Contributions to Economic Analysis &
Policy
Volume 5, Issue 1
2006
Article 17
Measuring the Effect of Multimarket Contact
on Competition: Evidence from Mergers
Following Radio Broadcast Ownership
Deregulation
Joel Waldfogel∗
∗
†
Julie Wulf†
Wharton School, University of Pennsylvania and NBER, waldfogj@wharton.upenn.edu
Wharton School, University of Pennsylvania, wulf@wharton.upenn.edu
c
Copyright 2006
The Berkeley Electronic Press. All rights reserved.
Measuring the Effect of Multimarket Contact
on Competition: Evidence from Mergers
Following Radio Broadcast Ownership
Deregulation∗
Joel Waldfogel and Julie Wulf
Abstract
This paper examines the effects of multimarket contact on advertising prices in the U.S. radio
broadcasting industry. While it is in general difficult to measure the effect of multimarket contact
on competition, the 1996 Telecommunications Act substantially relaxed local radio ownership
restrictions, giving rise to a major and exogenous consolidation wave. Between the years of 1995
to 1998, the average extent of multimarket contact in major U.S. media markets increased by 2.5
times. Importantly, the extent of change in multimarket contact varies across markets, and the
change in multimarket contact varies separately from the change in concentration. Using a panel
data set on 248 geographic U.S. radio broadcast markets, 1995-1998, we find that multimarket
contact has little effect on advertising prices. This paper contributes to the empirical literature
on multimarket contact by analyzing a different industrial context and using longitudinal data
surrounding an ownership deregulation.
∗
Corresponding author: Julie Wulf, Wharton School, University of Pennsylvania, 2023 SteinbergDietrich Hall, Philadelphia, PA 19104
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
1
Industrial economists and strategy researchers have long thought that multimarket
contact (MMC) might soften competitive behavior. Consequently, considerable
effort has been devoted to documenting the effects of multimarket contact on
competition, especially since Bernheim and Whinston’s formal characterization
(1990). A large group of cross sectional studies finds competition to be weaker in
markets whose participants also interact in other markets. A consensus has
emerged, based on this evidence, that multimarket contact reduces competition.1
Yet, measurement of the effect of multimarket contact requires exogenous
variation in multimarket contact that the cross-sectional context cannot plausibly
provide. Recognizing this, researchers in industrial organization and strategy
have looked to longitudinal data surrounding deregulatory episodes to measure
effects of multimarket contact on competition. Notably, Evans and Kessides
(1994) and Gimeno and Woo (1996, 1999) examine the relationship between the
changes in prices and changes in multimarket contact surrounding airline
deregulation, finding that markets with increased multimarket contact also
experience increased prices.
In this study, we make use of longitudinal data surrounding a major
ownership deregulation in radio broadcasting to revisit the question of whether
multimarket contact affects competition. The effect found in the airline studies
might be unique to the structure of that industry or specific to the deregulation of
airlines. Multiple studies in other industrial contexts using longitudinal data are
needed to determine the generality of this effect. We examine the effects of
changes in multimarket contact and ownership concentration on radio advertising
prices in the period surrounding the 1996 Telecommunications Act. While firms
had previously been limited to owning two stations per market and 40 stations
nationally, under the Act, they could own up to 8 stations in each of the largest
markets and an unlimited number nationally. The Act set off a major
consolidation wave, increasing local ownership concentration by more than 30%
over the four-year period. Beyond that, the average number of markets in which
the ten largest firms operated doubled from less than 15 to more than 30,
increasing multimarket contact significantly. In 1995, the two largest owners in
each of 248 markets interacted in 1.27 markets on average. By 1998, this
measure had increased to over 3—2.5 times the level in 1995. While this sharp
increase in multimarket contact was accompanied by an increase in local
ownership concentration—itself an important potential determinant of
1
Jayachandran, Gimeno and Varadarajan (1999) and Korn and Baum (1999) summarize the results
of the research on multimarket contact and competition. While the results are somewhat mixed,
the majority of the studies suggest that firms compete less aggressively (using various measures
such as pricing, entry/exit rates, profitability) in markets in which they jointly compete.
Combining the list of papers from both summaries, 22 of the 27 studies find that multimarket
contact reduces competition.
Contributions to Economic Analysis & Policy
2
Vol. 5 [2006], No. 1, Article 17
competitive conditions—the increases in multimarket contact and concentration
are not so collinear to preclude measurement of distinct effects. Over this period,
radio advertising prices rose substantially, by about 25%, although this price
increase seems to have originated outside of the radio advertising market. Similar
ad price increases occurred in other media markets. (See note 8 below.)
