert (img.channels () == 3);

  cvtColor (img, img, CV_BGR2Lab);

  split (img, color);
  imwrite (file + ".L.png", color[0]);
  imwrite (file + ".a.png", color[1]);
  imwrite (file + ".b.png", color[2]);

  return 0;
}

(I guessed that a loaded .jpg image would be BGR color order, based upon the documentation for imwrite). So let's see this in action: here's a picture of a police car.

Here's the L (luminosity) channel:

Here's the a* (magenta vs. green) channel:

Here's the b* (yellow vs. green) channel:

Car Brain!

2010-01-25T20:23:00.001-08:00

My New Year's resolution is to give my car a rudimentary visual system, which should give me plenty to blog about.

My initial goal is to train the car to identify police cars, mount 4 cameras to allow for panoramic view, and have some kind of audible or visual warning. Is identifying a police car feasible? I'm not a computer vision expert, but searching around these points appear salient:

single object class: the question is binary, "does this picture contain a police car?", which makes the problem easier.
pose variability is low: the cameras will be mounted at fixed points at my car, and both my car and any police car will (barring the "Dukes of Hazzard" scenario) have all four wheels on the ground, which should moderate the pose variability.
volatile illumination: ideally, the detector would operate under all weather conditions, day or night. so that means significant illumination changes.

Well, we see how far I get anyway.

Step #1 was to get a brain for my car, so I purchased a refurb Dell Mini 110-1030nr for $280 from Amazon and put Ubuntu Netbook Remix on it. I also got 1 webcam for now.

It is presumably underpowered for the image processing that will be required, but if I get that far that will justify spending more money. First, I need to install it in the car and have it do video capture; from the resulting data I imagine many interesting problems will suggest themselves.

Did Bush Spend the Peace Dividend?

2009-06-14T21:20:00.001-07:00

I'm no fan of Bush, but I find myself arguing against certain misconceptions about his fiscal performance which continue to persist. One such misconception is that he spent an insane amount of money on the military. Nominally (that is, in raw dollar amounts) this appears to be true, but while he did increase funding relative to the Clinton administration it was still very low by historical standards.

So let's start with a picture. Here is a graph of defense spending as a percentage of GDP from 1940 to 2003.So the end of the cold war did result in a peace dividend (lower defense spending) and this situation persisted through 2003. Bush did increase funding from the low point in 2000 but only to circa 1994 levels.

After 2003 the picture is more complicated because the Iraq war spending is not reported. I took Iraq war spending numbers and combined them with reported military outlays and then divided by nominal GDP. The result is summarized in the following graphic, where I include the levels from corresponding years in previous decades for reference.

Here we can see that even in 2006 with the Iraq war in full swing we spent less on total military expenditures relative to GDP than during 1976 (with the Vietnam war fully wound down) or 1986 (at the height of the cold war). The 1990s were definitely a period of relative low military expenditures but in historical perspective Bush's spending on defense was not very high even with Iraq war expenditures accounted for (and by the way, the entire practice of keeping things off the budget is very distasteful, shame on the Bush administration for that).

In particular we can still claim a peace dividend during the 2000s relative to the cold war.

On the Record

2009-06-09T11:37:00.000-07:00

The RIAA has released 2008 Year-End Shipment Statistics which allows us to appreciate graphically just how screwed the record industry is. Compact Disc sales are plummeting, legal digital downloads are growing and the net impact is (significantly) negative. Here are the results (adjusted for inflation, which makes the story even worse!):

One popular theory is that, like newspapers, record labels have lost pricing power because they can no longer bundle (in particular, putting one good song per album and charging $13 is a thing of the past). The RIAA (implicitly) dismisses this line of reasoning in the associated notes:

If digital singles are converted into an album equivalent (divided by ten) and added to both CDs and digital albums, the overall album unit decline in 2008 was 14 percent (635 million to 545 million).

See ... less overall units means people must be stealing music, because their demand for quality product has not decreased! Quick, pass some laws ...

However we have established that an album is not equivalent to ten good singles. Let's be generous and say an album is equivalent to 3 good singles. In this case the picture is rosier:

Viewed this way demand for music is about the same, with the difference being attributable to the economic climate. Furthermore, if two singles are considered equivalent to one good album, then demand has actually gone up.

Finally, it's amusing to note that a greatest hits album from the early 1970s is the 2nd highest selling album of all time, selling 48 million copies and therefore roughly valued at $500 million dollars. Clearly the media consumer had less capabilities in the past if they were willing to collectively pay $500 million for someone to assemble a set of previously released tracks into a single physical format.

More on the California Budget FAIL

2009-05-29T19:58:00.001-07:00

Armed with a Bud Light, I found two papers which suggest that from a tax burden standpoint, California is on the high side but not insanely so. The Public Policy Institute of California notes the following per-capita combined 2005-2006 state and local tax burdens for several large states:

New York: $6413
California: $4517
Illinois: $4081
National Average: $4001
Florida: $3693
Texas: $3235

From both a tax burden and overall revenue burden standpoint, California was 10th highest in the nation. If adjusted for incomes (which are higher in California), the ranking drops to 18th. That doesn't sound that bad, although the LA times quoted a more recent study saying our overall burden was now 6th, behind New Jersey, New York, Connecticut, Maryland and Hawaii. (That LA times blog is a bit misleading as it focuses on top marginal tax rates and only mentions overall tax burden at the end of the article).

