Towards a better measure of everything

Can there be a good single metric that can measure everything we care about—the things that make human life good? Probably not. But maybe we can try for less bad. Right now the metrics we use are bad.

I think there’s little general recognition of the fact that the Dow Jones Industrial Average is reported on every newscast, every night as if it is an indicator of how the economy and thus the country is doing on the whole.  The fiction that it represents the country is so embedded that it isn’t even explained any more.  Instead, it’s read off by the news anchor, in a perfunctory manner, unless of course the market drops, in which case the anchor sounds concerned and offers a perfunctory explanation, unless it drops by a lot, in which case an “expert” is called in to provide a sound bite on what was concerning the market that day.

How can a collection of 30 stocks be the measure of everything, rather than an indicator of how some HFT algorithms have decided to run momentum while the market is influenced by curious central bank policies?

It’s not just the DJIA that’s odd. As Adam wrote some time ago:

But consider GDP.  It’s just a single figure which tells us basically nothing about how the money that constitutes it is actually spent.  Which means, among other things, that GDP is e.g. indifferent to money spent on cleaning up pollution vs money spent treating people who have pollution-caused diseases.  There was an old Harper’s piece about this sort of thing.

What are our alternatives? There are dozens of such metrics today, including Bhutan’s Gross National Happiness, the Genuine Progress Indicator, and simple metrics like the Human Development Index.  A common objection to such metrics is that they are trying to measure something that can’t be measured: aggregate human well-being.  Another objection is that they take for granted the economic system and/or base metrics we use to measure certain properties of society today.

When you can’t win, change the rules: instead of aiming to accurately measure aggregate human well-being, we might aim for a couple of more tractable goals:

  1. To sever the false sense of connection between the stock market and general well-being.
  2. To provide a tight feedback loop between the metric(s) used and the levers used to change the system.

I’ve already written about the first goal, so let me explain the second.  Let’s take the Human Development Index as an example.  While the HDI measures a number of important dimensions, including health and education, one of its primary weaknesses is that it is computed once per year.  In any system, the length of the control loop—the lag between action and visible change—governs the system’s responsiveness.  For example, a home heater that takes a long time to respond to a change in the thermostat is inherently harder to control than one that responds quickly.  While few levers that can be used to alter our economic system’s course can be measured by the day, it’s likely that over the course of weeks or months trends will become visible.

In addition to improving responsiveness by using a shorter feedback, there’s a second benefit to daily updates.  For some time I’ve wondered whether the popularity of fivethirtyeight during the 2008 election was that it collected lots data in one place, was updated almost daily, and yet was able to summarize that information with a single number (or small set of numbers).  The popularity of Nate Silver’s daily updates seems similar to the appeal of using the DJIA as a proxy for the economy or progress.

What if we were to compute one or more daily metrics of economic, social, and/or cultural well-being?  What would be included in such a metric or set of metrics?  How would it be computed?  We have a preliminary idea of what this might include, focused on the United States for now, and plan to write more on this in the future.  For now, let’s give it a name: the DOMestic index—the DOW upside down.

Leave a Reply


Responses to “Towards a better measure of everything”