Not only were there exogenous changes in concentration and multimarket
contact, but because the Act relaxed local ownership limits by different amounts
in different-sized markets, it gave rise to exogenously different amounts of
variation in multimarket contact and concentration across markets. The structure
of ownership deregulation suggests an instrumental variables strategy for
empirically documenting the effect of multimarket contact on competition. The
relationship between market size and ownership limits is nonlinear by regulatory
fiat; it can be modeled with multiple terms in market size—for example,
population and its square—which provide two instruments for the two
endogenous variables.
Using a variety of empirical approaches, we find that neither multimarket
contact nor concentration had a robust effect on advertising prices in radio
broadcast markets in the U.S. during the years surrounding the
Telecommunications Act. These findings contribute to the empirical literature on
multimarket contact by analyzing a different industrial context and using
longitudinal data surrounding an ownership deregulation.
The paper proceeds in 3 parts. Section 1 presents background from
existing empirical and theoretical literature. Section 2 describes the data used in
the study, as well as the Telecommunications Act and the empirical strategy it
allows. Section 3 presents results. A brief conclusion follows.
BACKGROUND: THEORY AND EXISTING EMPIRICAL EVIDENCE
Most research on tacit collusion among firms has been in the context of singleproduct firms. The possibility that multiproduct firms recognize their rivals and
interactions across markets was first raised in 1955 by Corwin Edwards: the
“multiplicity of their contacts may blunt the edge of competition.” Mutual
forbearance—that is, competitors adopting “a live-and-let-live policy designed to
stabilize the whole structure of the competitive relationship”—was a concept that
was discussed and explored in several empirical studies.2 In 1990, Bernheim and
Whinston developed a theoretical model of multimarket contact and collusive
behavior that accelerated interest in the topic, spurring research that evaluated the
relationship between multimarket contact and the degree of competition. In what
2
Quotes from Scherer, 1980, p.340. For a list of empirical papers examining the effects of
multimarket contact on competition, see Korn and Baum (1999) and Jayachandran, Gimeno and
Varadarajan (1999).
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
3
follows, we first describe the intuition behind Bernheim and Whinston’s model
and then highlight the results of select empirical studies on multimarket contact.
In a game-theoretic analysis of multimarket contact, Bernheim and
Whinston establish that interdependencies between firms across markets can
facilitate tacit cooperation in otherwise competitive markets. The intuition is that
collusive outcomes are easier to sustain with multimarket contact because there is
more scope for punishing deviations in any one market. For example, when firms
compete in two markets, they recognize that any deviation from collusive
behavior in either market will be met with punishment in both markets.
In the formal modeling, multimarket contact serves to pool the firm’s
incentive constraints across markets. The incentive constraints are expressions
that compare the firm’s payoffs from deviating today to the payoffs from
cooperating over future periods. For example, in a duopoly market where each
firm receives half of monopoly payoffs, the payoffs from cooperating over future
periods may far exceed the payoffs from deviating (i.e. there is slack in the
constraint) and firms will cooperate. However, in a competitive market in which
firms share payoffs among many players, the opposite holds (i.e. payoffs from
cooperation are less than that from deviation) and firms do not cooperate. As this
two-market example illustrates, in pooling the constraints in both markets, the
slack in the constraint of the duopoly market relaxes the competitive market’s
binding constraint. Multimarket contact facilitates collusive behavior in the more
competitive market since the firms can shift the ‘slack enforcement power’ or the
ability to collude in the duopoly market to the more competitive market. Hence,
we would expect to see higher prices in otherwise competitive markets whose
participants have contact in other less competitive markets.
This formal characterization of multimarket contact has led to accelerated
interest in the topic by both strategy researchers and industrial economists alike
and, while the empirical evidence is somewhat mixed, the majority of studies find
that multimarket contact weakens competition. Price levels, profitability, and the
stability of incumbents are higher in markets with greater multimarket contact,
typically measured by the number of markets in which participants compete. For
example, based on a sample of 187 local banking areas in 1972, Heggestad and
Rhoades (1978) find greater market share stability of large bank holding
companies in local markets with higher multimarket contact. Baum and Korn
(1996) using event history analysis find lower entry and exit rates over the period
1979 to 1984 in California airline routes with higher multimarket contact. In
industries with higher multimarket contact, Scott (1982) finds higher profits
among U.S. manufacturers based on a sample of 437 firms in 1974. Fewer
studies find either no effect or the opposite effect of multimarket contact on
competitive behavior. For example, Rhoades and Heggestad (1985) find no
relationship between operating profits, service charges, or loan rates and fees and
Contributions to Economic Analysis & Policy
4
Vol. 5 [2006], No. 1, Article 17
multimarket contact using a larger sample of banks than the earlier study. Mester
(1987) finds lower operating profits among 171 California savings and loans in
markets with higher multimarket contact.
