Clarification please.
The journal of the American Economic Association published a highly policy-influential - hand-drawn - downward-sloping curve, although the data suggests an upward-sloping relationship.
The HER referee admits that Samuelson and Solow "made a mistake" and, therefore, ... the issue must not be raised in print?
In this instance (as in many others) the costs of referee anonymity outweighs the benefits: shouldn't the referee explain this logic?
----- Original Message -----
From: "William R (Rich) Hart" <[log in to unmask]>
To: [log in to unmask]
Sent: Tuesday, 22 April, 2014 1:05:47 AM
Subject: Re: [SHOE] Samuelson and Solow on the Phillips curve
In response to Robert Dimand.
We estimated the Samuelson and Solow Phillips curve (SSPC) over the 25 year
period 1934 to 1958 because Samuelson and Solow (1960, 192, Figure 2) state
in the caption to their hand-drawn Phillips curve: “…roughly estimated from
the last twenty-five years of American data.” They did not say ‘…roughly
estimated from the last twenty-five years of American data excluding the war
years.’ So, in the first instance, we included the war years because
Samuelson and Solow seemingly included the war years.
Second, Dimand claims we ignored problems with the WWII data in our HER
paper despite this having been brought to our attention in referee comments
at another journal. We were well aware of the problems with the WWII data
(who isn’t?) just as Samuelson and Solow were surely aware. We did not
ignore this issue (see below) so much as we considered it irrelevant to the
main purpose of our paper. Our paper was not about data problems or about
what data Samuelson and Solow should or should not have used in their
‘estimate’ of the Phillips curve (hence, what data we should or should not
have used in our estimate). The primary purpose of our paper was much less
ambitious—-it was to answer a simple question: in 1959-60 could Samuelson
and Solow, using the data they said they used and employing simple,
straightforward econometric techniques available at the time, produce the
Phillips curve they claimed as fitting the data over the twenty-five year
period 1934 to 1958? Our answer was no, a resounding no. Given that the SSPC
paper is one of the most widely read and widely cited papers in the
profession, one would think our results would/should be of interest to
economic historians.
More directly to Dimand’s claim that we ignored problems with the WWII data
in our HER paper, he is correct if by ignore he means we did not re-estimate
the SSPC over the 1934 to 1958 period excluding the war years. We did not do
this. However, it is incorrect to claim that we totally ignored the WWII
data problems.
One HER referee recommended rejection because:
"The principle problem is this. In their scatter plot of data on page 188,
Samuelson and Solow have circled 13 points (it is not clear exactly how many
there are) and said that the circled points are ones from 'recent years'.
It is visually pretty obvious that the hand drawn Phillips curve on page
192, although described as based on 25 years, is in fact based on the
circled points. .... My conclusion is that Samuelson and Solow made a
mistake in the way they described the hand-drawn curve. They did not in fact
use 25 years of data, they used 13, and their hand drawn curve is about
right."
In response to this comment, we re-estimated the SSPC over the period 1947-
1958. As we discuss in footnote 16 of the HER paper, the re-estimated SSPC
(that excludes WWII data) still does NOT fit the hand-drawn SSPC. The re-
estimated SSPC, like the Phillips curve estimated over the twenty-five
period 1934-1958, is also hump-shaped—-the unemployment rate and inflation
move in the same direction for unemployment rates below 4% and move in
opposite direction for unemployment rates above 4%.
Finally, Dimand’s comment brings to mind the many obstacles we faced in
getting this paper published. Samuelson and Solow hand-drew a downward-
sloping Phillips curve based on a 25-year period that includes WWII and nary
a peep of concern. Using their data, we estimate the Phillips curve over the
same twenty-five year period in an attempt to replicate their results (using
a statistical technique that would be exceedingly difficult to manipulate),
find that their results cannot be replicated, and are subject to a barrage
of criticism.
|
