Download Testing Austrian Business Cycle Theory? A

Testing Austrian Business Cycle Theory? A Rejoinder to Andrew Young
By
Robert P. Murphy
Adjunct Scholar
Mises Institute
518 West Magnolia
Auburn AL 36832-4528
[email protected]
William Barnett II
Professor of Economics and Chase Bank Distinguished Professor of International
Business
Joseph A. Butt, S.J., College of Business
Loyola University New Orleans
6363 St. Charles Ave.
New Orleans, LA 70118
(504) 864-7950
[email protected]
http://www.business.loyno.edu/faculty/wbarnett
and
Walter Block
Harold E. Wirth Eminent Scholar Endowed Chair and Professor of Economics
College of Business Administration
Loyola University New Orleans
6363 St. Charles Avenue, Box 15, Miller 318
New Orleans, LA 70118
c.v.: http://www.cba.loyno.edu/faculty.html
office: (504) 864-7934
dept: (504) 864-7944
fax: (504) 864-7970
[email protected]
[email protected]
1
Testing Austrian Business Cycle Theory? A Rejoinder to Andrew Young
Abstract:
Young (2005) attempts to test Austrian Business Cycle Theory (ABCT). The present
paper is devoted to an examination of this test. Our conclusion is that Young’s paper fails
on several grounds. In the first place, Young’s model fails on its own terms. In
particular, we come up with significantly different parameter estimates using the same
data. Beyond these serious objections, there is the more fundamental difficulty that
Young’s approach, even if conducted flawlessly, would still be an improper “test” of
ABCT.
Key Words
Austrian Business Cycle Theory; testing economic theories; illustrating economic
theories
JEL Category
E32
2
Testing Austrian Business Cycle Theory? A Rejoinder to Andrew Young1
I. Introduction
Young (2005)2 sets out to test Austrian Business Cycle Theory (ABCT).3 This is a very
welcome development, in that this theory has been around for a long time (Mises, 1912)
and has suffered unwarranted neglect at the hands of the economics profession, even
though one of their number had won the Nobel Prize in Economics (Hayek, 1931).4
Although we shall find some errors in Young (2005) we enthusiastically welcome this
publication to the economic literature.
In section II we provide a quick review of ABCT. In section III we provide a summary
of Young’s model, accompanied by an immanent critique, taking Young’s approach at
face value and demonstrating its internal problems. In section IV we broaden our critique,
by arguing that even if all of the problems in section III were corrected, Young (2005)
nonetheless provides no serious difficulty for the Austrian macroeconomist. This is
because Young has misconstrued the empirical claims of ABCT; even if we agreed with
his parameter estimates (which we do not) they would be entirely consistent with
Hayekian cycle theory.5 In short, we argue in section III that Young mishandles his own
model, while in section IV we argue that he chose the wrong model to begin with! We
conclude in section V.
II. A Review of Austrian Business Cycle Theory
In this section we offer a very brief review of ABCT, in order that the reader may fully
understand Young’s critique—and our own dissatisfaction with his “test.” Naturally this
description will not provide a comprehensive exposition; the interested reader is referred
to the citations in footnote 3.
1
We would like to thank NYU graduate student Paul Dower for his helpful comments, but especially for
independently running a regression that confirms our own results.
2
All references to Young are to his 2005 paper.
3
For an explication of ABCT, see Barnett and Block, 2005, 2006B; Block, 2001; Block and Garschina,
1996; Carilli and Dempster, 2001; Callahan and Garrison, 2003; Cochran, Call, and Glahe, 2003; Cwik,
1998; Garrison, 1994, 2001, 2004; Garrison and Bellante, 1988; Hayek, 1931; Mises, 1998; Powell, 2002;
Rothbard, 1993, Sechrest, 2001.
4
Co-Nobel Prize winner in 1974. See http://nobelprize.org/economics/laureates/
5
We sometimes refer to ABCT as the Hayekian cycle theory since it is often referred to in that way in the
literature. There is some justification for this nomenclature since Hayek (1931) was the first to use the
triangle. But Mises (1912) should perhaps be given priority in this regard since he published on this topic
earlier, albeit without utilizing the triangle to illustrate the theory. Garrison (2001) also deserves mention in
this regard since he was the first to “tilt” the triangle on its side, placing time on the horizontal axis, and
value on the vertical. Young (2005) follows the practice of Garrison, not Hayek.
3
Contrary to those who believe that the boom-bust cycle is natural to a market economy,
Ludwig von Mises (1912) argued that government intervention was the ultimate cause of
business cycles. Specifically, the government would attempt to curry popularity with the
voters by an easy money policy that kept interest rates below their “natural” levels. In
the Austrian view, interest rates are crucial market phenomenon that helps coordinate the
intertemporal structure of production. Free-market rates of interest indicate the tradeoffs,
on the margin, between the enhanced output that would be possible if more “roundabout”
methods of production were chosen, versus the loss in utility (time preference) that would
result from a further delay in consumption.
Within this context, Mises (and later Hayek, 1931) argued that artificially low interest
rates give false signals to entrepreneurs. Acting on the basis of faulty information, firm
owners behave as if more savings are available in the economy (since an increase in
genuine savings would also push down free-market interest rates). Consequently firms
would hire more workers and expand production capacity due to the lower rates. A
period of prosperity would ensue.
Unfortunately, this growth is illusory, according to Mises. Since the supply of savings
really hadn’t increased, there are insufficient capital goods and natural resources to
complete all of the new projects initiated by the overly optimistic business owners.
When they realize their errors, the bust ensues, with workers being laid off and
intermediate goods being sold off to other lines of production. In the long-run, the only
way to avoid the bust, according to the Austrians, is to avoid the preceding (artificial)
boom period.