The Center for Continuing Study of the California Economy has an interesting 2007 paper about California state taxation that echoes some of the above sentiments plus provides detail about the highly progressive (and thus volatile!) nature of California taxation. It also has this cool graph showing that prior to the passage of proposition 13 California was a "high tax" state, but proposition 13 brought us in line with other states.
This suggests to me that our budget shortfall is the result of excess spending, not insufficient taxation. There is perhaps some room to increase the tax burden but real progress will only come from spending cuts. The progressive nature of our taxation does increase the sensitivity to economic downturn but one could argue this is an automatic stabilizer and should be retained.

It's always scary to agree with an organization with an Orwellian name like the Reason Foundation, but they come to the same conclusion:

"A good rule of thumb in government budgeting is that the rate of spending increases should not exceed the rate of population growth, plus inflation ... Over the entire 18-year period, state spending grew at an average annual rate of 5.91 percent, while population plus inflation grew only 4.38 percent a year, on average ... Gov. Schwarzenegger and state lawmakers won't fix the underlying budget problems until they admit the state has a spending addiction."

But addicted to spending on what, exactly? An investigation for another day.

A Brief Look at California's Budget FAIL

2009-05-29T17:44:00.001-07:00

In a small effort to be a good citizen and understand wtf is going on with my state, I decided to make a picture of recent California expenditures and revenues. I took raw revenue and expenditure data and adjusted it for inflation and population. The result is per-capita expenditure (red) and revenue (green) in constant 2000 dollars.It looks like the budget was out of control in the pre-Schwarzeneggerian era (he assumed office Nov 2003), improved somewhat in 2004 and 2005, and then drunken sailor-esque spending kicked in while revenues trailed off.

A much more ambitious task would be to figure out on what the money is being spent ... hmmm ... I think I'll have a Bud Light instead.

There will be Confusion

2009-05-29T13:58:00.001-07:00

The price of oil, having reached over $66 per barrel today, has the financial intelligentsia confused, and some are downright angry. This month has seen both prices and inventories rise in the face of declining demand which is counterintuitive even to OPEC. As is fashionable these days there is speculation that oil is being used by the Chinese as a store of value. The more reasonable explanation is simple contango, i.e., oil companies are being paid alot of money to store oil because the futures market is anticipating a supply crunch.

Unfortunately, higher oil prices impede economic recovery, but the silver lining is they provide the economic incentive to change our energy strategy.

Profiting from Insanity

2009-05-28T19:52:00.000-07:00

What goes up, might come down if you wait long enough.
And I actually traded this one, starting Wednesday when the tide appeared to be turning. (I'm out now).

I have my eye on the Gap which is a similar story, namely up on a earnings report which was less than stellar. However it's still defying gravity and I don't see signs of lack of support yet.
This is just a gut feeling from watching the real-time quotes stream across my desktop while I work, but I suspect some big money is accumulating this stock.

Size Matters

2009-05-28T10:37:00.000-07:00

My startup is focused on internet advertising, and one thing we've commented on around the office is how computer monitors used to be much smaller. So what? Well, that means internet advertising used to be a much larger portion of the screen, because of IAB standard sizes were developed in the mid 1990s.

So I decided to investigate this hypothesis by cobbling together what I could find on screen resolution and banner advertising clickthrough rates. I fit the data to a 2 parameter model where one parameter was the clickthrough rate for screen resolutions higher than 1024x768 ("high resolution"), and the other parameter was the clickthrough rate for screen resolutions at or below 800x600 ("low resolution"). Note the resolution 1024x768 does not have it's own parameter, on purpose: there are only 5 data points, so I wanted to limit the model.The parameters end up being 0.2% for high resolution screens and 0.8% for low resolution screens.

Could a factor of 4 in clickthrough rate be recovered by switching to much larger ad units?

Natural Gas Vehicles FTW

2009-05-27T15:17:00.000-07:00

I guessed recently that speculation in natural gas prices was unfounded since we are literally swimming in the stuff, partially due to improvements in fracing. I was right, but it doesn't count, because I didn't believe it enough to buy puts.
The more important point is that this is good for America, because oil prices have been steadily increasing even though we are still in the midst of a deep recession. Natural gas is an important part of how America avoids a future Kurt Vonnegut style dystopia. So I was pleasantly surprised to discover the Civic GX is now being produced in North America. I was doubly surprised to discover just how many natural gas filling stations there are in my area, including one just 3 miles from my house:
Considering there are only circa 150 natural gas stations nationwide, this is pretty good; hopefully with the recent budget problems these will stay open.

You can also fill up at home using a device called Phill.

The Credit Crunch is Over, Sort of

2009-05-23T10:56:00.000-07:00

The TED spread is a key barometer of credit market risk aversion, and it has fallen to levels last seen before the credit crunch began. In other news, treasury yields are approaching levels last seen before the credit crunch (despite or because of quantitative easing?), and the dollar has lost it's "safe haven" credit crunch aura. These developments are positive as they indicate a dissipation of the massive fear in world markets over the last 2 quarters. However, as that fear dissipates, investors will start to view the US without their rose-colored "flight to safety" goggles on. There are still plenty of reasons to be pessimistic about the US, including tremendous loss of US household wealth (including my own!), resulting consumer deleveraging, and increasing federal budget deficits. (International markets have better prospects over the short term: for example, compare the US unemployment outlook with the Brazilian unemployment outlook.)