Identification of the effect of multimarket contact on competition is
inherently difficult because pricing, market concentration, and multimarket
contact all emerge as endogenous outcomes of market processes. While the cross
sectional approach in earlier studies may identify correlations between
multimarket contact and competition, the causal relationship is unknown.3 A few,
more recent empirical studies use longitudinal techniques to analyze the causal
relationship and find that increases in multimarket contact weaken competition.
Evans and Kessides (1994) find that major airlines during the period from 1984 to
1988 raise fares more on routes where average multimarket contact among all
competitors increases faster. During a similar period, Gimeno and Woo (1996,
1999) find that major airlines earn increasing yields on routes where their average
multimarket contact with competitors is growing more quickly. These studies
rely on the deregulation of the airline industry as a source of exogenous variation
to identify the effects of multimarket contact on airline fares and yields.4 Similar
to these studies, we exploit an exogenous change in regulation—the relaxation of
radio station ownership limits across varying-sized geographic markets—to
evaluate the effect of changes in concentration and multimarket contact on
advertising prices.5
DATA, MEASURES, AND EMPIRICAL STRATEGY
DATA
The data for the study are drawn from a radio station-level dataset that includes
revenue estimates, owner identity, programming format, and audience size for
3
An important endogeneity problem in this setting is due to unobserved heterogeneity across
markets. For example, a positive market shock may increase market price and induce large firms
to enter a market. Entry by large firms (that operate in many markets) increases multimarket
contact. Hence, we observe a positive correlation between market price and multimarket contact.
In this example, omission of market demand in the price regression leads to overstating the effect
of multimarket contact on market prices. Panel regressions using market fixed effects address
unobserved heterogeneity across markets.
4
Baum and Korn (1996) use longitudinal data on airlines over the period 1979 to 1984 and find
that multimarket contact lessens interfirm rivalry as measured by exit and entry (not prices).
5
While the common intuition—codified into the Merger Guidelines—is that increased
concentration raises prices, competition in the industry could operate differently. Stations
compete for listeners in part through their choice of the quantity of ads to air. Competition would
tend to reduce ad quantities and, via the market demand curve, to raise prices with ambiguous
effects on revenue overall. The actual effect of concentration on revenue per listener (our measure
of prices) remains an empirical question.
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
5
every station in 268 U.S. markets in 1995-1998. Our measure of listening is
average quarter hour (AQH) listening, the number of persons listening for at least
five minutes during an average quarter hour period. Most of the data come from
BIA Research (Media Access Pro). We obtain market level AQH listening for
Spring of each year from Duncan American Radio. The underlying dataset covers
5888 commercial stations and 2268 owners in 1998. Table 1a shows the number
of stations increasing slightly from 1995 to 1998 and the number of owners
nationwide declining by more than 25%. We compute measures of price,
multimarket contact, and ownership concentration at the market level for much of
the analysis.
MEASURES
Our study requires measures of prices and their determinants including ownership
concentration within markets and multimarket contact by firms across markets.6
Here we describe the calculation of these measures. We calculate each market’s
advertising price as the aggregate revenue across stations divided by the sum of
listeners across the same stations.7 We measure ownership concentration based
on the listening share of the largest firms in a market. Because the top 4 firms
generally account for 65-85 percent of radio listening in a market, we calculate
the collective share of the largest 2, 4 and 6 firms in the market as measures of
ownership concentration. In addition, we calculate the Herfindahl index (hhi) of
station listening by market. Table 1b shows the market averages of revenue per
listener and several measures of ownership concentration by year. Revenue per
listener has increased by 25% and concentration (measured by hhi) has increased
by 32% over the four-year period. Spot market prices increased by 18% over the
three-year period from 1995 to 1997. 8
6
In this study, we use market level data which are based on actual information about market level
revenue rather than estimates of firm revenue.
7
This measure has the shortcoming that, in principle, annual station revenue= (price per
ad/listener) * (# of listeners) * (# of ads). Since we don’t observe the number of ads, our “price”
actually equals the product of two expressions we cannot disentangle. That is, our “price”=(price
per ad/listener) * (# of ads). Balancing this disadvantage are two advantages. First, the revenue
data, while estimates for individual stations, are calculated to audited totals for many markets.
(See Duncan (1993)). The second advantage of these data is that they are available for all major
stations. Clearly, there is a problem if the revenue data reflect a formula rather than actual prices.