In this context, Young’s critique analyzes the sensitivity of employment to changes in the
interest rate. Because he believes his data support a low estimate, Young concludes that
the Austrian story cannot be the major cause of business cycles in the real world.
III. A Summary of Young’s Model, and Its Internal Difficulties
In Young’s model, a firm’s desired amount of job reallocation in any period t is given by:
Ri*,t  F[| rt 1  rt 2 |, Wi ,t ],
(1)
where r is the Federal Funds (FF) rate and W is a set of unobservable variables relevant to
the decision. (All equation numbers are consistent with Young’s paper.) We note that
the firm’s optimal decision in time t is based on the lag of the change in the fed funds
rate.
Because adjustments take time, actual adjustment need not equal desired adjustment. The
actual reallocation at any time t is given by:
Ri ,t  Ri ,t 1  Ri ,t   ( Ri*,t  Ri ,t 1 ).
(2)
4
Young assumes that 0< λ<1, where λ is a speed of adjustment parameter. This author
then approximates F as linear and combines his equations (1) and (2) to yield, after
rearrangement:
Ri ,t  (1   ) Ri ,t 1  a (| rt 1 |)  b (Wi ,t ).
(4)
Next Young assumes that changes in W are white noise, allowing him to write:
Ri ,t  1 Ri ,t 1   2 (| rt 1 |)   i  vi ,t ,
(6)
where δi=bλ(W0), vi,t=bλ(εi,t), β1=(1-λ), and β2=aλ.
Finally, Young aggregates each such equation for the individual firm into an industry
equation:
R j ,t  1 R j ,t 1   2 (| rt 1 |)   j  v j ,t ,
(7)
where every j variable is the summation of i=1, 2, 3, …, I, where I is the number of firms
in industry j. With data for j=1, 2, 3,…, N such industries, equation (7) “represents a
dynamic panel data model of reallocation” (Young p. 278).
Even at this early stage of the analysis, we have objections. First, why is the lag in the
change in the fed funds rate taken to be “the latest signal available to firms”? Under
Young’s assumption, at any time t a firm can only observe the FF rate through rt-1. This
seems odd, especially since the data Young uses are quarterly. Moreover, this isn’t
simply an ad hoc way to capture the time needed for desired reallocation to take place;
Young explicitly models this as the speed of adjustment parameter, λ. In other words,
Young’s model explicitly takes into account the distinction between desired reallocation
and actual reallocation. Presumably desired reallocation in any quarter could depend on
interest rates in that quarter itself, and so we cannot see how Young can justify a lag in
this particular independent variable.
Our second objection concerns the aggregation process, from individual firms to
industries. When Young moves from his equation (6) to (7), he carries through the β2
term without changing it. This is a mistake: Because the lag in (the absolute value of)
the difference in fed funds rates is the same for all firms and in the industry-wide
regression, the coefficient on this term in the industry-wide regression is being inflated by
a factor of I (the number of firms in the industry). For example, if a given shift in interest
rates coincides with (on average) a reallocation of 100 jobs in each firm, and if there are
50 firms in the industry, then the same interest rate change will coincide with an increase
of 5,000 jobs at the industry level. It is true that in this particular application the problem
is minor, since Young only works with industry data and uses the firm-level equations
5
only to motivate his later specification.6 However, we raise the point in case Young (or
others) pursues this line of research and in particular tries to tie macro events with
individual firm decisions.
The thorny issue of aggregation leads to yet another concern with Young’s modeling
approach. Specifically, by dealing only with the absolute value of changes in labor
allocation—that is, failing to distinguish between job growth versus decline—Young
opens himself up to a form of “double counting.”7 For example, suppose that one firm in
an industry gains market share at the expense of another; perhaps the former adds 500
jobs while the latter sheds 500. In Young’s first regression, the dependent variable is R
(total reallocation), which is “[j]ob creation plus job destruction” (p. 279). It seems then
that Young’s regression would attempt to relate the interest rate to this “change” of 1,000
jobs in the industry in question, even though these particular reallocations (by
themselves) wouldn’t affect total industry employment. Since what Young is ultimately
testing is Hayek’s theory of interindustry employment shifts, we feel he should have
explicitly addressed this potential concern. (For example, if some industries have higher
intraindustry job movement—for whatever reason—then this could distort a test of the
Hayekian theory.) Unfortunately, Young makes no mention of this possible danger.
We finally come to the most serious objection in our immanent critique. Simply put, we
believe Young has reported parameter estimates that are significantly erroneous! With
total reallocation R as the dependent variable, and with the first lag of R and the first lag
of (the absolute value of) the change in fed funds as the two independent variables,
Young reports parameter estimates of 0.819 and 0.391, respectively. This leads him to
conclude that “a standard deviation change in the federal funds (positive or negative) is
associated with a change in desired reallocation of about 11% of its standard deviation”
(p. 280). This is presumably a blow against the Mises-Hayek cycle theory, because “the
contribution of monetary policy [on job reallocation] is modest” (p. 281).
In this context, it is clear that Young believes large estimates for the coefficient on the lag
of R, and low estimates for the coefficient on the lag in fed funds changes, undermine the
Misesian-Hayekian theory (ABCT). It is with some irony then that we report our own
estimates of 0.156 and 0.544, respectively!8 Thus, putting aside all of the other immanent
objections, this one alone demonstrably reverses Young’s pessimism regarding the
Austrian story. This is because, using our own estimates, it appears that the previous
period’s reallocation has far less influence than changes in the interest rate. Thus, if
anything Young’s model bolsters ABCT (assuming that our parameter estimates are
correct and Young’s are wrong, and that it is logically coherent to test economic
theories).
