Recently in Studies Category

Further Issues with Brain Imaging

| No Comments
Day after day it seems we are told that neurosciences will dramatically change in how we conceptualize human behavior and ultimately culpability.  As I have noted extensively in the past, there are many problems with this view.  The newest difficulty was published last year in the Proceedings of the National Academy of Sciences:

Functional MRI (fMRI) is 25 years old, yet surprisingly its most common statistical methods have not been validated using real data. Here, we used resting-state fMRI data from 499 healthy controls to conduct 3 million task group analyses. Using this null data with different experimental designs, we estimate the incidence of significant results. In theory, we should find 5% false positives (for a significance threshold of 5%), but instead we found that the most common software packages for fMRI analysis (SPM, FSL, AFNI) can result in false-positive rates of up to 70%. These results question the validity of a number of fMRI studies and may have a large impact on the interpretation of weakly significant neuroimaging results.

Hat tip to Prof. David Bernstein of Antonin Scalia Law School for pointing out this article, a long but illuminating study about why it's misleading to compare murder rates in the United States with those in other "developed countries."  It starts:

Much of the political thinking about violence in the United States comes from unfavorable comparisons between the United States and a series of cherry-picked countries with lower murder rates and with fewer guns per capita. We've all seen it many times. The United States, with a murder rate of approximately 5 per 100,000 is compared to a variety of Western and Central European countries (also sometimes Japan) with murder rates often below 1 per 100,000. This is, in turn, supposed to fill Americans with a sense of shame and illustrate that the United States should be regarded as some sort of pariah nation because of its murder rate.

Note, however, that these comparisons always employ a carefully selected list of countries, most of which are very unlike the United States. They are  countries that were settled long ago by the dominant ethnic group, they are ethnically non-diverse today, they are frequently very small countries (such as Norway, with a population of 5 million) with very locally based democracies (again, unlike the US with an immense population and far fewer representatives in government per voter). Politically, historically, and demographically, the US has little in common with Europe or Japan.
The Brennan Center is an enthusiastically pro-criminal outfit affiliated with NYU. It produces a number of papers it calls "reports," but which are actually advocacy pieces for lowered sentencing.

It came out with one recently, titled "Crime in 2016: Final Year-End Data." Mostly, it presents a chipper picture.  But if you look hard enough...

Lying By Cherry-Picking

| No Comments
There are many ways to misuse numbers to intentionally create a false impression in the mind of the reader.  Such deception is morally no different than lying, in my view, even if one carefully avoids saying anything false.  Tyler Cowen at Marginal Revolution has pointed out an exceptionally egregious example from the New York Times.

The headline of the Times article is "Amid 'Trump Effect' Fear, 40% of Colleges See Dip in Foreign Applicants."  The article states:

Nearly 40 percent of colleges are reporting overall declines in applications from international students, according to a survey of 250 college and universities, released this week by the American Association of Collegiate Registrars and Admissions Officers.
The actual finding of the survey is "39% of responding institutions reported a decline in international applications, 35% reported an increase, and 26% reported no change in applicant numbers."  That is a complete nothingburger, but cherry-picking the first number creates a very wrong impression.

Cowen is too generous in the title of the post:  "This one is a real blooper and I cannot let it pass by."  The word "blooper" implies an innocent mistake or accident.  This looks like intentional deception to me, and that appearance is reinforced by the fact that the misleading story and atrocious headline are still on the NYT website three days after this has been all over the internet.  Additional commentary comes from James Freeman at the WSJ and Eugene Volokh at the Volokh Conspiracy.

We see similar cherry-picking in arguments about criminal justice, but this is such a clear and obvious example of the deceptive tactic that I thought it worth noting here.

Federal Criminal Statistics

| No Comments
The Bureau of Justice Statistics has released a report on Federal Justice Statistics 2013-2014.  Note the time lag.  That's part of the problem with justice statistics.

Another problem is that statistics are sometimes defined in ways that people would not expect.  "Tracking recidivism rates involved identifying prisoners released from federal prison following a U.S. district court commitment between 1998 and 2014."

What about people released from federal prison and subsequently prosecuted by state authorities?  Not in the definition.  Not tracked.

Three-fifths of federal arrests in 2014 were made in just 5 of the nation's 94 judicial districts -- Southern Texas, Western Texas, New Mexico, Arizona, and Southern California.  What do these five districts have that none of the others have?  You guessed it.

Since 2011, California has gone further than other states in the rapid dismantling of its tough-on-crime policies, so we have been keeping track of California crime rates as compared with the nation as a whole.  Here are graphs showing 2011 to 2015 with data from the FBI's Uniform Crime Reports.*

The "realignment" bill, AB109, took effect in October 2011, and one would not expect much effect in the first couple of months.  So we can consider 2011 to be a base year.  California shows a jump in violent crime the following year while the rate for the country as a whole was essentially flat.  California had a sharp jump in property crime for 2012, while the national rate was declining.
As has been reported on this blog and in many other places, crime across the cities of this country has been on a tear the last two years the likes of which we have not seen for decades.  After a generation of keeping our nerve and getting tough under Bill Clinton and George Bush, we decided that the real problem is thuggish police and overcrowded prisons.  So we started down the road of policing consent decrees and retroactively lowered sentences.

Liking the results?

But wait, it gets worse.  Lots worse.
The FBI issued this press release yesterday on its Preliminary Semiannual Uniform Crime Report.