In the face of both good and bad news, the US market is doing some very interesting things. My favorite story this week is Saks Fifth Avenue, which experienced a huge bump in stock prices after losing less money than analysts expected (seriously?), only to give most of it back by the end of the week.
Another example of temporary insanity is Deere & Co., which announced falling revenues and slashed forward guidance before the open on Wednesday, skyrocketed, and then fell back to Earth.
One analyst said it "defies logic that this company is trading higher on the news of further lowered guidance". Indeed.

Yet there is no shortage of craziness, especially in the retail sector. Barnes and Noble on Thursday had a horrible earnings report, but less horrible than what analysts expected. This is a company that even without the credit crunch is destined to be crushed under Amazon's boots, yet here is the market response:
I like to stay out of the market's way until I see signs that craziness is abating: I made modest but not maximal amounts of money on Saks and Deere this week because I waited until the euphoria had faded to buy puts. So I shall be watching Barnes and Noble for signs of buyer's fatigue.

Gold-silver ratio developments

2009-05-17T14:14:00.000-07:00

There has been chatter in the blogosphere about how the gold-silver ratio is historically anolomous. A picture is worth a thousand words, so here are GLD and SLV relative movements over the past 2 years:Pre credit-crunch, gold and silver were moving in tandem, but they have disconnected since then. What's going on? Everybody is asking that question, and probably those who actually know aren't talking. Although silver has industrial uses (unlike gold, which is primarily a value store), silver is mainly produced as a by-product of other mining activities like zinc, so falling economic activity leads to lower silver production: silver production dropped 5% in 2008 and another 6% decline is forecast for 2009.

A geopolitical point of interest is that China is accumulating gold; this is not a recent phenomenon (accumulation has been over the past 6 years), but people are sensitive to the dollar losing its status as a reserve currency. But apparently they've been buying all sorts of industrial metals and actually as a percentage foreign-exchange holdings their gold position has declined since 2003.

I'm noodling whether to invest in the gold-silver ratio declining; this would mean going short on gold and long on silver. However I need to get more information before committing actual money.

Would You Mail Fries with that?

2009-05-15T22:07:00.001-07:00

The recent hike in US postage fees has led to lighthearted discussion of the suitability of Forever Stamps for speculation. The relative difficulty of liquidating vast amounts of Forever Stamps, combined with the legal requirement that the Post Office not raise rates faster than inflation, suggests no.

However it's still fun to ask whether or not this rate hike is justified; is postage historically cheap or expensive? Wikipedia claims that postage prices relative to the consumer price index have changed very little in past three decades. However the consumer price index fails to take things into account that feel like inflation (or deflation) to me personally like the purchase price of a home (it uses rents instead, because "owner-occupied housing combines both consumption and investment elements, and the CPI is designed to exclude investment items" ... gee I wonder how asset bubbles ever form).

Clearly then we must find some other barometer to judge the cost of postage, and in the spirit of the Big Mac Index I nominate the Big Mac. Note this is actually a bit stupid because the Big Mac Index is designed to assess the same item being sold in different places at the same time; over time its price relative to other stuff should move around, e.g., due to beef prices. However the temptation was just too delicious to pass up. Armed with historical data gleaned from the archives (hey, the print subscription finally had digital utility!), I assembled the price history of a Big Mac over the last decade. Without further ado, here is the number of stamps it would take to buy a Big Mac varying over the past 10 years.As you can see, postage is currently relatively cheap even after the recent price hike, since it takes circa 8 stamps to buy a Big Mac.

Hedge Your Home

2009-05-12T13:52:00.000-07:00

An pair of ETPs designed to track the S&P/Case-Shiller Composite-10 Home Price Index will soon be available. In theory, these could be used to hedge against a drop in your home, speculate about a rise in home prices without actually purchasing a home, or generate extra income from your house akin to a covered call. In practice, the pair have a teeter-totter structure which is unintuitive (but necessary to track an arbitrary quantity like Case-Shiller), and the prices interact with the closing date of the underlying securities in August 2014 which make them more like a futures contract.

The ETPs fill a need because the obvious choice, real estate ETFs, are poorly correlated with Case-Shiller. However this is a definite wait-and-see, because if they really end up priced like a futures contract it might not prove useful for shorter term hedging.

America as a Price Taker

2009-05-08T21:44:00.000-07:00

Oil is creeping back up again in price; it recently crossed $58. This rally seems anticipatory rather than demand driven; US crude oil stockpiles are at 19 year highs and China moderated their oil consumption in Q1. With only partial information, I'm waiting for USO to pull back to circa $28; then I'll buy leaps. If domestic or developing world demand is ramping due to data I as humble blogger cannot see, then I'm waiting for Godot, and I'll miss the party. C'est la vie.

Long term, of course, oil will go up. As I alluded to previously, commodities prices will not necessarily moderate in the face of American economic weakness; in other words, as developing nations increase in economic power America will increasingly be a price taker, not a price maker.