6
In private correspondence (July 23, 2006) Young has admitted that this is a mistake but considers it akin
to a “typo,” since writing a different label for β2 would eliminate the problem.
7
We are grateful to Paul Dower for originally voicing this concern.
8
We ran a LSDV fixed effects panel data regression on the Excel data provided by Young himself (via
email, dated June 7, 2006). Thanks again are due to Paul Dower, who independently ran the same data
using a different software program, and achieved point estimates identical to ours.
6
IV. Even if Done Properly, Young’s Model a Poor Test of ABCT
In light of the fact that our own parameter estimates, given Young’s framework, if
anything seem to support Hayekian cycle theory, we are tempted to end our critique at
this point and declare ABCT vindicated by the data. However, that would be a
strategically risky move, because a different econometric test in the future might yield the
opposite result. Therefore in this section we discuss broader difficulties with Young’s
approach, and argue that even if we disregarded the serious problems discussed in section
III, Young’s model wasn’t a very good “test” of ABCT in the first place.
For the Austrian economist, the most fundamental problem we encounter is that apodictic
economic theories cannot be empirically tested at all. Rather, they are aspects of
praxeology,9 and, as such, can only be illustrated, not tested. For praxeologists, economic
theories are equivalent to mathematical or geometrical claims, such as the Pythagorean
Theorem, or to the assertion that the three angles of a triangle sum to 180 degrees. One
can indeed illustrate such declarations, but it would be meaningless to even attempt to
test them. Young (p. 2) in this regard points to the “unwillingness of the ‘Austrian
school’ to pursue econometrics.” But Austrians10 are by no means “unwilling” to pursue
econometrics; there are numerous praxeological publications that utilize such statistical
techniques. All of them, however, illustrate, but do not purport to test, economic law.11
Moreover, ABCT takes seriously the complexity and heterogeneity of real world. Thus it
is a mistake to think that one can illustrate, much less “test,” ABCT with a model, such as
Young’s, that attempts to capture resource (specifically, labor) reallocations with a
single-equation for each industry; this is simply inadequate to the task. That is, when all
is said and done, after his simple, algebraic manipulations, Young ends up estimating a
single equation regression model (Young, 278). And, that equation takes the form of a
dependent variable being regressed against the dependent variable, i.e., itself, lagged one
period, the difference between the FF rate lagged one period and the FF rate lagged two
periods, and a dummy variable representing: “a set of unobservables relevant to the
firm’s decisions.” (We did not see a constant term in the model.) At best this is a very
simplistic model.
This inadequacy is magnified by the fact that Young uses one, single, proxy variable –
the FF rate − to capture [the essence of] the Federal Reserve’s “monetary policy’s
stance.” Austrian economists do not accept models based on “the” interest rate, or “the”
9
Block, 1973; Batemarco, 1985; Fox, 1992; Hoppe, date, 1989, 1992, 1995; Mises, 1969, 1998; Rizzo,
1979; Rothbard, 1951, 1957, 1971, 1973, 1976, 1997a, 1997b, 1997c, 1997d, 1993; Selgin, 1988.
We are puzzled by the quote marks around “Austrian school.” It would appear to imply that there is no
such school of thought as the Austrian. But this is surely mistaken. See on this Rockwell, 1995.
10
11
See on this Callahan and Garrison, 2003; Carilli and Dempster, unpublished; Gallaway and Vedder,
1992; Keeler, 2001; Mulligan, 2002, 2005, 2006; Powell, 2002.
7
real wage, etc. Also, monetary policy is reflected not only in interest rates, but also in
other terms and conditions of credit.12 Even to capture the interest rate part more
accurately would require some sort of measure of the term structure that would take into
account not only “the” level, however measured, but also the intertemporal and risk
spreads. Furthermore, even if the stance of monetary policy could be captured by one
proxy variable, this one indicator would not be the Fed Funds rate – an overnight,
virtually-riskless rate available only to depository institutions. The rates that affect realworld allocations of resources are those that include premiums for default risk and
inflation premiums, and involve many different terms-to-maturity. Such rates are notinfrequently quoted as (varying) spreads over a riskless (US Treasury) rate, but also over
the three-month dollar LIBOR rate.
Additionally, there is no basis in anything other than mathematics and statistics for
assuming the validity of a single variable W, said to be “a set of unobservable variables
relevant to ‘the’ firm’s decision” (Young, 277). Moreover, he (Young, 278) later
assumes: “This requires assuming that W0 to be the same set of variables for all firms in
an industry.” It is true that he allows the magnitude of W to differ among firms in the
same industry by an error term - εit, but his assumption is that every firm; i.e., each
decision maker, bases his decision on the same set of (unobservable) variables. (He does
not say so, but we presume that Young assumes the error term is a random variable,
distributed N(0, σ2).) Entrepreneurship is central to Austrian Economics, including
ABCT, and this assumption of his effectively removes any entrepreneurial decision
making from the process. That is, save for a random variation, captured by the error
term, all firms in an industry make the same decisions. In effect, we have a
“representative firm,” which renders the mathematics and statistics tractable, but finesses
the truly intractable problems of aggregation.13
Again, there is no basis in anything other than mathematics and statistics for:
“Approximating F as linear…” (Young, 278). Nor, yet again, is there any basis in
anything other than mathematics and statistics for the assumption: “Changes in W are
assumed to be white noise…” (Young, 278).
Austrian economics, as all good economics should be, is forward looking. True we may
learn from the past, but all decisions must necessarily be forward looking. Unless
entrepreneurs consider the future to be a pure extrapolation from the past, they will not
base their decisions entirely on past data, save as it informs their expectations as to the
future. Thus Young’s model is problematical in the same sense that Friedman’s laggedexpectations model14 was. Moreover, although it is true that, as noted in this present
As we write, there is much talk of “quantitative easing,” by the Federal Reserve in an attempt to increase
the quantity of credit.