To address this, we also estimate regressions using cities in which revenue data are audited by an
accounting firm (either Hungerford or Miller Kaplan). Approximately 100 of the 268 markets
were audited by an accounting firm. (Refer to discussion in Section 3a.)
8
The magnitude of the increase in revenue per listener is consistent with price increases in both
radio and television advertising as documented in a Morgan Stanley Dean Witter (1999) research
report. Over the same time period, the report documents increases in radio ad prices (measured by
CPM or cost-per-thousand-listeners) of 32.1% in network radio and 23.4% in spot radio ads. The
Contributions to Economic Analysis & Policy
6
Vol. 5 [2006], No. 1, Article 17
The calculation of multimarket contact in a market is more complicated.
Table 2 shows the amount of multimarket contact between pairs of firms in
geographic markets in 1998. The table includes the largest ten firms measured by
the number of stations owned in 1998. (The diagonal elements show the instances
of multimarket contact with the own-firm or, more directly, the number of
markets of operation for the firm—Clear Channel Communications owned radio
stations in 80 geographic markets, while Capstar Broadcasting Partners operated
in 62 markets.) Zeroes in this table indicate that a firm pair has no geographic
overlap. Clear Channel overlaps with each of the other largest firms (i.e.
multimarket contact measure greater than zero for all values in column 1), yet
Cumulus Media shares markets with only two of the largest firms (with mainly
zeroes in column 3). A value of one indicates that the firms have stations
coexisting in a single market. Values of 2 or more literally reflect multimarket
contact. Clear Channel overlaps in more than fifteen markets with four of the
largest firms. Slightly more than one-half of the forty-five firm pairs represented
in Table 2 have some market overlap (i.e. multimarket contact greater than or
equal to 2), while about one-third have no geographic overlap.
How does one quantify the amount of multimarket contact in a market? A
market consists of N participants that may encounter each other in different
markets, in addition to this one. It seems natural to consider the pairs of
participants in a market, then to average the multimarket contact across these
pairs. Thus, if there are N firms in the market, one measure of multimarket
contact is the average of the number of contacts across all markets between N(N1)/2 pairs of firms. A practical problem with this “simple average” approach is
that because ownership within markets is fairly concentrated, firms are not
symmetric. Computing the average multimarket contact over all firms, including
the “fringe” of firms that may own only one station nation-wide, might give
undue weight to firms that are unimportant in price-setting. To address this
concern we calculate the amount of multimarket contact between the top 2, 4 and
6 participants in each market (based on listening share for each firm’s local
collective stations). (Note that the numbers of firm-pairs among the top 4 and 6
firms in each market are 6 and 15, respectively.)
The following example illustrates this measure of multimarket contact. In
Philadelphia, the top four firms based on listening share in 1998 are Greater
report documents even higher increases in television ad prices: 37.7% in network television and
46.9% in spot television ads. Source: Morgan Stanley Dean Witter (March 3, 1999), “Ad Hoc:
Media Pricing Trends—A Look Under the Hood.” Table 1, pg. 3. We take averages of
men/women categories for both radio and television and broadcast / spot categories for television.
The average spot prices in Table 1b come from SQAD Spot Radio and represent CPM (cost per
thousand) estimates for Arbitron Metro Survey Areas (MSAs).
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
7
Media, Chancellor Media Corporation, Evergreen Media Corporation, and
Beasley Broadcast Group. There are 6 possible firm-pairs among these firms and
each pair meet in a number of markets nationwide including the Philadelphia
market. Hence, our multimarket contact measure between the top 4 firms in
Philadelphia equals the average of the total number of markets in which each
firm-pair meets. Based on this method of calculation, Table 1b shows that
average multimarket contact for all markets almost tripled over the period (2firm) from 1.27 to 3.17.
EMPIRICAL STRATEGY
The Telecommunications Act of 1996 substantially relaxed both local and
national ownership restrictions. Prior to the Act, a firm could own up to 40
stations nationally and no more than 3 in each market.9 Under the Act a firm
could own up to 8 stations in each of the largest markets and faced no national
ownership limits. As has been documented elsewhere (Berry and Waldfogel,
2001), this policy shift unleashed a wave of consolidation that substantially
increased ownership concentration in local markets. The question here is whether
the policy change also gave rise to exogenous variation in the extent of
multimarket contact and, moreover, whether the change in multimarket contact is
separate from the change in concentration.
In 1995 the largest chain (the firm owning the most radio stations within
the sample markets) was Jacor Communications, with 52 stations in 23 city
markets (Table 3).10 A few other firms operated networks of similar reach.
