Tuesday, 4 December 2012

DNA, Nametags, burglars and rapists

 The Night Stalker revisited

Serial offenders sometimes catch the public interest, but only when their crimes are particularly gruesome. Many criminals are serial offenders, and this includes many burglars. Criminals establish a routine, a modus operandi, and stick to it. It may be a lack of imagination, or simple pragmatism. If a way of committing a crime gets them what they want without detection, they persist in it. Far from being anything special, serial offenders are simply criminals with a habit of offending.  It’s a bit like the Olduvai Tool Set, a collection of stone implements our ancestors made, with minimal variation, for 600,000 years.  In that era we were very conservative or not very bright, most probably both.

Anyway, back to South London. Burglary has a significant psychological impact, particularly on the elderly, who are helpless. The implied sanctity of the home is violated, with a consequent pervasive fear and loss of security on the part of the victims.  However, Delroy Easton Grant was more than a burglar. He specialised in raping elderly women and very occasionally elderly men. He did so with such violence that one woman’s life was threatened by a perforated bladder. He probably committed over 200 offences (other estimates are much higher) starting in 1990, and was not caught till May 2009.

Why did it take so long? Contrary to popular TV detective series, it is hard to catch an accomplished criminal who strikes at random, with long dormant periods, and who leaves very few forensic clues. This is no criticism of the Police. One cannot protect every old lady in South London for years on end simply because a particular violent rapist may strike again.

One possible reason for their failure is that Police encountered difficulties because of the assailant’s race. Victims described him as a black male aged between 25 and 40, about 5'9" to 5'11" tall, of slim athletic build. Police at one stage issued a confusing photo-fit picture, in which the balaclava clad assailant seemed to have a white face. That distraction aside, race was an issue because an advanced DNA analysis of his sperm showed that he was very probably from the Windward Islands ( St Lucia, Barbados, St Vincent, the Grenadines, Tobago or Trinidad). Police identified around 21,000 possible suspects that fitted such a profile. So, the solution seemed simple: DNA as many of those suspects until a match was found.

Despite a promise that DNA profiles which did not match the assailant would be destroyed a number of putative suspects declined to give a DNA sample. To some it seemed that the procedure was stigmatising an entire community, and 125 flatly refused to provide samples. This misunderstanding of the process, or distrust of the Police, perpetuated the horror. Police were left with 1000 potential suspects.
However, consider the only other forensic clue the rapist left: a size ten footprint of a particular brand of trainers. Would it have been better to trace that? It might have been. A DNA profile is unique, while anyone can buy a publicly available brand of trainers, though the rarer brands such as the one found in this instance might be traceable. It was a positive indicator, but hard to track down.
So, looking at the task from a Police perspective, they were trying to find a black rapist without the full support of the relevant target population of suspects, and had a DNA signature which they simply had to match against a name.

Is it easier to catch a burglar than an episodic gerontophile rapist? Frankly, the Police have more experience of burglars. Every criminal has several signatures. Almost as good as DNA, footprints, fingerprints and the like is the MO, the modus operandi. What Police knew was that the assailant seemed to have an unerring skill in tracking down the homes of the elderly.  He never broke into a house occupied by anyone other than a lone elderly occupant. He once targeted three houses in a single street. He picked detached or semi-detached houses and bungalows but never flats. Therefore he must have spent much time reconnoitring. To do this almost randomly over a wide geographic area he needed a car or motorbike, or both.

The Police had hit a blank wall with their DNA enquiries. They could only wait till those who had refused to proffer their DNA got arrested for other crimes. Unknown to them, they were up against an even bigger problem. Delroy Easton Grant had been incorrectly listed as having been eliminated from suspicion. There are 63,000 Grants in the United Kingdom. Delroy as a first name is disproportionately found among black men, so in this case the nametag was a distractor, not an identifier. A young policeman made a clerical error and accessed the wrong Delroy Grant, who was already on the Police database but whose DNA did not match that of the rapist. One Delroy covered for another. As a consequence the local police did not follow up the car licence plate number that would have brought them to the real burglar Delroy Easton Grant, whose DNA would have shown him to be the Night Stalker rapist. While teams were trying to do the fancy stuff, and push the DNA analysis to its outer limit, getting geneticists to come up with a possible estimate of the assailant’s appearance, an elementary error was made on the tag which constitutes a name. So, through carelessness, the unique identifier of the genetic code was mistakenly thought not to match the supposedly unique code of a given name.