12
13
This constitutes further mainstream avoidance of the complexity and heterogeneity of the real world. For
more on this theme, see Anderson, 2001, 2002; Barnett, 2003A, 2003B, 2004; Callahan, 2001; Herbener,
1996; Jablecki, 2007; Leoni and Frola, 1977; Mises, 1977, 1998; Murphy, 2008; Rizzo, 1979; Rothbard,
1988, 1993; Shostak, 2002
14
See, for example, Friedman (1970), eq. 35 and accompanying text, p.228.
8
rejoinder, Young’s model implies that decisions in period (quarter) t are made based on
FF rates from period t-1 and earlier, and that more current FF rates are available to the
entrepreneur, nevertheless, even that is insufficient as, as noted above, for several
reasons: the market for FF, and thus the FF rate, is not available to non-depository
institutions; entrepreneurs are not concerned normally with overnight rates, riskless or
risky; and, current rates, FF or others, are historical, and only future rates are relevant.
Beyond these major methodological difficulties, Young’s attempt to empirically test
ABCT fails on several grounds. First, all of Young’s chosen industries come from the
manufacturing sector. Yet this is the single worst sector to pick. As Young’s own
Hayekian diagrams15 indicate (p. 277), central bank-induced expansions have the smallest
effect in the middle of the triangle. Thus, Friedrich Hayek himself might have hoped for
Young’s reported results, so long as he supplemented them with similar studies in, say,
mining and retail sectors that showed a larger impact of the Federal Funds (FF) rate on
reallocation.
A second problem is that Young’s empirical data is limited to that concerning job
reallocations subject to Fed oriented changes in the interest rate. He thus in his figure 2
implicitly makes an assumption concerning size of the labor force in the various stages of
production; to wit, his assumption is that there would be a perfect correlation between
size of the labor force in each of the various stages of production, and the value, or
contribution to GDP, of any given stage of production. To say the least, this need not be
the case. If it is not, Young’s calculations are to that extent rendered inaccurate. Let us
put this into other words. Young is “testing” for number of workers. His observations are
numbers of jobs. If the number of employees was proportional to the stages in his two
sets of diagrams, well and good, at least insofar as this present point is concerned. For,
then, if a given stage gained (lost) job slots, we could unambiguously infer that that stage
had increased (decreased) in size. But this does not at all apply if the number of workers
in a given stage is not proportional to the value it contributes. And, there is simply no
reason to suppose that this is correct. For example, in the earlier stages, heavy industry,
there are fewer workers per value created and more capital equipment. In the later stages,
think retail, there are more employees, and less valued capital. To put this in yet other
words, the vertical axis in his diagrams represents value of product at any given stage.
Value of product, not number of laborers involved. But his statistical work is couched in
terms of number of employees, not value of product. Thus there is a disconnect.
The loss of signs (both in job reallocation and interest rate changes) through taking
absolute values is another defect of the Young test. It would have been far more
appropriate—and perfectly tractable with the data available—to examine whether interest
rate increases were positively correlated with job creation in late stages (wholesale,
retail) and with job destruction in early stages (mining).
15
Young (2005) quite properly utilizes the Hayekian triangle to illustrate ABCT. This has been the well
established practice since 1931 within the Austrian tradition. However, Barnett and Block (2006B) have
severely criticized this procedure.
9
On a related note, it is not clear why Young places so much emphasis on the distinction
between the amount of current reallocation due to the previous period’s adjustment,
versus the previous period’s change in monetary policy. After all, the Austrian School is
(in)famous for its insistence on the roundabout structure of production, and the fact that
policy changes may take years to reap their undesirable consequences. For example,
Austrian economist Murray Rothbard (1963) blames the Great Depression on the
unsustainable credit expansion throughout the 1920s, and not—as a monetarist might—
on particular Fed blunders during 1929 and the early 1930s.16 In this light, a Hayekian
might look at Young’s reported results and ask, “So what?” If Young had found that job
reallocation is influenced more by tax policy (say) than by interest rate policy, that would
have constituted a more legitimate objection to the Hayekian story.17 But the fact that a
given period’s job reallocation is due more to the previous period’s reallocation, versus
the previous change in the fed funds rate, by itself is fairly neutral. Again, this is because
the initial fed funds change—which may have occurred, say, years in the past—might
take time to work its way through the capital structure. In a simple econometric test such
as Young’s, where only the previous period’s interest rate change is included, this
situation might very well generate parameter estimates similar to the ones that Young
reports.
The previous two objections concern timing and sign changes, and manifest themselves
in yet another difficulty: Young uses the quarters 1972:2 to 1993:4; he does so
apparently only because Davis (1996) did so. But this start and end point are irrelevant to
ABCT, which focuses on a single boom-bust cycle. This time period chosen by Young
includes several business cycles. His conclusions are to that extent problematic. Within
the ABCT, the effect of a given change in interest rates may be different, depending on
the stage of the business cycle. If unemployment is virtually absent, and an unsustainable
boom has been underway for several years, reductions in the interest rate might have very
little influence on the number of workers in a given sector. Rather than looking at
consecutive years, Young would have done better to look at either a single boom (or bust)
period, or several booms, or several busts.
