We spent last Saturday with UCLA scholar Bill McKelvey. Currently Professor of Strategic Organizing at the Anderson School, Bill ranks among the nation's foremost authorities in complexity leadership theory. He makes a compelling case about the dangers of the longstanding social-sciences bias toward normal distributions in which most scholars "manage the averages" and "ignore the extremes."
Carl Friedrich Gauss's Hanover experiments in the early 19th Century gave birth to normal or Gaussian distributions, the well-known bell-curve shapes in which the vast majority of statistical elements exist within the customary two or three standard deviations from the mean. McKelvey reminds us that most social science has been engineered to organize around normal averages and avoid the extremes, valuing statistical tidiness over inconvenient variance.
He suggests that we really live in a Paretian world, underscoring the value of the Power Law distributions - commonly called the 80-20 Rule - conceived by Vilfredo Pareto. McKelvey believes there is so much to learn from exploration of extremes found in the statistical tails of quantitative research that are so difficult to explain.
To understand our increasingly complex world, McKelvey underscores the utility of scale-free dynamics that can cope with the fact that emergence may not occur in homogeneous, linear ways and that one extreme event may be well worth researching. He quips that upon discovering a talking pig most social scientists would wait to find a large "n" size so as to have enough talking pigs to study, inevitably leading to some kind of normal distribution used to describe an abnormal event. McKelvey says that lone talking pig is worth studying now, right now, as the extreme event it is and that researchers don't need a large "n" size to gain legitimacy. Besides, they'll never achieve legitimacy anyway if there is only one talking pig.
McKelvey brilliantly uses a 2006 Malcolm Gladwell The New Yorker article on homelessness to underscore the value of studying extremes. He uses Gladwell's data to show that over a defined period 50% of homeless people in New York City use a shelter for one night at an administrative cost of $62 each, 30% use a shelter for two nights at a cost of $132 each, and 10% use shelters for several three-week periods at a cost of $5,179 each. The vast majority of the city's attention is paid to this 90% portion of homeless people. Yet, say, given 10,000 homeless people over the period my math shows a public cost of $310,000 for the 50%, 396,000 for the 30%, and $5,179,000 for the 10%. In Paretian terms, of course, it's not surprising to learn that the remaining 10% of the "chronic homeless" cost $24,800 each or $24,800,000. So, while most public policy and operational activities are invested in the 90% that comprise $5,885,000 in expense, a small fraction (10%) of the overall total actually comprises four times that annual cost. By ignoring the 10% so-called outliers, we forfeit our ability to effect real savings and make real changes to people's lives.
A Guassian approach to homelessness dwells on the 90% while, in McKelvey's terms, a Paretian approach addresses the needs of the vastly more expensive 10%. The City of Denver embraced the extreme, provided city-paid apartments for the chronic homeless and has already reduced costs from $45,000 per chronic case to $15,000 per year. Bill McKelvey has a point. Extremes matter.
Carl Friedrich Gauss's Hanover experiments in the early 19th Century gave birth to normal or Gaussian distributions, the well-known bell-curve shapes in which the vast majority of statistical elements exist within the customary two or three standard deviations from the mean. McKelvey reminds us that most social science has been engineered to organize around normal averages and avoid the extremes, valuing statistical tidiness over inconvenient variance.
He suggests that we really live in a Paretian world, underscoring the value of the Power Law distributions - commonly called the 80-20 Rule - conceived by Vilfredo Pareto. McKelvey believes there is so much to learn from exploration of extremes found in the statistical tails of quantitative research that are so difficult to explain.
To understand our increasingly complex world, McKelvey underscores the utility of scale-free dynamics that can cope with the fact that emergence may not occur in homogeneous, linear ways and that one extreme event may be well worth researching. He quips that upon discovering a talking pig most social scientists would wait to find a large "n" size so as to have enough talking pigs to study, inevitably leading to some kind of normal distribution used to describe an abnormal event. McKelvey says that lone talking pig is worth studying now, right now, as the extreme event it is and that researchers don't need a large "n" size to gain legitimacy. Besides, they'll never achieve legitimacy anyway if there is only one talking pig.
McKelvey brilliantly uses a 2006 Malcolm Gladwell The New Yorker article on homelessness to underscore the value of studying extremes. He uses Gladwell's data to show that over a defined period 50% of homeless people in New York City use a shelter for one night at an administrative cost of $62 each, 30% use a shelter for two nights at a cost of $132 each, and 10% use shelters for several three-week periods at a cost of $5,179 each. The vast majority of the city's attention is paid to this 90% portion of homeless people. Yet, say, given 10,000 homeless people over the period my math shows a public cost of $310,000 for the 50%, 396,000 for the 30%, and $5,179,000 for the 10%. In Paretian terms, of course, it's not surprising to learn that the remaining 10% of the "chronic homeless" cost $24,800 each or $24,800,000. So, while most public policy and operational activities are invested in the 90% that comprise $5,885,000 in expense, a small fraction (10%) of the overall total actually comprises four times that annual cost. By ignoring the 10% so-called outliers, we forfeit our ability to effect real savings and make real changes to people's lives.
A Guassian approach to homelessness dwells on the 90% while, in McKelvey's terms, a Paretian approach addresses the needs of the vastly more expensive 10%. The City of Denver embraced the extreme, provided city-paid apartments for the chronic homeless and has already reduced costs from $45,000 per chronic case to $15,000 per year. Bill McKelvey has a point. Extremes matter.