Category Archives: Statistics

Some thoughts on data visualization and Oregon’s unemployment

Economists often have something interesting to say.  However, they often do a terrible job of making their points in a way that those outside of economics can understand.

For some time, I’ve developed an admiration for the approach preached by Edward Tufte. He is widely considered a pioneer in the field of data visualization—also know as making-tables-and-graphs.

I won’t go into Tufte’s approach here, but I thought I’d show how using his approach can radically improve the presentation of economic data.  In this example, I’m trying to show Oregon’s unemployment record relative to the rest of the U.S.

This first figure, below, shows unemployment for Oregon and the U.S. as a whole.  The chart was made in Excel by highlighting the three relevant columns (month, Oregon unemployment, and U.S. unemployment) and clicking the line chart button.

The result is a standard ugly Excel chart.  Even worse, it does not tell an especially interesting story about Oregon’s unemployment.  Oregon seems to have slightly higher unemployment than the U.S. as whole, but also seems to track U.S. unemployment fairly closely. Big whoop.

The problem with the figure above is that it does not say anything about unemployment in the other states individually.

The next set of figures use a much bigger dataset.  It reflects the unemployment rate for each state in each month from January 1976 through October 2012. That’s more than 22,500 observations.

The figure below is better, but not much better.  Oregon’s unemployment is the heavy black line.  The gray background represents the range between the state with the highest unemployment and the state with the lowest unemployment.

Notice that the graph removes much of what Tufte calls “chart junk”—a frame around the graph, ugly and unnecessary gridlines,  a legend, and the month of “Jan” in the x-axis.

While the graph above is much cleaner and better looking, it is fatally flawed.  With a few exceptions, Oregon looks to be in the middle of the range of unemployment rates across the states.

I know this is not true.

In fact in 213 out of 442 months (48 percent of the time), Oregon has been in the top 10 for high unemployment among the states.  The figure above does not adequately show this fact.

The figure below is a major improvement.  It shows the percent unemployment for each state represented by tiny gray dots, with Oregon represented by the red dots. (I would have preferred a red line for Oregon, but Excel layers dots over lines. Thus a red line would be “under” gray dots and looks amateurish.)

With the figure below, it is easier to see the point that I am trying to make: Oregon’s unemployment tends to be among the worst in the country for many points in time.  This can be seen pretty easily in the years 2001 through 2005.  But the point is not as well made in 1981 and 1981 when Oregon’s unemployment was the fourth or fifth highest in the U.S.

The figure below is another improvement.  The black line is Oregon’s unemployment (the black line is darker than the gray dots, so Excel’s layering problem goes away.)  The red dots represent months in which Oregon’s unemployment is in the top 10.

While the graph is much better and makes the point I was looking to make, something about it bothers me and I feel it is cluttered.

The fact that Excel does some weird thing that makes it look like there are white gridlines really bugs me.

Here we go. One last improvement.

In the figure below, I went back to the gray background representing the range between the state with the highest unemployment and the state with the lowest unemployment. The black line is Oregon’s unemployment and the red dots represent months in which Oregon’s unemployment is in the top 10.

The final figure seems to best make the point I was trying to make.  We see Oregon’s unemployment, we see the range of unemployment among the states, and we see how often (and when) Oregon’s unemployment was among the worst in the U.S.

New study: A right-to-work law in Oregon would give a big boost to employment and incomes

In Oregon, employers can have an agreement with unions that make union membership—and the payment of union dues—an employment requirement.  Refusal to stay in the union or to pay dues can result in termination.

Right-to-work laws provide job seekers the right to work for an employer whether or not they choose to join the union. Twenty-three states have right-to-work laws, with Indiana enacting its legislation yesterday. Research has found that as a group, right-to-work states have enjoyed more rapid employment growth, better job preservation, and faster recoveries from recession.

A recently released study from Cascade Policy Institute examines the impacts right-to-work legislation would have on Oregon. The study is consistent with the vast majority of peer-reviewed research in finding that if Oregon were a right-to-work state, we would see improved employment and income growth. For example, if Oregon enacted right-to-work legislation this year, in five years, the state would have 50,000 more people working than if it maintained the status quo. Similarly, in five years, Oregonians would have $2.7 billion more in wage and salary income by enacting right-to-work legislation.

The study is Fruits, E. and Pozdena, R. J. (2012). Right-to-Work is Right for Oregon: A Comprehensive Analysis of the Economics Benefits From Enacting a Right-to-Work Law. Cascade Policy Institute.

How many parameters does it take to make an elephant?

A post by John Cochrane on the pitfalls of over-differencing data brought up the broader topic of over-parameterization. One commenter mentioned a wisecrack attributed to John Von Neumann:

In desperation I asked Fermi whether he was not impressed by the agreement between our calculated numbers and his measured numbers. He replied, “How many arbitrary parameters did you use for your calculations?” I thought for a moment about our cut-off procedures and said, “Four.” He said, “I remember my friend Johnny von Neumann used to say, with four parameters I can fit an elephant, and with five I can make him wiggle his trunk.”