We also point out that the Hayekian theory is quite subtle, and admittedly does not lend
itself to easy empirical tests. Although this may be regrettable, one should be aware of
the fact when interpreting an analysis such as Young’s. Consider a concrete example of
this difficulty. In the standard Hayekian analysis, if the interest rate is originally at the
“natural” level, a Fed cut will cause an unsustainable expansion, an artificial boom
period. Now even if the Fed continually pumps in additional credit to keep rates
artificially low, this illusory prosperity cannot continue forever. At some point there will
not be enough physical capital goods to complete all of the various projects underway,
16
On this see also Robbins (1934); Garrison (2001, 2004); Rothbard (1993)
17
This would be far from a fatal objection, however. ABCT merely asserts that its story is true; it would be
improper to interpret it as claiming that the phenomena it discusses need always and ever be the most
important explanations of a business downturn. War, for example, or a gigantic natural disaster, might well
account more for a bust than a slight tinkering with interest rates.
10
and some firms will need to shut down and fire their workers. This ultimate reckoning
would not necessarily be tied to a “policy innovation” on the part of the Federal Reserve,
which indeed might be cutting rates even further in an attempt to postpone the bust. As
this scenario—which is entirely consistent with ABCT—demonstrates, one might find
massive reductions in employment without prior increases in the interest rate. In any
case, this level of amalgamation, economy wide, is incompatible with ABCT. For there,
there will always be some stages that gain employment even during the depression (the
higher stages), and others that lose job slots even during the boom (again, the higher
stages).
Finally consider the 20 industries used in the study (see appendix). The first few, (20-23)
may be categorized as consumer goods. Young speaks in terms of a focus on
manufacturing, but then lists these four clearly non-manufacturing categories. These four
(20-23) are represented most reasonably, in Young’s figure 1, as “retailing” or “later
stages.” The remainder are almost certainly to be characterized as either producers’
goods, or consumer durables. Thus, they would most properly be placed in his “earlier
stages.”18 While some of these 20 can possibly be associated, specifically, with various of
the stages depicted in his figures, others of them clearly transcend two or more or even all
of these stages. They cannot all be considered members of the “manufacturing” stages in
his figure 1.
There are grave problems here, insofar as ABCT is concerned. For this theory does not at
all speak in terms of total employment. Rather, the Hayekian model emphasizes
employment or value or capital at specific stages. The boom shifts resources of all types
and varieties, certainly including labor, to the earlier stages. In terms of employment of
labor, jobs during the initial phase of the business cycle increase in the earlier stages, but
decrease in the later ones. In the bust, yes, job slots decline in the early stages, but they
increase in the later ones. Young’s is a “test” of total employment. ABCT knows of no
such aggregation. Thus, even if all of our other criticisms of Young were for some reason
discounted, this one alone would be fatal to his thesis.
Let us draw but one more implication from this point. Consider the two triangles in
Young’s figure 2. Let us focus on that stage depicted in the triangle where the two
hypotenuses cross. There, employment remains constant, regardless of whether we are
moving from boom to bust or in the opposite direction. But this is the only section of the
triangle for which this is true.
The “job reallocation data” he used was based entirely on the manufacturing sector of
the economy, in which employment was falling during the period of the study from
about 26% to about 14% of the total employment. The following is the abstract of the
paper Young (279) cites as describing the data.
“We rely on a decomposition of employment changes into job creation and job
destruction components - and a novel set of identifying restrictions that this
18
For ambiguities in representing consumer durables in the ABCT triangle, see Barnett and Block, 2006B.
11
decomposition permits - to develop new evidence about the driving forces behind
aggregate fluctuations and the channels through which they operate. We implement our
approach to identification using quarterly postwar U.S. data on oil shocks, monetary
shocks, and manufacturing rates of job creation and destruction. Our analysis delivers
many inferences: 1) The data favor a many- shock characterization of fluctuations in
employment and job reallocation, 2) Theories of employment fluctuations that attribute a
predominant role to aggregate shocks must in order to fit the data involve
contemporaneous effects of such shocks on job destruction that are at least as large as the
effects on job creation, 3) Theories in which aggregate shocks primarily affect the first
moment of the cross-sectional density of employment growth imply that allocative
shocks have bigger contemporaneous effects on destruction than on creation and, hence,
that allocative shocks reduce aggregate employment, 4) Allocative shocks drive most
fluctuations in the intensity of job reallocation, 5) Oil shocks drive employment
fluctuations through a mixture of allocative and aggregate channels, 6) Monetary shocks
trigger job creation and destruction dynamics that fit the profile of an aggregate shock.”
Because (Young, 279) states: “Job reallocation data is quarterly…” we assume that the
FF rate data is also. This raises the issue of whether the (St. Louis Federal Reserve
Bank’s data base “Federal Reserve Economic Data” commonly referred to as FRED) data
used was that for the target rate or that for the effective rate. As they differ, regularly, it
would seem that the appropriate one is the actual (effective) rate. Moreover, as can be
seen at this url: http://research.stlouisfed.org/fred2/categories/118, the target rate series
goes back only to 9/27/1982 and is a daily series, whereas the effective-rate series, daily,
weekly, and monthly, can be found as far back as July, 1954. Then the question arises
which of these series did he use, daily, weekly, or monthly? Monthly seems the most
logical, but, regardless, the actual data used must have been an average (no matter
whether the Fed’s or Young’s) of daily figures. The FRED data comes from the Board of
Governor’s Statisatical Release H.1519 and is “a weighted average of rates on brokered
trades.”20 A quick look at the relevant FRED graphs or data show that the daily figures
varied, sometimes considerably, during the various quarterly periods. What have such
averages of historical data to do with entrepreneurial decisions? Nothing.
Young does not discuss the dummy variable at all; as W0 is the same for all firms in an
industry (a truly heroic assumption when dealing with ABCT, as the role of the
individual entrepreneur is central to all of Austrian economics), we must assume that it
varies among industries. On what basis are the values of W0 assigned to each industry?
The dependent variable, “job reallocation data” (Young, 279), is meant to capture the
effect of reallocation of resources; i.e., changes in the structure of production, as per
ABCT. As ABCT is a “capital based” theory, it would seem more appropriate to use
data, if such exists, on “capital goods reallocation.” Moreover, it is quite possible that
changes in the structure of production could occur without “job (or even capital goods)
reallocation.”
