Recently in Studies Category

I have noted on this blog a few times the hazards of putting too much faith in studies conducted with mathematical models.  See this post in 2011 on the interesting book Models Behaving Badly and this one last year on the disastrous failure of a sports prediction model.

One good thing about election prediction models is that we get quick and conclusive results about whether they are right or wrong.  Charles Forelle has this story in the WSJ about the meltdown of British pollsters in last week's election.

The website FiveThirtyEight, run by statistics guru Nate Silver, who made his name with accurate predictions of U.S. presidential elections, used a model developed by British academics. It came up with 278 seats for the Conservatives and 267 for Labour, and put the probability at 90% that the Conservatives would win between 252 and 305 seats. They won 331.
The difference between the prediction and the outcome was double what was supposed to be a 90% confidence interval.  That's more than failure; that's crash and burn.

The reason I bring this up is to remind everyone to be skeptical when someone asserts that some proposition is proven because "studies show" it to be true when the studies are based on mathematical modeling.  Such models can provide indications but not definitive proof, and it is a serious error to rely on them entirely.
"Even a dog distinguishes between being stumbled over and being kicked," Justice Holmes famously said. 

The role of intent in distinguishing criminal acts from noncriminal accidents and higher-degree offense from lower ones is deeply ingrained in our law.  It may be deeply ingrained in our brains.  Robert Sapolsky reports for the WSJ on a study of scanning people's brains as they read about intentional and unintentional killings.  The study is Treadway, et al., Corticolimbic gating of emotion-driven punishment, Nature Neuroscience 17, 1270-1275 (2014).
It is always important to remember that most people with mental illness are not involved in crime nor are they prone to violence.  In fact, treatment for mental illness is fairly ubiquitous in our society.  The next time you are at a family gathering or among your colleagues there is a fair chance that someone in your presence has been struggled with depression, addiction or some of form of mental illness. 

But we still hear claims that there is no relationship between mental illness and crime or violence.  We hear this despite numerous studies, including several population-based studies, that have shown time and time again that there is indeed a relationship between mental disorder and crime.  The latest issue of Psychiatric Services provides the latest evidence:

Bipolar disorder is a severe and prevalent psychiatric disease. Poor outcomes include a high frequency of criminal acts, imprisonments, and repeat offenses. This critical review of the international literature examined several aspects of the complex relationship between individuals with bipolar disorder and the criminal justice system: risk factors for criminal acts, features of bipolar patients' incarceration, and their postrelease trajectories.

Publications were obtained from the PubMed and Google Scholar electronic databases by using the following MeSH headings: prison, forensic psychiatry, criminal law, crime, and bipolar disorder.

Among patients with bipolar disorder, the frequency of violent criminal acts is higher than in the general population (odds ratio [OR]=2.8, 95% confidence interval [CI]=1.8-4.3). The frequency is higher among patients with bipolar disorder and a comorbid substance use disorder than among those without either disorder (OR=10.1, CI=5.3-19.2). As a result, the prevalence of bipolar disorder among prisoners is high (2%-7%). In prison, patients' bipolar disorder symptoms can complicate their relationship with prison administrators, leading to an increased risk of multiple incarcerations. Moreover, the risk of suicide increases for these prisoners.

Criminal acts are common among patients with bipolar disorder and are often associated with problems such as addiction.

California v. National Crime Rates

| No Comments
Here is an update on California versus overall United States crime rates.  The table below shows the rates per 100,000 population for the FBI's violent crime index and property crime index for 2011, nine months of which predates the Realignment program, and 2013, the most recent year with full data available.  The data are from the downloaded files on the FBI's Crime in the United States reports for the respective years.

The nation as a whole had a 6% drop in property crime over the two-year period, while California had a 3% increase, a difference of 9%.

How Dumb Do They Think We Are?

| No Comments
Newsflash:  Incapacitating people who commit crime has no effect on the amount of crime.

Do you believe that?

A far left advocacy group, the Brennan Center, wants you to.  The press release from their latest propaganda (heralded, as ever, as a "study") states:

Increased incarceration had some effect, likely in the range of 0 to 10 percent, on reducing crime in the 1990s. Since 2000, however, increased incarceration had a negligible effect on crime.