Natural gas is different in ways that are good for America but bad for speculation. Natural gas prices are radically below last years levels, and there are good reasons for this. First, the US is swimming in natural gas, thanks to a new extraction technology recently developed. Second, the infrastructure to import or export natural gas has been neglected; this adds up to the US being a price maker and not a price taker. However, natural gas has a lower carbon footprint than coal (for electricity) or oil (for transportation) and it is a proven technology for both cases (although, a large scale switch to natural gas for transportation is harder than for electricity). Clearly long term the United States is going to greatly increase its natural gas consumption. A movement into natural gas for transportation would be great for national security, the environment, and our balance of payments; but it is not imminent, and demand is forecast to fall this year. So I'm waiting to invest.

Yield Going Up

2009-05-07T21:01:00.000-07:00

Every American should be keeping a close eye on treasury bond yields, since the willingness of foreigners to finance our profligacy is key to our survival. Today at a record-sized auction of treasuries yields went up, meaning buyers were demanding better prices.

Generally yields are back to their November 2008 levels, and this is with the fed printing money to purchase them. Looks like the flight to safety is officially over, time to invest in dollar hedges again. I like the Australian dollar: I like to think it's because of their high yield, exposure to the commodity rebound, and integration with the Chinese economy; but in fact I'm still enamored with Crocodile Dundee.

More Fertility Fun

2009-05-06T20:58:00.000-07:00

Yesterday I investigated a putative relationship between housing and fertility. I suspected that unemployment was probably a more reliable predictor. Well, unemployment data is pretty easy to come by so I poked around a bit. A one-year lag between unemployment and total fertility rate for recent data yields a visually pleasing fit.Unemployment for Q1 2009 is currently estimated at 8.1% with the March estimate at 8.5%. So this model is more pessimistic than the housing model, and predicts a 2010 total fertility rate of 2.02 (vs. 2.04 for the new housing sales based predictor); or possibly worse if unemployment continues to degrade over the year.

So the models are predicting between 0.06 and 0.08 drop in total fertility rate in a 2 year period (2008 was 2.1). This kind of movement is not unprecedented; there was a 0.05 drop from 1991-1993 and again from 2000-2002. It would be an usually large drop, but this is an usually harsh recession, so it does not seem unreasonable.

One interesting question to ponder are the possible mechanisms for lower fertility during recessions: less sex happening, higher use of birth control, or more abortions?

Housing and Fertility

2009-05-05T20:30:00.001-07:00

I mused that the credit crunch was killing the housing market, which in turn might be killing fertility; after all, does losing one's home really set the mood for love? First I took a look at new home sales, courtesy of the US census bureau. I normalized the housing sales by population to get something akin to a rate.
Wow, look at that housing market crash! Next I took a look at recent total fertility rate, cobbled together from two sources. 2.1 is the magic replacement fertility level; apparently, we're experiencing a baby boomlet. However, in past recessions fertility rates dropped, so the boomlet might not last.
It's possible to regress the housing sales against the fertility rate and get a good p value if a two-year lag is used. It predicts a total fertility rate of 2.04 for 2010. Will this prediction hold up? My gut tells me the decision to have children is probably better related to unemployment numbers, and housing has been so bonkers this decade that using it to predict anything is questionable. However one of the fun things about blogging is the time capsule aspect, so I'll check back in 2011.

Another oddity I dug up while looking into this: multiple births have been steadily rising in this country for some time.

Arbing the Global Savings Glut

2009-05-03T18:05:00.000-07:00

I blogged recently about how foreign capital continues to invest at elevated levels in US federal debt, even as private debt instruments fall out of favor. I mused that one thing Obama could do to arb this would be to issue credit cards to Americans.

So I was very interested when I heard about Build America Bonds. These are municipal bonds, but unlike typical municipal bonds they are not US federal tax-free. Instead, the federal government subsidizes 35% of the interest rate payments, i.e., the issuer of the bond offers a higher yield to compensate for paying taxes. This might seem like a tortured dance until you realize foreign investors are not subject to US federal tax on their bond income. Therefore, up to now, they have avoided tax-free municipal bonds with higher "effective pre-tax yields". Clearly Build America Bonds were designed to attract their attention, and sure enough a recent auction by California was gangbusters. (Ok, a less cynical possible design rationale is that domestic investors from the lower marginal tax rate brackets will find these bonds more attractive than tax-free bonds; somehow I don't think that was the aha moment behind these.)

Given that the domestic personal savings rate is trending higher, and marginal tax rates are almost certainly going up in the next few years, classical tax-free municipal bonds will have more domestic money chasing them with a higher effective pre-tax yield. In addition, the Build America Bonds provide an alternative market competing for yield which attracts foreign interest. It's a good time to be a local issuer of debt: Obama has handed you a federally subsidized credit card!

Diseases, Labor Supply, and Income Equality

2009-05-01T22:43:00.000-07:00

All the recent discussion of a global pandemic got me thinking about the Black Plague. The Black Plague was actually really good for labor; it is commonly believed that by killing 2/3 of the population in Europe, the Black Plague radically altered the supply of labor and forced feudal lords to end peasantry.

What about the converse? Lately, there has no real threat from disease or war and the population has been growing nicely. That made me wonder if perhaps the supply of labor has been increasing and therefore the balance of power has been shifting to capital.