19
See “Source(s),” http://research.stlouisfed.org/fred2/series/FF?cid=118.
20
See footnote one, http://www.federalreserve.gov/releases/h15/update/.
12
Finally, let us consider the matter of units/dimensions as is appropriate for any work
using applied mathematics (Barnett, 2004). Young (279) says that the “jobs reallocation
data” is “percents of total jobs.” That is, the dependent variable is a percentage or pure
number; i.e., something such as 100 jobs/1,000 jobs = 0.1. (We assume that to be the
form as we have not access to the NBER paper cited by Young as describing the data.)
The first problem with such a percentage is it assumes that all jobs are homogeneous, an
assumption that flies directly in the face of Austrian economics and, indeed, of all
common sense. However, assuming arguendo, that such a percentage is meaningful,
because the dependent variable is such a percentage, that necessarily means that the first
regressor, the lagged dependent variable and the second regressor, the difference
between the FF rate lagged one period and the FF period lagged two periods, present no
problem re dimensions, as they are also in percentages, and thus require no nonsensical
assumptions re the relevant parameters. However, this is not so: the dummy regressor is
different.
First, what, pray, is the dimension of W? Unless each of the “set of unobservables” that
constitute W is measured in percentage terms and then aggregated in some way, e.g., a
weighted average, to yield W as a percentage, or they are first aggregated and then the
aggregation used to create a percentage, W must be dimensioned. (It is hard to imagine
that the relevant data re each of the variables in the “set of unobservables relevant to the
firm’s decisions” are percentages thereof. Moreover, it is most likely that some of these
are subjective in nature and thus not grist for the mill of cardinal mathematics.)
Moreover, what the dimension of a “set of unobservables” might be is an interesting
question. Assuming some or all are dimensioned, but not all have the same dimension,
how would they be meaningfully aggregated so that the dimension of W was consistent
with the different dimensions of its various elements? However, again, assuming
arguendo, that the elements of W could be meaningfully aggregated into W with its own
dimension consistent with those of its elements, it would then be necessary to see what is
required to insure that δ = bλ((W) a pure number. From Young’s eq. 2 we know λ is a
pure number. Thus, for δ to be a pure number the dimension of b must be the inverse of
that of W. Moreover, as there is a different W for each industry, there is no reason to
think that the elements of the different Ws are the same from industry to industry. That
would mean that as the dimension of W varied among industries, so, also would the
dimensions of b have to vary. To conclude this point, we doubt that Young’s model
would look sensible if it were to pass the test of dimensional analysis.
V. Conclusion
We welcome Young’s econometric test of the Austrian Business Cycle Theory, which in
our view is one of Mises and Hayek’s tragically underappreciated contributions to the
profession. Having said that, we unfortunately find that Young does a poor job in testing
his own model. In particular, we find parameter estimates that would lead to the exact
opposite conclusion from Young’s. Finally, we find that Young’s model was never a
proper test of ABCT in the first place. Contrary to a widespread view, Austrian
economists do not fear empirical illustrations of their theories. They merely insist that
13
they be fair and correct. For that matter, we have no objection to neoclassical economists
“testing” Austrian theories either; such practices are compatible with their own
methodology, if not ours. But, again, fairness and correctness should be the order of the
day.
References:
Anderson, William. 2001. “The Limits of Mathematics” November 29
http://www.mises.org/story/839
Anderson, William. 2002. “Mathematics and Economic Analysis.” April 02;
http://www.mises.org/story/926
Barnett, William II. 2003A. “The Modern Theory of Consumer Behavior: Ordinal or
Cardinal?” The Quarterly Journal of Austrian Economics. 6 (1): 41 − 65;
http://www.qjae.org/journals/qjae/pdf/qjae6_1_3.pdf
Barnett, William II. 2003B. "Dimensions and Economics: Some Problems" The
Quarterly Journal of Austrian Economics. 6 (3): 27 − 46;
http://www.qjae.org/journals/qjae/pdf/qjae6_3_2.pdf
Barnett II, W. 2004. “Dimensions and Economics: Some Problems.” Quarterly Journal
of Austrian Economics. Spring. Vol. 7, No. 1, pp. 95-104;
http://www.mises.org/journals/qjae/pdf/qjae6_3_2.pdf
Barnett, William II and Walter Block. 2005. “Professor Tullock on Austrian Business
Cycle Theory,” Advances in Austrian Economics, Vol. 8, pp. 431-443
Barnett, William II and Walter Block. 2006A. "On Gallaway and Vedder on Stabilization
Policy" Quarterly Journal of Austrian Economics. Vol. 9, No. 1, spring, pp. 57-81
Barnett, William II and Walter Block. 2006B. “On Hayekian Triangles.” Procesos De
Mercado: Revista Europea De Economia Politica; Vol. III, No. 2, Fall, pp. 39-141;
http://tinyurl.com/2zkvj7; http://mises.org/journals/scholar/block18.pdf
Batemarco, Robert. 1985. "Positive Economics and Praxeology: The Clash of Prediction
and Explanation," Atlantic Economic Journal, July, 13(2), pp. 31-27.
Block, Walter. 1973. "A Comment on 'The Extraordinary Claim of Praxeology' by
Professor Gutierrez," Theory and Decision, June, 3(4), pp.377-387.