Convincing the public that there is little or no relationship between (1) increased incarceration of people who commit crime and (2) enormous crime reduction over the last quarter century is critical to the efforts of the pro-criminal lobby to sell miniaturized sentences (which they understandably call by the opaque name "sentencing reform").  The lobby knows by now  --  in part because of its humiliating failure in Congress  to pass the Smarter Sentencing Act  --  that the public simply is not going to buy slashing sentences as long as it understands that a crook who's in prison is not ransacking your house while you're at work, or selling heroin and similar goodies to your teenager.  Hence the effort to convince us that incarcerating criminals has nothing or next to nothing to do with crime reduction  -- arguably the most important domestic policy success of the last fifty years.

Related Newsflash:  The centuries-long link between crime and punishment just disappeared.

These people are a hoot.

UPDATE:  The Heritage Foundation, which takes the same robust pro-"reform" position as the Brennan Center but is a great deal more honest, recently took the view, through its distinguished Fellow, John G. Malcolm, that increased incarceration could be credited with between 25 and 35 percent of the last generation's crime reduction.  See Mr. Malcolm's remarks starting at 7:35 of this tape.  Somebody's telling a whooper, and it's not John Malcolm.

Fingerprint Error Rate: Zero

| No Comments
Natalia Zea reports for CBS Miami on a study of fingerprint match reliability.

The fingerprint examiners correctly matched every single print in the tests, with only 3 percent of the inaccurate matches caught by a second examiner, which is part of normal protocol at crime labs across the country.
That's rather awkward phrasing.  Presumably it means that 3 percent of the matches found by the first examiner were not correct but all of the errors were found and corrected on the second examination.
The kerfuffle over the measles vaccine is off-topic for the blog.  It is also ultra vires (not virus) for CJLF.  I have watched the discussion with some interest, though, particularly the propensity of media outlets of a given orientation to ascribe anti-vaccine beliefs to the other side.  Crunching some actual numbers was a fun project (in a nerdy sense of fun) for a soggy Saturday here on the Left Coast, and I found some interesting things about how political orientation relates to an essentially nonpolitical subject.

Correlation, Causation, and Education

| No Comments
How many times have we heard "correlation does not prove causation"?  Too many to count.  How many times have we heard that elementary truth recited and then ignored, as people proceed to argue for policy changes based on correlation alone.  Almost as many.

Charlie Wells has this article in the WSJ on financial education of kids and their financial behavior as adults.  It has nothing to do with crime, but it sounds a cautionary note about the argument we hear all the time.  "Studies show that educational program X is correlated with positive outcome Y.  Therefore we must spend more on X to produce Y, and that is more important and more cost effective than the solution only you ignorant rednecks believe in, Z."  It does not follow.

Correction Squared

| No Comments
An amusing little notice from the Bureau of Justice Statistics in my mailbox today:

The Corrections Statistical Analysis Tool (CSAT) - Parole has been corrected on the BJS website

Dude, Where's My Car? (Update)


Nearly two years ago, we noted a disturbing statistic from half-year preliminary crime data from the FBI.  Here is an update from later and more complete data.

Of the categories of crime tracked by the standard statistics, auto theft is the category most likely to have been affected by California's prison realignment program.  Auto theft was always a felony.  Before realignment, it was therefore a crime for which a state prison sentence was always a possibility, although the judge had discretion to give a lesser sentence.  After realignment, auto theft is a crime for which a person can never go to state prison.  Not for the 97th offense.

A New Forensic Tool for Rape Cases?

| No Comments
Rachel Feltman reports in the WaPo:

A new study suggests that the microbes present on pubic hair -- which vary from person to person -- could be used as evidence in sexual assault cases. This particular research is in its early stages, so you probably won't hear about genital microbes in a courtroom anytime soon. But the study is just one example of the effort to turn the incredible diversity of the bacteria that live on human beings into a high-tech forensic toolkit.
*                                      *                                   *
The researchers, led by Silvana Tridico from Murdoch University, took scalp and pubic hair samples from seven individuals (three male and four female, with one co-habitating couple in the mix). While hair from the head had around 50 kinds of bacteria a pop, and seemed to be influenced by the environment, pubic hairs had over 70 kinds of bacteria each, which were highly individualized. That's in line with previous studies on the vaginal microbiome, which has shown an unexpected diversity distinguishing one individual from another.