Certainly the US has developed an increasingly inequal income distribution over the past 30 years. The US census bureau has historical data on the Gini coefficient you can download and play with, which will yield a graph like this.Lower numbers are "more equal", so we've been getting worse. Reaganomics at work! Since it has been growing over the past 30 years we could regress against anything that has been growing during this time and get what looks like a meaningful correlation, including regressing against either population or working age population. So that seems like cheating. Therefore I tried not to cheat by looking at the current distribution of the Gini coefficient across the world courtesy of the CIA world factbook. However I couldn't find anything by regressing this against measures of "labor supply". For instance, here's a scatter plot of Gini coefficient vs. working population fraction.
So back to the US historical data. I decided to feel ok about regressing it against the balance of payments normalized by GDP. This is only partially a measure of labor supply in that importing alot means people from other countries are doing work for you.Don't take this as proof of causality; it's too easy to find all sorts of stuff that essentially monotonically increase over the past 30 years. However, if you already think that globalization has shifted the balance of power from labor to capital by radically increasing the labor supply (which I do), then you can take this as an estimate of the size of the effect. That's not a meaningless result, because we can say that eventually the US will have to have a net zero balance of payments; the globe will not labor for us indefinitely. Thus, once we are in balance, we can predict what the Gini coefficient of the US will be: about 0.41. This is substantially lower than the current Gini coefficient (circa 0.47), comparable to Russia's current Gini coefficient and the US coefficient in 1982.

Ross Perot's giant sucking sound notwithstanding, to get back to the Great Society 1968 levels of equality (Gini coefficient 0.38) we'll need to change more than our trade policies; we'll have to change the tax code and social net as well.

Ecommerce and Unemployment

2009-04-27T11:34:00.000-07:00

Nicholle and I have two incomes, two kids, and no time. Not surprisingly we heavily utilize ecommerce sites: Old Navy for clothes, Zappos for shoes, and Amazon for everything (Amazon Prime rocks). We sometimes buy our groceries online, but not that often since delivery windows suck.

Having heard alot about rising unemployment, I mused that ecommerce sites are presumably more efficient with human resources than traditional retail. In other words, we might be exacerbating unemployment via ecommerce.

It turns out Amazon has about 21,000 employees and generates about $19B in revenue. Contrast with Wal Mart, with has about 2,100,000 employees and generates about $400B in revenue. Clearly, Amazon is more efficient per employee. There are probably lots of reasons for this (e.g., different inventories), but blindly extrapolating Amazon could scale to handle Wal Mart's business using only 20% of the employees. That's about 1.5 million people out of work.

Another classic example of automation replacing many low paying less interesting jobs (retail service) with a smaller number of higher paying more interesting jobs (retail site programming), as predicted by Kurt Vonnegut.

The Credit Crunch Hits Home (Ownership)

2009-04-22T12:56:00.000-07:00

Home ownership is the new whipping boy, not surprising given its central role in the credit crunch. The economist has run several articles questioning the value of home ownership to society, and other opinion outlets are chiming in. The lynchpin of the argument is labor mobility; home ownership decreases it and reduces economic efficiency. This leads to the recommendation to reduce incentives for home ownership in order to better approximate the socially optimal level of home ownership, e.g., reduce or eliminate the home mortgage deduction.

Interestingly, technological developments have been favoring labor mobility. Job search engines make it easier to find the jobs. Social networking lowers the personal cost of moving by making it easy to stay in touch with the people from whom you move away and to get in touch with the people to whom you move toward. Community recommendation sites reduce the cognitive load in becoming an expert in a new area. Mapping sites and GPS technologies mean getting lost is a thing of the past. Rental listing sites makes it easy to find a place to live.

Given these technological advancements, and a weaker domestic labor market due to globalization, I expect that the new generation will be far more mobile than my generation. I do not think telecommuting is a mitigating factor; after all, if you can do the job from Kansas, you can do it from Bangalore, thus physical presence will command a premium. I think the real danger going forward is not oversupply of home ownership, but undersupply. This is doubly true given young people will have the experience of a housing market crash increasing their level of risk-aversion.

Therefore, reducing incentives for home ownership is an improper policy response.

Sailor Capone

2009-04-20T13:18:00.000-07:00

Columbian drug-dealers now have their own submarine navies for the purposes of drug smuggling. Adm. James Stravidis, Miami-based U.S. Southern Command, was quoted:

“In simple terms, if drug cartels can ship up to 10 tons of cocaine in a semi-submersible,” Stravidis wrote, “they can clearly ship or ‘rent space’ to a terrorist organization for a weapon of mass destruction.”

Sounds dangerous. Ask yourself, who created the economic incentive for this innovation?

Prohibition 2.0 makes as much sense as pervasive indiscriminatory use of antibiotics; we are literally evolving super-criminals. Today they have their own navies, tomorrow will they have their own space program?

We need to decriminalize it ASAP instead of financing terrorist R&D.

Full Faith and Credit

2009-04-18T11:39:00.000-07:00

As an entrepreneur I like to know which way the wind is blowing. Having heard alot lately about the era of frugality, I thought I'd look into it myself. Fortunately, all the data you can eat about our economy is merely a mouse-click away, which is a good thing given that journalism is dying.

A starting point is the personal savings rate data, courtesy of the commerce department. The trend from the mid-80s right up to the credit crunch was steadily decreasing personal savings rate, which actually dipped negative briefly. Recently it has shot up, although it's a bit difficult to see on this graph. The last few monthly data points are 0.8, 1.4, 2.6, 3.0, 3.8, 4.4, and 4.2. That basically puts us back to where we were in the mid-1990s and well below the savings rate of the 1980s (who knew the decade of greed was so virtuous?). In absolute terms it's not that shocking, although the rate of change has caused alarm.