Block, Walter and Kenneth Garschina. 1996. "Hayek, Business Cycles and Fractional
Reserve Banking: Continuing the De-Homogenization Process," Review of Austrian
Economics, Vol. 9, No. 1, pp. 77-94;
http://www.mises.org/journals/rae/pdf/rae9_1_3.pdf;
http://www.mises.org/journals/rae/pdf/r91_3.pdf
14
Block, Walter. 2001. “Yes, We Have No Chaff: A Reply to Wagner’s “Austrian Business
Cycle Theory: Saving the Wheat While Discarding the Chaff,” Quarterly Journal of
Austrian Economics, Vol. 4, No. 1, Spring, pp. 63-73.
//www.mises.org/journals/qjae/pdf/qjae4_1_4.pdf
Callahan, Gene. 2001. “Logical Economics vs. Mathematical Economics ” February 17;
http://www.mises.org/story/616
Callahan, Gene; and Garrison, Roger W. 2003. “Does Austrian Business Cycle Theory
Help Explain the Dot-com Boom and Bust?” Quarterly Journal of Austrian Economics
6(2): 67-98.
Carilli, Anthony M. and Gregory M. Dempster. 2001. “Expectations in Austrian Business
Cycle Theory: An Application of the Prisoner’s Dilemma.” Review of Austrian
Economics. Vol. 14, No. 4, pp. 319-330;
http://www.gmu.edu/rae/archives/VOL14_4_2001/4_carilli&dempster.pdf
Carilli, Anthony M., Gregory M. Dempster and Henrik Rasmussen. Unpublished. “Is
Austrian Business Cycle Theory Still Relevant?”
Cochran, John P., Call, Steven T. and Glahe, Fred R. 2003. “Austrian Business Cycle
Theory: Variations on a Theme.” Quarterly Journal of Austrian Economics 6(1): 67-74.
Cwik, Paul. 1998. “The Recession of 1990: a Comment.” Quarterly Journal of Austrian
Economics 1(3): 85-88.
Davis, S.J. & Haltiwanger, J.C. and Schuh, S. 1996. Job Creation and Destruction. MIT
Press
Fox, Glenn. 1992. "The Pricing of Environmental Goods: A Praxeological Critique of
Contingent Valuation," Cultural Dynamics, Vol. V, No. 3, pp. 245-259
Friedman, Milton. 1970. “A Theoretical Framework for Monetary Analysis.” The
Journal of Political Economy. 78 (2): 193-238.
Gallaway, Lowell, and Vedder, Richard. 1992. Out of Work: Unemployment and
Government in 20th Century America. New York: Holmes and Meier
Garrison, Roger W. 1994."Hayekian Triangles and Beyond," in Jack Birner and Rudy
van Zijp, eds., Hayek, Coordination and Evolution: His Legacy in Philosophy, Politics,
Economics, and the History of Ideas. London: Routledge, pp. 109-125.
Garrison, Roger W. 2001. Time and Money: The Macroeconomics of Capital Structure.
London: Routledge; an important contribution to Austrian business cycle theory
15
Garrison, Roger W. 2004. “Overconsumption and Forced Saving in the Mises-Hayek
Theory of the Business Cycle” History of Political Economy vol. 36, no. 2 (summer)
Garrison, Roger W. and Don Bellante. 1988. "Phillips Curves and Hayekian Triangles:
Two Perspectives on Monetary Dynamics," History of Political Economy, vol. 20,
no. 2 (Summer), pp. 207-234.
Hayek, Friedrich A. 1931. Prices and Production, London: Routledge.
Herbener, Jeffrey M. 1996. "Calculation and the Question of Mathematics," Review of
Austrian Economics, 9(1), pp. 151-162
Hoppe, Hans-Hermann. 1989. "In Defense of Extreme Rationalism: Thoughts on Donald
McClosky's The Rhetoric of Economics," Review of Austrian Economics, 3, pp. 179-214.
Hoppe, Hans-Hermann. 1991. "Austrian Rationalism in the Age of the Decline of
Positivism," Journal des Economistes et des Etudes Humaines, Vol.2, No. 2; reprinted as
Hoppe, Hans-Hermann. 1992. "On Praxeology and the Praxeological Foundation of
Epistemology and Ethics," Herbener, J., ed., The Meaning of Ludwig von Mises, Boston:
Dordrecht
Hoppe, Hans-Hermann. 1994. "Austrian Rationalism in the Age of the Decline of
Positivism," in Austrian Economics: Perspectives on the Past and Prospects for the
Future, Vol. 17, Richard M. Ebeling, ed., Hillsdale, MI: Hillsdale College Press, pp. 5996.
Hoppe, Hans-Hermann. 1995. Economic Science and the Austrian Method. Auburn, AL:
Ludwig von Mises Institute
Hoppe, Hans-Hermann. 2006. "Austrian Rationalism in the Age of the Decline of
Positivism" in The Economics and Ethics of Private Property: Studies in Political
Economy and Philosophy, 2nd, ed., pp. 347-379; Auburn AL: The Mises Institute
Jablecki , Juliusz. 2007. “How Many Traders Can You Fit into a Model? August 23;
http://www.mises.org/story/2646
Keeler, James P. 2001. Empirical Evidence on the Austrian Business Cycle Theory.”
Review of Austrian Economics. Vol. 14, No. 4, pp. 331-351
Leoni, Bruno, and Frola, Eugenio. 1977. "On mathematical thinking in economics," The
Journal of Libertarian Studies, Vol. 1, No. 2, Spring, pp. 101-110
Mises, Ludwig von. [1912] 1953. The Theory of Money and Credit [originally published
in German in 1912]. New Haven, CT: Yale University Press.
16
Mises, Ludwig von. 1969. Theory and History: An Interpretation of Social and Economic
Evolution. New Rochelle, NY: Arlington House.