All of the offenses in the violent crime category--murder and non-negligent manslaughter, rape (revised definition), rape (legacy definition), aggravated assault, and robbery--showed increases when data from the first six months of 2016 were compared with data from the first six months of 2015. The number of aggravated assaults increased 6.5 percent, murders increased 5.2 percent, rapes (legacy definition) increased 4.4 percent, rapes (revised definition) rose 3.5 percent, and robbery offenses were up 3.2 percent.
*             *             *
In the property crime category, offenses dropped 0.6 percent. Burglaries were down 3.4 percent, and larceny-thefts declined 0.8 percent. However, motor vehicle thefts increased 6.6 percent.

We at CJLF will be looking at the data more closely and will report what we find on this blog.
An implicit assumption in the outcry for releasing "nonviolent offenders" is that criminals specialize, and a person in prison for, say, burglary, is no more likely to commit a violent crime than regular law-abiding people are.  Last month, California voters approved an initiative for releasing supposedly "nonviolent" criminals by a landslide even while they rejected an initiative to repeal the death penalty by a greater margin than they did four years ago.  That indicates the extent to which the "nonviolent offender" myth has taken hold.

But it's a bunch of hooey.  Today the U.S. Bureau of Justice Statistics released supplemental data on Recidivism of Prisoners Released in 30 States in 2005: Patterns from 2005 to 2010.

How many prisoners released in 2005 were rearrested by 2010, and for what crimes?

For those committed for violent offenses, 33.1% were rearrested for another violent offense.  For those whose most serious commitment offense was classified as a property offense, 28.5% were rearrested for a violent offense.

Is that a big difference?  No, it is a small, bordering on trivial, difference.  The premise that "nonviolent offenders" can be released without placing law-abiding people at increased risk of violent victimization is just plain wrong.

Study: Racist Police Narrative is False

| No Comments
For the past eight years, the world has been told that American law enforcement is infected with racism.  The President, the Attorney General, most civil rights leaders, most of academia, seemingly every Democrat politician in the country, and the national media have repeatedly cited the disproportionate number of arrests, convictions and incarceration of young black males as proof that white police officers and prosecutors are literally rounding up blacks and throwing them in prison.  We have been told that the racial bias is so prevalent that white police officers routinely single out black suspects and murder them.  The Black Lives Matter movement was formed specifically to fight this perceived injustice and over the past couple of years more than a few police officers have been assassinated by BLM followers.  John Lott, Jr., of the Crime Prevention Research Center and Carlisle Moody of the College of William and Mary have released a study mentioned below in the News Scan which found that black suspects are more likely to be shot by black police officers and female police officers than by white police officers.

Poll Analysis Kerfuffle

| 1 Comment
The things people get angry about ...

Trouble in Poll-Land

| 1 Comment
We talk a considerable amount about polls on this blog, enough to devote a category to them.

But polling is getting harder, and the pollsters have done some belly flops in recent years, including the Kentucky Governor election and Brexit.  Ryan Knutson has this article in the WSJ on the challenges and responses.

The problem of the growing number of people who have only cellphones, no landline, and the legal prohibition on robocalling cell phones is well known.  Here is another problem I did not know about.

In 1997, 36% of households sampled agreed to participate in a poll, according to the Pew Research Center. Now it is 9%. This means thousands more calls must be made for a telephone survey to reach a sufficient sample.
Wow.  It's not just more calls.  How do we know the 9% who will talk to pollsters are representative of the 91% who will not?  Pollsters can match on demographics, but demographics are not everything.  Is it possible that willingness to take the poll correlates to views on the questions being asked, even after demographic adjustments are made?  That seems to me to be entirely plausible.

A variety of new polling methods are being used, but until they have a track record we won't know how valid they are.
As noted in a CJLF press release last month, the FBI's Uniform Crime Reports (UCR) reported a 3% increase in violent crime in 2015 over 2014.  Today, the Bureau of Justice Statistics reported that its National Crime Victimization Survey (NCVS) showed no statistically significant change.  The lesson here is in the limitations of statistics.
As Bill noted earlier, today's FBI statistics report, Crime in the United States -- 2015, reports an 11% increase in the number of murders.  The murder rate (murders per 100,000 population), rose a slightly lower but still horrific 10%.  See Table 1A.

How unusual is this?  I dove into the historical data to find out.  The Bureau of Justice Statistics online data gives us murder rates back to 1960.  The rates range from 4.4 to 10.2 murders per 100k and are expressed to only one decimal place, so there is some rounding error, but they are good enough for a quick take.

Putting the rates in a spreadsheet and calculating the percent change for each year over the previous year, we see that only twice in over a half century has the rate jumped double digits in one year and once just a tad below that threshold.  Those three years were 1966-1968, when crime was rising at a horrific rate.

We have never had a double-digit drop, but several years come close.

The standard deviation for the changes is 5.9%.  This year's change is not quite two standard deviations from the mean, but it's close.  (The mean is near zero.) 

In plain English, this is a very unusual one-year jump, although not unprecedented.

New Study Shows ...

| No Comments
The Onion reports:

Highlighting the gaping security holes that continue to persist 15 years after the attacks, an encouraging report released Thursday by radical extremist think tank the Caliphate Institute determined that the United States is no safer than it was before 9/11. "Despite efforts to expand digital surveillance and coordinate information-sharing among intelligence agencies, we discovered that the ability of the U.S. government to assess and eliminate potential terrorist threats has not substantively improved since September 11, 2001, which came as a shocking and welcome finding," said Selim Amir, chairman of the fundamentalist K Street research institute, which is staffed by prominent jihadist thinkers, visiting Sharia law scholars, and retired senior members of al-Qaeda.
The Onion is, of course, a satire publication.  The kernel of truth beneath the satire is how studies by organizations with agendas are so often uncritically reported as if they were done by neutral seekers of truth and as if they are the definitive word on the subject.

Monthly Archives