Desperate to avoid a paradox of thrift, the federal government has stepped in and started spending like a drunken sailor. The US government can really hold its liquor, actually, since with the US debt data from Wikipedia it is easy to see that this is business as usual, except for a brief period of fiscal responsibility in the second half of the Clinton administration. (I've applied the GDP deflator to the data to attempt to normalize the dollar to the year 2000).

So Americans save more individually and the government borrows more collectively; that's the Keynsian playbook, at least for the last 6 months. Prior to the last 6 months, however, we were borrowing heavily at the Federal level and not saving personally, so naturally interest rates increased due to the lack of supply of savings, right? Wrong. Witness the yield on the 30 year treasury bill during this period (that gap in the graph is from when the 30 year was temporarily discontinued).

As Americans saved less and less, the cost of borrowing for the government plummeted (or perhaps the converse?). Here's the treasury yield parametrically plotted against the personal savings rate (the quadratic fit is just phenomenological eye-candy). I colored the most recent data point red just to emphasize this trend has been broken.

Basically foreign liquidity rushed in to fill the gap; while oil-exporting countries would be a reasonable guess as to the source of funds, Asian countries are a larger factor, specifically Japan and China. This global imbalance is at the heart of the financial crisis, leading to basic questions such as: why are developing nations saving so much, and why are they investing their savings abroad?

In any event the future, just like with oil, it is less about the United States and more about the developing world. Arguably, the past two decades were the era of frugality, in a global sense. Previously individual Americans could tap into the global flood of liquidity via instruments such as sub-prime mortgage backed securities, but investors are no longer biting. Currently, however, they are willing to loan to the country collectively (i.e., the federal government). In Q4 2008, investors were literally paying the federal government to hold their money for them. Under these conditions it makes perverse sense for the federal government to borrow heavily. For instance, scaling each year's federal deficit from above by the treasury bill rates over that interval yields something I call the "interest load" (not a standard term).

This indicates that interest rates are so low that even with record budget deficits we will have historically reasonable interest rate payments. Obviously, that could change fast if investors no longer find our federal government a good investment.

If the global era of frugality continues, however, all of us together can borrow much more cheaply than any of us can individually: how best to exploit that? We could arbitrage it directly and have the federal government issue credit cards or offer debt consolidation services (the Obama housing plan bears some resemblance to the latter). Humor aside, retail therapy has a low multiplier value, so the government is going to focus on social services and infrastructure spending.

I implicitly promised a weather forecast at the beginning of this post, so here it is. The global era of frugality will persist in the near term due to inertia, delegating the problem of spending to the US federal government; thus sectors related to public-sector spending will be hot. Americans were defying gravity in the mid-2000s so the personal savings rate cannot go down; but given that standard investment vehicles will be giving tepid returns Americans will not return to mid-1980s savings levels. In this environment the recovery will be soft, the labor market less strong, and taxes higher; therefore I like ideas related to minipreneurship. Etsy-like platforms offer a way to supplement the income of a formerly dual-income (or comfortably retired) household in a work-life balance compatible way for the seller, and offer a low-cost way to achieve a unique retail therapy experience for the buyer. Unfortunately all my ideas for Etsy-style platforms are frustrated by government regulation, but that's a topic for another post.

The Underbelly of the Internet

2009-04-17T09:51:00.000-07:00

My startup makes purchase-cycle prediction models (among other things), meaning that we figure out what people are doing on the internet days, weeks, and months before a purchase; and then we invert that to target advertising.

As part of the testing of the technology, we bought remnant run-of-network inventory on a major ad network and used our model to present Ebay auctions in banner ad space. While the experiment was a success, I am permanently scarred by the experience, since I spent alot of time looking at the sites that we were placed on, and it was not a pretty picture.

The issue was not the content. There were some sites that were either illegal or low value, e.g., video sharing (warezing) sites and automatically generated appropriated content republishers, but these were not typical. Indeed, there were alot of interesting fan sites (e.g., bon jovi, detroit red wings), in addition to real innovation (I discovered failblog.org this way, and they are now a regular source of personal mirth).

The ads were the issue. First, most of these pages' layouts are hellish because they are trying to maximize monetization by having multiple ad units. Second, the ads being shown were at best uninteresting and at worse facilitating fraud. The four ads shown here occurred constantly. We call ads like this "belly fat" around the office, in honor of the one with a stomach pic. You should be thankful that I'm not showing the animated version of the belly fat ad, cycling between the before and after picture.

None of the ad networks are immune to belly fat. Google Adsense might seem like a paragon of virtue but they were the source of the cheddah gets cheddah ads, which are the equivalent of those late night commercials with the guy wearing a suit that has question marks all over it. When Millard calls for an end to the tyranny of the click, I hear "end of belly fat".

If someone has built an interesting website serving an audience of, e.g., Bon Jovi fans, then why are they showing belly fat ads? One problem is that they are too small to attract the attention of advertisers who are interested in that audience; technologies like OpenX market can assist by aggregating these tail publishers and reducing friction. However the problem is not just size, otherwise large publishers like Martha Stewart wouldn't be complaining about the lack of creative innovation in web advertising.