American Radio Systems operated 45 stations in 18 city markets, and SFX
Broadcasting operated 37 stations in 17 city markets. By 1998 (only two years
after the Act), the largest firm operated over 400 stations in as many as 80 city
markets. While no firm operated in over 23 city markets in 1995, in 1998 six
firms operated in over 23 city markets.
It is clear from Table 3 that firms operated more stations per city market in
1998 relative to 1995 hence explaining the increase in local ownership
concentration. The fact that firms increased the number of markets of operation
9
Prior to September 1992, a firm could own up to 12 AM plus 12 FM stations nationally. On
September 16, 1992, the national ownership limits were increased to 18 AM plus 18 FM stations.
On September 16, 1994, the limits increased to 20 AM plus 20 FM stations. (Source: FCC)
10
While Jacor’s ownership of 52 stations appears to contradict the ownership rules then in effect, it
appears not to be an error. Other authoritative accounts, such as Noam (2004) write that Jacor
owned 54 stations during 1995 (see http://www.vii.org/papers/medconc.htm, accessed October 24,
2005). And FCC documents indicate that some stations “owned” by Jacor were actually
separately owned but operated jointly with Jacor properties during 1995 (see
http://www.fcc.gov/Bureaus/Mass_Media/Orders/1996/fcc96380.txt, accessed October 24, 2005).
Contributions to Economic Analysis & Policy
8
Vol. 5 [2006], No. 1, Article 17
also suggests that incidence of multimarket contact between firms—as when two
firms both operate in a number of markets—would increase.
Tables 4 and 5 illustrate the dimensions of exogenous variation that we
will exploit to measure the impact of multimarket contact and ownership
concentration on market prices. The ownership restrictions were relaxed to
different extents in markets of different sizes, thus giving rise to different extents
of increase in both multimarket contact and ownership concentration across
markets of different sizes. Thus, we can address the possible endogeneity of
multimarket contact and concentration measures by using two terms in market
size, measured by either population or by the number of stations in a market prior
to the changes in ownership restrictions.
Table 4 presents averages of both multimarket contact and concentration
ratios of the two largest firms by 1998 population quartiles in 1995 and 1998. A
number of things are apparent from this table. First, consider the statistics in the
lower part of the table. While larger markets are less concentrated, the increase in
concentration over the period is the greatest in the largest markets. In 1995,
concentration is higher in smaller markets than in larger markets [0.60 in the
smallest market (Q1) vs. 0.38 in the largest market (Q4)]. Second, concentration
in all markets increased between 1995 and 1998 by 24%, and the increase is
proportionately larger in larger markets (45% for Q4 vs. 12% for Q1). In contrast
with the concentration measures in Table 4, multimarket contact is higher in
larger markets [1.9 in the largest market (Q4) vs. 1.0 in the smallest market (Q1)].
Multimarket contact increases dramatically between 1995 and 1998 (139%
overall), with sharp increases of more than 3.5 times in the larger population
markets (265% for Q4).
Table 5 shows the local limits before and after the Act, by market size
measured by the number of stations in the market. The patterns are qualitatively
similar to those presented in Table 4. That is, while markets with more stations
are less concentrated, the increase in concentration over the period is the greatest
in the markets with the most stations. Moreover, multimarket contact is greater in
markets with more stations and the increase in multimarket contact is
proportionately larger in those markets. We propose to use the variation in
multimarket contact and ownership concentration that followed the Act to
measure the effects of multimarket contact and ownership concentration on
competition.
We have two endogenous variables, multimarket contact and
concentration, and one conceptual instrument, market size (measured by either
population or the number of stations). But the ownership limits were relaxed to
different extents in markets of different sizes, and a review of Tables 4 and 5
shows that the changes in multimarket contact and concentration bear different
nonlinear relationships with market size. Hence, more than one term in market
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
9
size, such as population and its square, can serve as plausible instruments for both
multimarket contact and concentration.
REGRESSION RESULTS
LONGITUDINAL SPECIFICATIONS
Fixed effects regressions using longitudinal data overcome the biases in OLS
estimates that arise from fixed unobserved heterogeneity. In this sense, our
approach goes beyond many of the earlier studies based on cross-sectional results.
We begin by estimating market fixed effects regressions using pooled data for all
years and the 2-firm measures of multimarket contact and concentration. In the
first column of Table 6, we regress the logarithm of average revenue per listener
in a market (i.e. our measure of price) on 2-firm measures of multimarket contact
and concentration.11 We include year effects and report robust standard errors
clustered by market. We find a positive and significant coefficient on
multimarket contact and a positive, but insignificant coefficient on concentration.
