Documentation for the replication files for the following papers:

Stewart, K.G., and J.C.H. Jones (1998) "Hedonics and demand
analysis: the implicit demand for player attributes," Economic
Inquiry 36, 192-202.

Ferguson, D.G., Jones, J.C.H., and K.G. Stewart (2000) "Competition
within a cartel: League conduct and team conduct in the market for
baseball player services," Review of Economics and Statistics 82(3)
422-430. (This paper has a subsequent correctional note.)

Stewart, K.G. (2009) "Nonjointness and scope economies in the multiproduct
symmetric generalized McFadden cost function," Journal of Productivity Analysis
32(3), 161-171.

Stewart, K.G., and J.C.H. Jones (2010) "Are sports teams multiproduct firms?"
Empirical Economics 39(2), 487-514.

DATA FILES (for data sources see the published papers)

Team.dat    team characteristics; documented in Team.txt

City.dat    city characteristics; documented in City.txt

HedPrice1.dat   Hedonic prices estimated from data on all players on
                the team for whom salary data was available
HedPrice2.dat   Hedonic prices estimated from data on players with at
                least 3 years experience in MLB, for whom salary data
                was available.

HedPrice1.dat and HedPrice2.dat are the files of hedonic prices on
player characteristics that are used in the estimation of the demand
system. Both have the same format consisting of 156 lines
(6 seasons x 26 teams); the columns are:

p1          hedonic price for experience
p2          hedonic price for batting skill
p3          hedonic price for pitching skill
p4          hedonic price for star status
nobs        number of observations (players) in the salary regression
            used to estimate the hedonic prices

The hedonic prices were obtained from estimated salary equations
using a large data set on individual players; this data set remains
proprietary and so has not been publicly posted.

HedPrice1.dat and HedPrice2.dat differ only in that HedPrice2.dat was
estimated using data on "experienced" players; thus nobs for
HedPrice2.dat is always less than that for HedPrice1.dat. This was
done at the request of the REStat referees.

Thus the estimation results for the translog systems over the full
sample given in the EI and REStat papers differ numerically, although
not substantively, for 2 reasons:
(1) the hedonic prices are slightly different; and
(2) the REStat results are for the homothetic version of the demand
system, in order that no output measure need be specified, consistent
with the broader theoretical context of that paper.

COMMAND FILES

GL.tsp Estimates the Generalized Leontief hedonic factor demand
            system for baseball player characteristics reported
            in the EI article.
            User must edit output to equal WINS or ATTEND, as needed.
            Uses data files: Team.dat, HedPrice1.dat

TL1.tsp     Estimates the non-homothetic translog hedonic factor
            demand system for baseball player characteristics reported
            in the EI article.
            User must edit output to equal WINS or ATTEND, as needed.
            Uses data files: Team.dat, HedPrice1.dat

TL2.tsp     Estimates the homothetic translog hedonic factor
            demand system for baseball player characteristics reported
            in the REStat article, including subsidiary results such
            as the Hausman tests.
            Uses data files: Team.dat, City.dat, Hedprice2.dat
            Remarks: City.dat includes the instruments used in the
            Hausman tests

REStat.tsp  Estimates the 3-equation model of the REStat article.
            Uses data files: Team.dat, City.dat

The following command files replicate the results of SJ (2009) and
Stewart (2009) for the multiproduct symmetric generalized McFadden
hedonic factor demand system.
All use the data files Team.dat, Hedprice1.dat

1. SGMhausman1.tsp      Hausman tests (Table 1 of SJ) implemented
   SGMhausman2.tsp      in single-output models M1, M2, M3.
   SGMhausman3.tsp      Use additional data file City.dat

2. MSGMm0mx.tsp         Tests single-output models M1,M2 as restricted
                        versions of two-output model M0 (Table 2 of SJ)

3. MSGMseparability.tsp Tests separability
                        (Table 2 of SJ and Table 1 of Stewart)

4. MSGMvuong.tsp        Vuong selection tests of one- and two-output
                        models, and model selection criteria
                        (Tables 3 and 4 of SJ)

5. MSGMcost1.sha        Separability not imposed.
   MSGMcost2.sha        Separability imposed.
   These printouts yield fitted cost shares and their means; predicted
   cost shares and their standard errors (Table 5 of SJ).
   Values are used to plot isocost curves (Figure 1 of SJ).

6. MSGMdemandelas.sha   Demand elasticities and their standard errors
                        (Table 6 of SJ)

7. MSGMmorishima.sha    Morishima elasticities and their standard errors
                        (Table 6 of SJ)

8. MSGMinputelas.sha    Factor input elasticities with respect to output
                        (Table 7 of SJ)

9. MSGMcostelas.sha     Cost elasticities (Table 8 of SJ)

10. MSGMpsrts.sha       Product-specific returns to scale
                        (Table 8 of SJ)

11. MSGMscope.sha       Scope economies (Section 6 of SJ)

12. MSGMnonjointness.sha    Table 2 of Stewart

Notes:

TSP is used for command files that estimate the non-concavity-constrained model.
Shazam is used for command files that estimate the concavity-constrained model,
as only it would yield convergence in this case. For the same reason, Shazam
is used for MSGMnonjointness.sha because it estimates a normalization of the
model (normalization N2 of Stewart) for which TSP did not yield convergence.