The publisher bears some blame, because they could make the decision to show nothing but rarely do (in defense of publishers, ad networks keep publishers in the dark preventing them from making decisions like this). However I believe the fundamental difficulty is the barge pole problem: quality advertisers cannot embrace the tail of the internet because the landscape is so variable that they risk being potentially associated with content that is unprofessional, offensive, or even illegal. Only belly fat is brave (shameless) enough to wade in, and that makes the situation worse, because who wants to be shown on a site that also shows unethical near-fraudulent ads?

Interestingly, even cable television suffers from the barge pole problem, and cable channels are light-years more elegant than the long-tail of the internet.

Green cars not just for hippies anymore

2009-04-16T16:16:00.000-07:00

The price of oil has ranged from $40 to $150 per barrel over the past 18 months. According to the IEA, world oil demand ranged from 82.5 to 87 million barrels per day during this interval. That's a 5% change in demand and a 300% change in price.

High oil prices have spurred alot of discussion, reviving peak oil and it's new cousin plateau oil. I think the more interesting and less controversial observation is significant economic growth cannot happen unless either the supply of oil is increased, or demand of oil per unit GDP is reduced. Otherwise, when the world economy recovers, oil demand will increase by 5%, causing a 300% increase in prices, causing the world economy to stall. (I highly recommend watching Shai Agassi's electric vehicle proposal although I am skeptical about his particular solution).

The last time we had an oil shock was the 1970s, and since that time we have decreased the use of oil for generating electricity, which was one factor in improving our GDP per barrel of oil (we did this by cheating a bit, since we replaced oil with natural gas and coal, mostly). Since then we've been getting increasingly efficient in terms of GDP per barrel, despite the popularity of SUVs. However we've had substantial economic growth so we actually returned to our 1970s peak of usage near the turn of the century and have since eclipsed it. In addition the rest of the world has increased their use of oil, in particular China, which has gone from 2 to 7 billion barrels per day during this period.

About 45% of our oil use is for personal transportation, so we can make substantial progress by improving our automobiles. We will not experience another Great Moderation until we do so.

School Choice

2009-04-14T21:05:00.000-07:00

I'm pro-choice ... pro school choice, that is.

I'm passionate about school choice so I get into many heated debates about it. Here are some common objections I encounter along with my responses.

It's just a way to subsidize the rich leaving the public school system. Certainly a poorly structured voucher system would have this property, so let's make vouchers large enough to matter (e.g, $20,000 per year), and mandate that acceptance of the voucher constitutes complete payment (i.e., no additional tuition allowed).

It would lead to balkinization. Common fears include promoting racial segregation or the teaching of creationism. I would prevent this by holding voucher receiving institutions to the same curriculum and admissions requirements as public schools.

People are unable to make this decision. Sometimes I hear claims that the working poor are too busy for the cognitive load this decision demands. No one is brazen enough to admit to me that they think people are just plain too stupid to make this decision, although I suspect they think so. Even if you grant that, a reasonable automatic default works well for 401K plans and could work here as well (i.e., default to the geographically closest public school, since public schools can still take vouchers). I also suspect that many low income parents would find the time to research their children's education.

Special needs children would be neglected. So let's let the voucher system reflect the increased costs of educating special needs children. For instance, one certified special needs, the size of the voucher could double to $40,000 per year.

I'm throwing around some big numbers for the vouchers, but I'd be willing to pay vastly more in taxes for education if the tax increases were combined with significant reform. Fundamentally I think education is an excellent investment for society, not just in pure economic terms but also for the health of our democracy. However my wife is a former elementary school teacher (now a practicing attorney), and the inside look at the public school system I got during her tenure convinced me that the current system is unable to effectively absorb a radical increase in funding without systemic overhaul.

Hot Lattes

2009-04-13T20:27:00.000-07:00

Coffee sex shops reminds me of Idiocracy.

Notwithstanding the above, people actually appear to getting more intelligent over time, a phenomenon called the Flynn Effect.

Advertising is dead; long live advertising

2009-04-13T15:51:00.000-07:00

I've been reading alot about the future of advertising lately, mainly because I'm thinking about the future of my startup.

On one end of the spectrum, some are saying that advertising will cease to exist. Of course, now we need to define advertising, because no one is predicting an end to commerce or marketing. "Advertising" in this context refers to unsolicited awareness-based marketing. The argument is that the same advances in communication technology that make it easy for anybody to blog to everybody imply that vendors can easily reach target consumers. In addition, since unsolicited interstitial advertising no longer works, tools and process will arise that make it easy for customers to find vendors that meet their needs (the beginnings of the trend visible in Google, Yelp, Amazon, etc.). In this glorious future (present?), media companies whose business model is based upon audience aggregation and collocation of content and advertising cannot survive. Heads up: when ads fail, you're going to have to start paying for professionally produced stuff. In some domains, perhaps, amateur stuff will be so good we won't notice or care (aside: Kutiman is a professional producer), but it does seem some genuine problems will be created.

On the other hand are those who say any talk of the end of advertising is at best too premature to be useful. The real change underway as audiences move online is towards measurability. Direct response advertising is emininently measurable so much of the early advertising activity on the internet has (obsessively?) focused on direct response, but this view ignores other value that advertising can generate. If you are an artistic type, you want to break out of the direct response straightjacket and have another creative revolution. If you are a math type, you either work with a direct response vehicle that facilitates brand awareness, or you want to find a way to make branding measurable. The latter can literally mean mindreading, data mining techniques, or fancy focus group techniques.