As mentioned earlier, one concern might be that the revenue data just
reflect a formula and that we are learning about the relation between multimarket
contact, concentration, and this formula and not about the relation with actual
prices. To address this concern, in Table 6 column (2), we estimate the baseline
regression, but use only those city markets (approximately 100) in which revenue
data are audited by an accounting firm (either Hungerford or Miller Kaplan). We
find similar results as reported in column (1) based on all markets. Importantly,
the coefficient on 2-firm multimarket contact (0.0033) is very close in magnitude
in comparison to the estimate based on all markets (0.0037), but is weakly
significant at the 10% level. The coefficient on concentration is positive and
larger in magnitude in comparison to column (1), but is imprecisely estimated.
In Table 6 columns (3) and (4), we replicate the regressions in columns (1)
and (2), but use the 4-firm measures of multimarket contact and concentration.
Both coefficients on multimarket contact are positive and larger in magnitude in
comparison to those based on the 2-firm measure, but neither coefficient is
statistically significant. In both regressions, we find negative and insignificant
coefficients on concentration. When we replicate these regressions using the 6firm measures of multimarket contact and concentration, the multimarket
coefficients are qualitatively similar to those based on the 4-firm measure
(unreported).
11
This semi-log specification is the same as that estimated in Evans and Kessides (1994) and we
report these results to allow a direct comparison to their findings.
Contributions to Economic Analysis & Policy
10
Vol. 5 [2006], No. 1, Article 17
The range of multimarket contact coefficients in Table 6 is between
0.0033 (for the 2-firm measure) and 0.0117 (for the 4-firm measure). How large
are these coefficients? Recall from Table 1 that the 2-firm multimarket contact
measure increases from 1.27 in 1995 to 3.17 in 1998, while the 4-firm measure
increases from 1.24 to 2.08. Multiplying the 0.0037 (coefficient on multimarket
contact in column (1) that is significant at the 1% level) by the change in 2-firm
multimarket contact yields a 0.70 percent change in prices attributable to the
change in multimarket contact. The analogous exercise for the 4-firm measure in
column (4) yields a 0.98 percent increase in price. Only the estimates in columns
(1) & (2) are significantly different from zero with one estimate being only
weakly significant. Nevertheless, one could infer that there is an effect of
multimarket contact on prices. However, it is quite small.
ROBUSTNESS CHECKS
Next we explore the robustness of the relatively small effect of multimarket
contact on radio prices as documented in Table 6. While fixed effects control for
time-invariant market effects and we cluster standard errors by market, we may
still be understating standard errors due to serially correlated residuals. In Table 7
column (1), we address this by adjusting standard errors for serial correlation
[AR(1)] and find that the coefficient on multimarket contact remains positive, but
is no longer significant. Also, the size of the coefficient is less than half that in
our baseline regression (Table 6 column 1).
Of course, this difference might be attributable to the smaller sample,
since we lose the first year in the AR(1) specification. In column (2), we repeat
the baseline regression on the sub-sample that excludes 1995 observations and
find that the coefficient on multimarket contact is positive, slightly larger, but
remains insignificant. As expected, the standard errors in the AR(1) specification
are slightly larger than the clustered standard errors in column (2) (however, only
those associated with the multimarket contact coefficient). We find qualitatively
similar results when we estimate regressions using only audited markets
(unreported). Thus we find that the small and significant positive coefficient on
multimarket contact in our baseline regression is not robust to adjusting standard
errors for serial correlation. Yet, since we lose 1995 in these analyses, this may
be simply due to the inability of capturing large changes in both prices and
multimarket during a critical year.
Of potentially greater concern than incorrect inference due to understated
standard errors, our fixed effects results may be influenced by time trends in
advertising prices that vary by market. To evaluate this, in Table 7 column (3), we
return to the sample based on all four years and control for market-specific trends
by adding an interaction term between city-market effects and a trend variable.
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
11
We find a positive coefficient on multimarket contact and a negative coefficient
on concentration, but neither coefficient is statistically significant. Once again,
the positive and statistically significant coefficient on multimarket contact that is
reported in our baseline regression (Table 6 column 1) is not robust when we
control for market-specific trends. Moreover, we find qualitatively similar results
when we estimate regressions using only audited markets (unreported).
The small but significant effect of multimarket contact on prices
documented in our baseline regression is not robust to either adjustments for serial
correlated residuals or to controls for market-specific trends. Next, we turn to
estimating instrumental variables regressions to acknowledge the potential
endogeneity of multimarket contact and ownership concentration.