Into this impasse came a new team. They decided to try to catch a burglar who spent a lot of time reconnoitring, and about whom they had some geographic leads. This was a very staff intensive procedure (the total cost of catching Grant was £10 million) and involved staking out likely suspects and likely places. Car license plates provided the other set of unique identifiers. In a way, they looked where the light was brightest in terms of their procedural capacities. This is often a sensible procedure.
At this point it might be interesting to take a Bayesian approach, in which the evidence about the true state of the world is expressed in terms of degrees of belief.

In The Theory That Would Not Die: How Bayes' Rule Cracked the Enigma Code, Hunted Down Russian Submarines, and Emerged Triumphant (2012) Sharon Bertsch Mcgrayne  described, amongst many other things, the use of Bayesian statistics to catch submarines. The initial indicator would be a radio location based on a transmission from the submarine. Sending a plane straight to the location might have seemed sensible. However, between the transmission and the plane arriving on site, many hours would have elapsed. It would be better to guess where the submarine was going to.  One could make some educated guesses and express those as prior probabilities.

How  best way to look?  Operational research showed that if a spotter on a plane saw anything, it was only in the first 15 minutes of their watch. Looking down at the sea is deadly boring. So, they made everyone change windows every 15 minutes. Research showed that if a spotter spotted anything it was on the horizon, or just below the horizon, and the sunlight had to be in the right place. So, they made the spotters look at that zone and nowhere else. In this way, coupled with a well thought out search routine (using Bayesian estimates of the most likely areas of sea given the time since the original intercepted transmission) they increase their operational effectiveness.

Operational research can be applied to policing. Given that names are so important as tags on human beings, making triple sure that the names are right should be a priority. Criminals often change their names, or give false names, or use different variants on their names in different circumstances. They also give false home addresses, and conduct their businesses from yet other addresses, and use cars registered to other people at yet other addresses.

Delroy Easton Grant, a Jehovah's Witness and father of eight, was a carer for his disabled wife but also living with another woman. He turned out to have been born in Jamaica, and was dark-skinned. He had a very long criminal career, beginning with petty theft and going on to armed robbery and vicious attacks on his partners. Neighbours found him warm, charming and always friendly. He followed cricket, liked to fish and enjoyed community barbeques where he would share jokes and reminisce about his childhood in Jamaica.

Some of the errors:
1                     Mixing up the names. The nametag must be firmly attached.
2                     Being more specific about ancestral DNA than was warranted.
3                     Issuing a misleading photo-fit suggesting he was white or light skinned
4                     Searching for a specific rapist rather than an accomplished burglar with a characteristic MO.
5                     Under-financing the operation in the early stages.

One moral of this story is that the psychological pull of DNA as a search tool may be entrancing the police, to the detriment of systematic data collection and normal, standard police work. It may be far better as a confirmatory technique in any setting in which a substantial minority will not cooperate with testing.
Another moral is that the search for ancestral origins identified the wrong Caribbean island, and added dangerous noise to the already well known fact that he was a black man.

Another possible moral is that psychological profiling can be a distraction, since it offers apparent specificity without sufficient reliability. It was another step too far into sophistication.

It may be best to look where the light is brightest, in this case for any person out late at night apparently checking out houses but never flats. Grant was least at risk when he was raping his victims, because he always disconnected electricity and phone lines. It was the elaborate reconnoitring which put him most at risk.  So, the task was not to catch him, but to catch burglars like him, in the areas where his rapes had been reported, in the hope that one of the burglars would be him.

Caution: This note was written with the benefit of hindsight, and based on publicly available accounts of the investigation, not on any internal documents. Almost every post hoc review of persistent criminal operators shows that they might have been caught earlier. The positive predictive value of criminal indicators is low, which is why so much police work is boring and routine. Even a good indicator may not be specific enough. Being a part-time cab driver, as Grant was, is a good cover for burglary, but most part-time cab drivers are not burglars.

Catching a wily criminal is hard. Searching extensively where the task is easiest (because a high frequency behaviour leads in a few cases to a very low frequency behaviour) may seem paradoxical, yet it has its advantages.

No comments:

Post a Comment