Mises, Ludwig von. 1977. "Comments about the mathematical treatment of economic
problems," The Journal of Libertarian Studies, Vol. 1, No. 2, Spring, pp. 97-100
Mises, Ludwig von. [1949] 1998. Human Action, Scholars’ Edition. Auburn: Mises
Institute. http://www.mises.org/humanaction/pdf/HumanActionScholars.pdf
Mulligan, Robert. 2002. "A Hayekian Analysis of the Term Structure of Production,"
Quarterly Journal of Austrian Economics 5(2): 17-33
Mulligan, Robert.2005. "The Austrian Business Cycle: a Vector Error-correction Model
with Commercial and Industrial Loans," Journal of Private Enterprise Fall, 22(1): 51-91
Mulligan, Robert. 2006. "An Empirical Examination of Austrian Business Cycle
Theory," Quarterly Journal of Austrian Economics 9(2): 69-93
Murphy, Robert P. 2008. “Austrian Realists.” July 17; http://mises.org/story/3028
Powell, Benjamin. 2002. “Explaining Japan's Recession.” Quarterly Journal of Austrian
Economics 5(2): 35-50.
Rizzo, Mario. 1979. "Praxeology and Econometrics: A Critique of Positivist Economics,"
New Directions in Austrian Economics, Louis Spadaro, ed., Kansas City: Sheed Andrews
and McMeel, pp. 40-56
Robbins, Lionel C. 1934. The Great Depression. London: The Macmillan Co.
Rockwell, Llewellyn H., Jr. 1995. Why Austrian Economics Matters. Auburn AL: The
Mises Institute; http://www.mises.org/etexts/why_ae.asp
Rothbard, Murray N. 1951. “Praxeology: Reply to Mr. Schuller.” American Economic
Review, December.
Rothbard, Murray N. 1957. "In Defense of Extreme Apriorism," Southern Economic
Journal, January, 23(1), pp. 314-320.
Rothbard, M. 1963. America's Great Depression. Kansas City: Sheed Andrews.
http://www.mises.org/rothbard/agd.pdf
Rothbard, Murray N. 1971. “Lange, Mises and Praxeology: The Retreat from Marxism.”
Toward Liberty. Vol. II, Menlo Park, CA: Institute for Humane Studies, pp. 307-321.
Reprinted in The Logic of Action One: Method, Money, and the Austrian School. Glos,
UK: Edward Elgar Publishing Ltd., 1997, pp. 384-396.
17
Rothbard, Murray N. 1973. "Praxeology and the Method of Economics," Phenomenology
and the Social Sciences, M. Natanson, ed., Evanston, IL: Northwestern University Press,
vol. 2, p. 311 342; and Austrian Economics: A Reader Vol. 18, Richard M. Ebeling, ed.,
Hillsdale, MI.: Hillsdale College Press, 1991, pp. 55-91.
Rothbard, Murray N. 1976. "Praxeology: The Methodology of Austrian Economics," in
The Foundations of Modern Austrian Economics. Edwin G. Dolan, ed., Kansas City:
Sheed and Ward, pp. 19-39;
http://www.econlib.org/library/NPDBooks/Dolan/dlnFMAContents.html
Rothbard, Murray N. 1988. “Chaos Theory: Destroying Mathematical Economics from
Within?” The Free Market. Vol. VI, No. 3;
http://www.mises.org/freemarket_detail.asp?control=296
Rothbard, Murray N. 1993. Man, Economy, and State, 2 vols., Auburn, AL: Ludwig von
Mises Institute
Rothbard, Murray N. 1997a. "Praxeology as the Method of the Social Sciences," in The
Logic of Action One. Murray N. Rothbard, ed., UK: Edward Elgar Publishing Limited,
pp.28-57.
Rothbard, Murray N. 1997b. "Praxeology: The Methodology of Austrian Economics," in
The Logic of Action One. Murray N. Rothbard, ed., UK: Edward Elgar Publishing
Limited, pp.58-77.
Rothbard, Murray N. 1997c. "Praxeology, Value Judgments, and Public Policy," in The
Logic of Action One. Murray N. Rothbard, ed., UK: Edward Elgar Publishing Limited,
pp.78-99.
Rothbard, Murray N. 1997d. "In Defense of 'Extreme Apriorism'," in The Logic of Action
One. Murray N. Rothbard, ed., UK: Edward Elgar Publishing Limited, pp.100-108.
Rothbard, Murray N. 1993. Man, Economy, and State, 2 vols., Auburn, AL: Ludwig von
Mises Institute
Sechrest, Larry J. 2001. “Capital, Credit, and the Medium-run.” Quarterly Journal of
Austrian Economics 4(3): 63-77.
Selgin, George A. 1988. "Praxeology and Understanding: An Analysis of the
Controversy in Austrian Economics," Review of Austrian Economics, (2), pp. 19-58; and
Praxeology and Understanding, Auburn, AL: Ludwig von Mises Institute, 1990.
Shostak, Frank. 2002. “What is wrong with econometrics?” April 17.
http://www.mises.org/story/940
18
Young, Andrew T. 2005. “Reallocating labor to initiate changes in capital structures:
Hayek revisited.” Economics Letters. Vol. 89, No. 3, pp. 275-282
Appendix
Here are the 20 industries used in the study:
20 = Food
21 = Tobacco
22 = Textile
23 = Apparel
24 = Lumber
25 = Furniture
26 = Paper
27 = Printing
28 = Chemicals
29 = Petroleum
30 = Rubber
31 = Leather
32 = Stone, Clay, and Glass
33 = Primary Metals
34 = Fabricated Metals
35 = Nonelectric Machinery
36 = Electric Machinery
37 = Transportation
38 = Instruments
39 = Miscellaneous
Source: Davis, et al, 1996, chapter 3; see also
http://www.econ.umd.edu/~haltiwan/download/QUARTER/RZI2BD.DOC
19