Lott cooks some more statistics

A while ago I wrote how Lott had seriously misrepresented NCVS data and given dangerous advice when he claimed that in robberies and assaults passive behaviour “is by far the most likely to result in injury”. Lott has now posted a response where he falsifies a table in an attempt to prove his point.

Here’s the table Lott posted to support his claim:

Percent Injured after Self Protection Action
  Robbery Assault
Any SP with gun 7.7% 3.6%
Chased, tried to catch O 9.6% 9.0%
Ran/drove away; tried to 4.9% 5.4%
Screamed from pain, fear 22.0% 12.6%
Threatened O without weapon 5.8% 13.6%
No SP measures at all 23.6% 55.2%

Notice how the numbers in in the last row are higher? I guess that proves that taking no self-protection measures is more dangerous… Except that Kleck’s table 7.1 (the table that Lott extracted from) actually has these numbers in it:

Percent Injured after Self Protection Action
  Robbery Assault
Any SP with gun 7.7% 3.6%
Chased, tried to catch O 9.6% 9.0%
Ran/drove away; tried to 4.9% 5.4%
Screamed from pain, fear 22.0% 12.6%
Threatened O without weapon 5.8% 13.6%
No SP measures at all

Notice how the correct table doesn’t have any numbers in those last two cells? If you think about it, you’ll realize that it doesn’t make sense to have numbers in those cells. By definition, the number is zero. If you don’t take any self-protection action, it is just not possible to be injured after your self protection action. So where did the numbers in Lott’s version come from? Well, Lott copied them from a different column. He has used the numbers for whether the victim was injured at all, even though his table is labeled differently.

To see the difference this makes, here are the corresponding numbers for the other rows:

Percent Injured either before or after
  Robbery Assault
Any SP with gun 12.8% 27.9%
Chased, tried to catch O 34.4% 58.2%
Ran/drove away; tried to 32.3% 38.4%
Screamed from pain, fear 69.3% 94.1%
Threatened O without weapon 30.0% 57.1%
No SP measures at all 23.6% 55.2%

Once you use comparable numbers, passive behaviour doesn’t look so bad.

Now, to be fair, this table doesn’t show that screaming is more dangerous than doing nothing since most of the people who screamed did so after they were injured, so you can’t blame the injury on the screaming. Unfortunately neither table tells us what we need to know to get an idea which is the safest choice. Really you’d have to ask the “No SP” people something like: “Consider the point in time when you had a chance to scream or try to run away or something. Were you injured after that point in time?” And I’m not sure whether you’d get a sensible answer from such a complicated question.

None of this, however, justifies Lott fabricating a table by mixing numbers from two different columns.

Comments

  1. #1 Look closer
    February 15, 2004

    How is it relevant that someone is injured before they brandish a weapon? Presumably if the attacker did not know about the gun (or hear the scream) how would this effect the attack rate?

    Once the self-protection measure is attempted we need to how many injuries follow from that.

    “Now, to be fair, this table doesn’t show that screaming is more dangerous than doing nothing since most of the people who screamed did so after they were injured, so you can’t blame the injury on the screaming.”

    You seem to have a glimmer of understanding.

    But conclude “Unfortunately neither table tells us what we need to know to get an idea which is the safest choice.”

    Well, yes they do. Take self-protection measures as soon as you can. They are shown to be better than doing nothing.

    Once you’re injured things change. Don’t let yourself get to that point.