"The advent of DNA profiling has resulted in an increase of sexual offenders using condoms, which they take away, post-assault," Tridico said in a statement. "The implication of this present study is that the transfer of bacteria between victim and offender, in rape cases, may provide a new way of linking the offender to the victim, in instances in which no human DNA is transferred."

Missing Data on Justifiable Homicides

| No Comments
The quality of any study cannot be any better than the quality of its data.  Modern statistics computer programs can produce a dazzling array of colorful charts and graphs, but if the input data are faulty, it's all pretty nonsense.  As we used to say when I was an IT guy, "garbage in, garbage out."

The main source of data for people studying crime has long been the Uniform Crime Reports compiled by the FBI.  These numbers are initially reported by local police agencies.  There are some known problems and some controversies.  Collection and submission of data by the agencies is voluntary.  There are even charges of intentional manipulation in some cities.

Then there are the Supplemental Homicide Reports.  These provide much more detail on homicides, but compliance is even more spotty.  Rob Barry and Coulter Jones report in the WSJ that there are yawning gaps in the data on justifiable homicides by police.  For example, Fairfax County, Virginia, did not report justifiable homicides at all because, well, these are crime reports and justifiable homicides are not crimes.  Of 105 large police agencies contacted by the reporters, justifiable homicides from 35 of them did not appear in the FBI records at all.

What's a researcher to do?  You must know the limitations of your data.  I once started a project using the SHR where a key data point was the circumstance of the homicide.  I had to shelve it because so many cases had that data point missing that the input data were essentially worthless.  Bad research producing very wrong results can be done by people unaware of the limitations of their data.  Of course, if the researcher is actually an advocate seeking to bolster the Politically Correct position on a controversy, then truth doesn't matter.  Damn the limitations, full speed ahead!

Underreporting of Crime

Last week I had this post on the Uniform Crime Reports, and a commenter noted that skepticism was in order due to underreporting by some police departments.  Across the pond there appears to be a major kerfuffle on this point.  David Barrett reports in the Telegraph:

Almost a million crimes a year are disappearing from official figures as chief constables attempt to meet targets, a study by the police watchdog has disclosed.

Its report exposed "indefensible" failures by forces to record crime accurately, and said that in some areas up to a third of crimes are being struck out of official records.

Her Majesty's Inspectorate of Constabulary said violent crimes and sex attacks were particularly vulnerable to being deleted under "inexcusably poor" systems.

Although the report stopped short of accusing police of widespread "fiddling" it said there was an "undercurrent of pressure not to record a crime across some forces" and "wrongful pressure" by managers.

It means violent criminals and even rapists are not investigated, potentially allowing offenders to strike again.
The FBI has finally come out with the 2013 statistics for Crime in the United States, almost two months later than last year.  The good news is that crime is down from 2012 about 5% in crimes per 100,000 population nationwide on both the violent (-5.1%) and property (-4.8%) scales.

Last month, we noted the good news that California crime was down, but we were interested in seeing the national figures for comparison to sort out national trends from possible effects of California's sentencing "realignment."  Last year's post making that comparison is here.

We look at property crime as the primary indicator, as persons convicted of violent crimes, either for the present offense or as priors, are not eligible to be shunted off to county jail under realignment.  Many property crime convicts are, and given that the jails are overcrowded in most counties they either get released early or they push out other inmates for early release, likely other property crime convicts.

California's overall property crime rate is down less than the national average, -3.8% versus -4.8%.  Auto theft is the category tracked by the FBI that is most likely to be affected by realignment, because all auto thefts in 2013 (pre-Prop. 47) were realignment-eligible felonies, while other categories are mixed eligible/ineligible or felony/misdemeanor.  Consistent with the realignment-effect hypothesis, California's improvement in auto theft lagged considerably behind the nation, -2.8% versus -4.0%.

Comparing 2010, the last full year before realignment, with 2013, property crime has dropped 7.2% in the nation while rising slightly, 0.8%, in California.  Auto theft is down 7.3% nationally but up 5.3% in California.

Political Bias in Psychology?

| No Comments
In the latest issue of the New Yorker, Maria Konnikova has a story that asks, is the field of psychology politically biased?  As the popular saying goes, to ask the question is to answer it.  Anyone who has spent any time within the field of psychology knows, if they're being honest, that the answer to that question is an emphatic "yes" all around.  But what is the nature of the bias? 

Monthly Archives