INSTRUMENTAL VARIABLES SPECIFICATIONS
As discussed earlier, both multimarket contact and ownership concentration are
potentially endogenous variables and the extent of the change in each is also
potentially endogenous across markets. But it is also possible that the change in
multimarket contact over time is endogenous. If the change in multimarket
contact is related to the change in prices, then our fixed effects approaches will
yield biased estimates of the effect of within-market variation in multimarket
contact on prices. To address this, we estimate instrumental variables regressions
using nonlinear terms in two measures of market size, population and the number
of stations in a market prior to the changes in ownership restrictions.
The relationship between the variation in both endogenous variables and
each instrument has been previously documented in the tables of summary
statistics [market population in Table 4 and number of stations in a market (or
policy bands) in Table 5]. In the last two columns in Table 7, we report results of
our instrumental variables regressions. In column (4) we instrument multimarket
contact and concentration with 1995 population of the market (and its square). In
column (5), we instrument with the number of stations in a market (or policy
bands). The policy bands are defined as a series of dummy variables that represent
how the Act changed ownership limits in each market as illustrated in Table 5.
For example, if a market in 1995 had 14 or fewer stations, then the policy band
dummy equals one and zero otherwise. Analogous dummy variables are created
for two other categories: markets with 15 or more, but less than 30 stations; and
markets with more than 30 stations.12 In the regressions in columns (4) and (5),
none of the variables of interest are significantly different than zero.
12
population of a market in 1995, and
and ( population 2 ) i
where populationi is
it is a dummy which is 1 after the Telecommunications Act
In particular, our instruments are populationi
it
Contributions to Economic Analysis & Policy
it
12
Vol. 5 [2006], No. 1, Article 17
COMPARISON TO PREVIOUS LONGITUDINAL STUDIES
Previous research has estimated the effect of multimarket contact on price
competition. Studies based on longitudinal data that exploit exogenous variation
surrounding a deregulation are generally limited to the airlines industry.13 These
studies find the multimarket contact has a positive effect on prices that is both
statistically and economically significant. For example, Evans and Kessides use
the exogenous variation in multimarket contact caused by airline deregulation and
find that increases in multimarket contact from the 25th to the 75th percentile
results in an increase of over 5% in airline prices.
By contrast, based on a similar empirical strategy in a different industrial
context, we find weak evidence of an effect of multimarket contact on price
competition. In this paper, we report the results of semi-log regressions that are
based on similar specifications to the regression results reported in Evans and
Kessides (hereafter EK). As mentioned earlier, we find a small economic effect
of multimarket contact on price—the growth in multimarket contact between
1995 and 1998 explains a 0.7 percent increase in average revenue per listener. In
addition to being small, this effect is not statistically robust to: (a) adjusting
standard errors for serial correlation or (b) controlling for market-specific trends.
Finally, instrumenting multimarket contact and concentration measures using two
measures of market size (market population and policy bands) results in
coefficients that are not statistically different than zero. By contrast, EK in a
comparable regression of the logarithm of average route price on multimarket
contact find that an increase of one standard deviation in multimarket contact
leads to a 3.5% increase in prices with a t-statistic greater than 8.14 So, EK
estimate a larger and statistically significant effect of multimarket contact on
prices in airlines (3.5%) in comparison to our estimate in radio broadcasting (less
than 1%). To sum up, we find an economically small effect that is not robust to a
goes into effect. The policy band instrument is, analogously, the policy band dummies interacted
with it .
13
An exception to this is Gelfand and Spiller (1987) who examine multiproduct interactions in the
Uruguayan Banking Sector in the late 1970s surrounding a partial relaxation of legal entry
barriers.
14
This estimate comes from the Table V regression results in EK on page 359. In Section V-A of
the paper, EK address the potential for understated standard errors due to regressing airline prices
that vary within the route on measures of multimarket contact that vary only across routes. To
address this, they define their dependent variable as the logarithm of the average ticket price at the
route level which is comparable to our dependent variable—logarithm of average radio prices in a
city market. A one standard deviation increase in their measure of average route contact (0.08)
results in a 3.5% increase in average route price (0.08*0.443 coefficient on average route contact
from Table V, column 1).
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
13
reasonable set of analyses suggesting that the multimarket effect on radio prices is
not significantly different than zero.
One possible explanation for the differences between EK and our findings
is that the amount of exogenous variation in multimarket contact differs between
the samples. However, the changes in multimarket contact over the periods
studied are roughly comparable: multimarket contact in airlines doubles from
1984 to 1988 and it increases by 2.5 times in radio from 1995 to 1998. Another
possible econometric explanation might be due to differences in samples: EK
analyze only the 1000 largest city-pair routes, while we investigate all radio
markets.