I certainly believe that the industry in trending towards measurability; it's why I'm in this business. However I suspect that historically advertisers have overpayed for marketing, and are now entering a period where they will underpay: a multi-decade hype cycle, if you will. In Seth Godin's terms, we are switching from a period of "glamorous" ad buying to "rational" ad buying. Ultimately, advertisers will converge to the proper level of spending, but from below, and to entice them towards optimal spend the industry will have to find a way to relate all marketing activity to ROI. It is only when measurability has been conquered that the next creative revolution can truly start, because only then can the value of a great advertisement be understood.

The Best of Both Worlds

2009-04-11T16:24:00.001-07:00

Segway, not content to be bigger than the PC, is now teaming up with GM to save the American auto industry.

Which reminds me ...

GM's North American operations have to get much smaller to be viable. From a macroeconomic perspective, it would be a good thing for that happen quickly and without incident. The problem is that many real people work at GM and its suppliers. Compassion aside, asking these people to "suck it up" is not efficient, since people will resist changes that are obviously not in their best interests through the political process.

Americans accepted the risks associated with an efficient labor market during times of prosperity, but as unemployment rises we can expect a backlash. France is the prototype of runaway labor populism, and the resulting high unemployment has explosive consequences. Can we avoid this fate?

Enter Flexicurity . Pioneered by the Nordic countries, it is a combination of a flexible labor market (like America), a generous social safety net (not like America), and responsibilities for the unemployed to continue receiving assistance (even more stringent than America).

I like the idea of getting workers out of obsolete jobs as fast as possible; otherwise, it's a waste of human capital. By providing a generous social safety net, we can improve the alignment of incentives between citizen and society. While I expect there will be fraud and abuse, ultimately, the inefficiencies from defrauding generous welfare have to be balanced against the inefficiencies that result from other policy choices, such as handcuffing businesses in the labor market.

Since we cannot afford our current entitlements, I'm willing to pay more taxes for policies based upon flexicurity. This is not despite the fact that the labor market for my particular skills is strong and should stay so during my working lifetime, but because of it. I'm going to pay for the curity whether or not I directly need it: protectionism, subsidies, and distorting incentives will get me if taxes don't. I'd rather the goal of sustainable labor market efficiency be tackled directly, so that I can benefit from the flex via higher wages, meritocracy over seniority, and increased job choices.

This line of thinking forces me to be in favor of some kind of health care reform. Currently I prefer a Mccain style plan to adjust the tax code, rather than what Obama has been proposing lately.

An Economic Theory of Dirty Dishes

2009-04-11T14:22:00.000-07:00

The dishes really piled up last week. I attribute it to an era of frugality inspired batch of late night bulk cooking, which generated a bunch of large dirty pots which we would clean in the morning. It never happened, and it wasn't until this weekend that we finally got a handle on it. I was reminded of my theory of dirty dishes however.

The theory basically says that the marginal cost of cleaning a single dish is essentially constant, but that the marginal utility of cleaning a single dish decays as a function of the total number of dirty dishes. (This is because, when are there are many dirty dishes, there is little sense of progress on each one.) Therefore, if a spike in the number of dirty dishes occurs, the equilibrium supply of dirty dishes becomes trapped away from zero. The only way out is to wait for the marginal cost of cleaning a dish to fall, i.e., for the weekend to show up so that you are feeling less lazy.

But hey, a picture is worth a thousand words right? Let's assume you have 10 dishes, and a sink that holds 6 dishes, and can hold 2 without the dishes being visible from afar. Then a model of the utility of a particular state of the kitchen is given by the silly picture. (Not captured: once all dishes are dirty, there is some utility to have a clean dish when you need to eat something.) This implicitly represents the demand for clean dishes: points where the slope of this curve are high correspond to states of the kitchen where you would pay the most to have a dish cleaned.

Of course by paying I mean actually cleaning something. I model the cost of cleaning one dish as roughly independent of the number of dirty dishes; instead it varies over time based upon psychological factors such as exhaustion and alternate uses of time at the current moment. There are economies of scale, however: the cost of cleaning subsequent dishes falls rapidly since one is "in the zone". For simplicity, I'll model the cost of cleaning subsequent dishes as zero. Therefore, the decision to do all the dishes hinges upon whether the marginal utility of doing one dish is sufficiently great. Since our unit of utility is arbitrary, I'll just say the units are "cleaning 1 dish". In this case, we only need the one silly graph I've already produced above, with the understanding that it will scale greatly for different people, or even for the same person at different times of the week.

So the theory admits several outcomes merely by scaling the above graph, including equilibrium at all dishes dirty (aka college), all dishes clean, and some dishes clean. Importantly there is a range where you normally manage to keep all the dishes clean, but if you experience a black swan in the number of dirty dishes you can get permanently behind. Since this dynamic happens for more than dirty dishes, this argues for having an infrequent housekeeper visit, e.g. biweekly, to prevent undesirable absorbing states.

It also makes one interesting testable prediction: you can get your spouse to clean more dishes by hiding most of the dirty dishes, and producing them only when they start to clean them. I'll be trying that one out on Nicholle.