Sounds like a quip from Big Bang Theory, huh?

Turns out, one can draw an elephant using four parameters, and use a fifth to wiggle the trunk.

Mayer, Khairy, and Howard (2010). "Drawing an elephant with four complex parameters." Am. J. Phys. 78.

In related news, a paper published in the Proceedings of the National Academy of Sciences, estimates that it takes at least 24 million generations for a mouse-sized animal to evolve to the size of an elephant.

Oregon and Washington minimum wage increases in New Year may hurt more than they help

2 Replies

Eight states will begin the New Year with a higher minimum wage under state laws that mandate minimum wages increase with inflation. Washington State will become the first state in the nation to set its minimum wage above $9 an hour. Oregon’s minimum wage will follow closely behind with an increase to $8.80 an hour.

The minimum wage is a textbook example of a price floor resulting in too many workers chasing too few jobs, especially among those applicants with the fewest skills.  The result is higher wages for those who get a job, but no jobs for many who are seeking employment.

Statistical research on Oregon and Washington’s minimum wage increases finds that the states’ higher minimum wages have, on net, a negative impact on employment and wages.

Higher minimum wages in Oregon and Washington are associated with reduced employment: Oregon and Washington’s higher minimum wages are associated with a statistically significant reduced probability of being employed.

Younger members of the labor force are more likely to be adversely affected by increases in the minimum wage: Oregon and Washington indexing policies produce annual increases in the minimum wage that, in turn, are likely to increase unemployment, especially among the young. The table below shows the impacts of higher minimum wages on youth unemployment over time. As Oregon and Washington’s minimum wages increased over time relative to the federal minimum, the states’ youth unemployment increased relative to what it would have been otherwise.

Higher minimum wages have no statistically significant impact on wages of Oregon and Washington hourly wage earners: Some proponents of higher minimum wages argue that the increases have a “ripple effect” for employees just above them on the pay scale.  However, statistical analysis of Oregon and Washington does not find any “ripple effect.” Indeed, controlling for employment impacts, increasing minimum wages has no statistically or economically significant impact on incomes.

Bottom line: Minimum wage indexing seems to impose employment costs with no measurable income benefits.

For more information on data and statistics, please see Fruits, E. (2009). The Impact of Minimum Wage Indexing: Employment and Wage Evidence from Oregon and Washington. Employment Policies Institute.

Urban legends: Shaky statistics behind Portland’s claim of having the most strip clubs of any city

1 Reply

If you have ever visited Portland, it’s fairly certain that someone will proclaim that the city has more strip clubs per capita than any other city. Indeed, Portland’s reputation is international.  Earlier this year, the UK’s Guardian newspaper dropped the factoid in its review of Voodoo Donut.

More recently (in an article that doesn’t seem to have much of a point), The Economist magazine paints a scene in which progressives bike side-by-side with the prurient:

Peaceful, green, and liberal, Portland has a reputation for being unusually socially conscious. So visitors are sometimes surprised to learn that it is a plausible contender for the title of lewdest place in America. It has more strip clubs per head than any other city; in its compact downtown, sex shops are scattered amid the bookstores, coffee bars and social services.

Aside from that fact that there is hardly a scattering of sex shops and—truth be told—downtown is somewhat bereft of bookstores, The Economist article repeats Portland’s most famous statistic that it has more strip clubs per person than any other city.

Is it true?

Ask a Portlander if the statistic is true and he or she will say, “Sure it’s true, just ask anyone!”

It’s not a statistic that the Census Bureau collects. The Chamber of Commerce does not spend any energy counting its members who own strip clubs.

So where did the statistic come from?

The oldest article I could find online came from a 1995 Willamette Week article which suggests the newspaper conducted its own survey:

And what of that “most strip clubs” boast? Our Internet survey of Las Vegas, the gold standard of urban debauchery, reveals 30 clubs, which works out to 5.85 strip joints per 100,000 residents. San Francisco, that legendarily libidinous burg, is estimated by SF’s adult weekly, The Spectator, to have 17 strip clubs, or 2.2 per 100,000 residents. By comparison, Uncovered and Exotic list 41 strip clubs within the Portland city limits. With a whopping 7.74 clubs per 100,000 residents, Portland solidly trounces these two centers of vice in number of brass poles per citizen. In your face, San Francisco!

Bottom line: A local paper began with the assumption that only Las Vegas and San Francisco could possibly beat Portland in the metric of strip clubs per person.  The paper counted clubs, divided by population, and—voilà!—a legend was born.  The methodology is pretty shaky, so you should take a generous grain of salt next time you’re told Portland has the most strips per person of any city.  If this were PolitiFact, the claim would be rated “Barely True.”