This week the Wall Street Journal’s data geek looks at data goofs throughout the ages: (Added “data” before “goofs” to clarify – 2nd pass.) (Put quotes around the term “goofs” that appears in italics – 5th pass.)
- This month, the U.K. Office of National Statistics calculated construction output for the three-month period ending in May instead of June. Diagnosis – failure to update the spreadsheet properly. (Added periods to U.K. and put a period at end of sentence – 1st pass.)
- In 2004, the U.S. Bureau of Labor Statistics couldn’t produce the producer price index. Its computer feed crashed.
- More recently, BLS misreported the number of mass layoffs because of an error in Arizona data. The bureau had to file a correction. (Revised to make clear it was an error in Arizona data, not an error in what was reported for Arizona – 1st pass.) (Made paragraph into two sentences to heighten clarity – 2nd pass.)
- And the classic – in 1999, NASA lost a $125M Mars orbiter because software failed to convert from English units to metric. As Homer would say: “D’oh!” (Aside: To emphasize the importance of units in an answer, my brother’s first tech teacher insisted all answers be given in furlongs per fortnight.) (Made it “$125M” instead of “$125 million” to follow style I prefer in this blog – 2nd pass.)
Somehow, the WSJ report omitted Standard & Poor’s $2 trillion error in projecting U.S. budget deficits. (Replaced “he” with “the WSJ report” to make it clear that the omission was by the WSJ reporter. Otherwise it reads like my brother’s tech teacher
made omitted the error – 2nd pass.) (Changed “made” to “omitted” – 4th pass.)
As actuaries, we’ve all been there. The WSJ article gives the common prescrpitives:
- Automate wherever you can. A standardized data feed will always err less than a person typing into a spreadsheet.
- Double- and triple-check. Quadruple-check if you can get away with it. When I want to make sure something is right, for example, I go over it in detail from beginning to end. If I find a mistake, I go through it again, top to bottom. If I find another mistake, I go through it again. And again and again and again, until I find no more mistakes. The last pass is a perfect one. (Added “The last pass is a perfect one” to emphasize the point of the paragraph – 3rd pass.)
That doesn’t eliminate errors, as any careful reader of this blog will know. And sometimes I don’t check as thoroughly as I’d like. But it does get rid of the worst of them. (Added “And sometimes I don’t check as thoroughly as I like” to make it clear that sometimes I fall short of the standard I set for myself – 2nd pass.)
(No errors discovered – 6th pass.)