One strike and you're out! (or Reinhart-Rogoff vs. New Zealand 1951)

The 1951 New Zealand waterfront strike closed the country's ports for 151 days and dealt the most serious blow to Reinhart and Rogoff's 2010 growth calculations.

Most of the discussion of the Reinhart-Rogoff 2010 debt paper since the publication of the Herndon, Ash and Pollin's critique has focused on the Excel coding error that inadvertently deleted the first five countries in the sample. (On the whole irrelevance of the study for establishing a purported 90% debt cliff, see my previous post.) While this struck many as the most egregious instance of data shoddiness in their study, it's impact on the discrepancy HAP found for mean growth rates for debt over 90% was in fact negligible.

What's really driving the 2.3% discrepancy, all of 2%, in the mean growth rate of the most highly indebted countries, is RR's treatment of the five New Zealand observations in this debt category in the period 1946-1951 (see HAP p. 10). Thus RR (cherry)picked only the 1951 value, -7.6%, leaving out the other four: 7.7%, 11.9%, -9.9%, and 10.8%, thereby lowering NZ's average growth rate when its debt was over 90% from 2.6% to -7.6%. Since average country growth rates are weighted equally in RR's scheme (as opposed to weighting each country-year observation equally, as HAP propose), this one observation is weighted as much as, for example, the 19 years of British over-90% indebtedness.

One can accuse RR of cheery picking and using biased (or at least debatable and not highlighted) weighting schemes in their study or not, as one wishes. But another aspect of the New Zealand story seems to have escaped attention.

In 1951 New Zealand was in the midst of an export boom caused by the outbreak of the Korean War (the 1950 growth rate of +10.8% is already an impressive indicator of this).

So why does growth collapse in 1951 to -7.6%?

Simple. The economy had been paralyzed by the 151 day waterfront dispute that closed down the ports and that was the longest and most violent instance of industrial unrest in New Zealand's history (a country then as now with a very high trade dependence). No wonder the growth rate tanked. It had nothing to do with NZ's indebtedness and in fact was an exception in an otherwise booming era. (See here and here for historical background. Thanks to littlegreyrabbit for raising this issue at Retractionwatch.)

In their by now numerous  but ambivalent mea culpas, RR appeal for understanding in view of the difficulties of recreating historical data. But why this is an excuse for the lacunae in their use of the panel data (the 2010 paper gives no indications of these gaps or the reasons for choosing some available years and not others) is never made clear.

But their failure to inquire deeper into why the New Zealand growth data show such an erratic pattern that had nothing whatsoever to do with the country's indebtedness seems to be symptomatic of many econometric studies that cavalierly gloss over the nitty-gritty of the historical record. Since so much of their 2010 result hinges on this one observation, the lack of such an inquiry seems all the more inexcusable in economists with such an expressed concern with that historical record.

Moral of the story: When you mess with New Zealand data, one strike and you're out!


PS: For some real historical nitty-gritty on the 1951 NZ waterfront lockout/strike, watch this superb labor history video.

[Postedit 27 May 2013: For final summing up of these issues, see post Reinhart-Rogoff vs. New Zealand: Final Round?]