More broadly, and possibly more likely, the results raise the question of
whether sector differences account for the different effect of multimarket contact
on competitive behavior. For example, airlines set prices for large numbers of
customers, and the prices are not only public information but they are also
observable at a relatively low cost. This may make it easier for competitors to
detect cheating and, in turn, strengthen incentives for tacit collusion. The number
of ad buyers in radio is small (compared to air travelers). Ad prices are negotiated
and not necessarily publicly known, even if rate cards are public. As a result it
may be more difficult to monitor behavior and maintain tacit collusion in radio
markets.
Another possible explanation is that joint ownership and the existence of
close substitutes may affect the ability of competitors to sustain collusion. While
there is joint ownership in multiple advertising media by radio firms (i.e. radio,
television and bulletin boards owned by a single firm), the same is not true in
transportation: airline firms are distinct from firms operating in railroad, bus, and
rental cars sectors. Finally, there may be differences in the deregulation of airlines
and radio that may account for the difference.
CONCLUSION
The consensus of previous empirical research across many industrial sectors
suggests that multimarket contact reduces rivalry among competitors. However,
the results are open to question because most of the studies are based on cross
sectional datasets, and measurement of the effect requires exogenous variation in
multimarket contact that the cross sectional context cannot plausibly provide.
More recent papers based on longitudinal studies surrounding deregulation in the
airline industry find support for the theory of mutual forbearance: increases in
multimarket contact between competitors serving the same airline routes lead to
higher prices. We argue that additional industrial settings should be explored
using longitudinal datasets.
Contributions to Economic Analysis & Policy
14
Vol. 5 [2006], No. 1, Article 17
The 1996 Telecommunications Act significantly relaxed ownership
restrictions in radio broadcasting leading to massive consolidation following the
Act. The extreme exogenous changes in multimarket contact combined with
well-defined geographic markets in radio broadcasting make it a desirable setting
to empirically measure the effect of multimarket contact on price competition.
We find a small economic effect of multimarket contact on radio advertising
prices surrounding the deregulation of radio ownership and this effect is not
robust to a reasonable set of analyses. Moreover, our estimates indicate that the
change in multimarket contact can explain very little (almost one thirty-fifth) of
the large increase in advertising prices between 1995 and 1998. Because the
concentration measures are significant in none of the fixed effects regressions,
there is no portion of the change in ad prices that we can attribute to increased
concentration. The findings in this paper suggest that the effect of multimarket
contact on competitive behavior may differ across industries.
Why do we find little effect of concentration or multimarket contact on
prices in radio advertising? One possibility is that radio advertising is not a
distinct market. That is, television advertising may be a close substitute to radio
advertising. A number of recent studies ask whether advertising on a particular
broadcast medium is in the same market as ads on other media (Ekelund, Ford
and Jackson, 1999; Silk, Klein and Berndt, 2001; and Bush, 2002). The findings
in these studies are mixed. Using 2001 BIA data to estimate own-price elasticities
and cross-price elasticities, Bush (2002) finds evidence suggesting that the local
radio advertising market is not distinct from local television or newspaper
advertising markets. By contrast, Ekelund et al. (1999) use 1995 BIA data and
estimate own-price elasticities that suggest radio advertising is a distinct market,
although their data do not allow them to distinguish between local and national
advertising. Our finding of little discernable effect of multimarket contact or
concentration on prices suggests that radio advertising is not a distinct market.
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
15
REFERENCES
Baum, Joel A. C. and Helaine J. Korn (1996), “Competitive Dynamics of
Interfirm Rivalry,” The Academy of Management Journal 39(2), pp. 255291.
Bernheim, Douglas and Michael D. Whinston (1990), “Multimarket Contact and
Collusive Behavior,” RAND Journal of Economics 21, pp. 1-26.
Berry, Steven T. and Joel Waldfogel (2001), “Do Mergers Increase Product
Variety? Evidence from Radio Broadcasting,” The Quarterly Journal of
Economics 116(3), pp. 1009-1025.
Bush, C. Anthony (2002), “On the Substitutability of Local Newspaper, Radio,
and Television Advertising in Local Business Sales,” Federal
Communications Commission, Media Bureau Staff Research Paper, pp.
2002-2010.
Duncan, James H. (1993), American Radio, Spring 1993, Indianapolis: Duncan’s
American Radio.
Ekelund, R.B. Jr., G.S. Ford and J.D. Jackson (1999), “Is Radio Advertising a
Distinct Local Market? An Empirical Analysis,” Review of Industrial
Organization 14, pp. 239-256.
Edwards, Corwin D. (1955), “Conglomerate Bigness as a Source of Power,” in
Business Concentration and Price Policy, National Bureau of Economic
Research Conference Report, Princeton, NJ: Princeton University Press,
pp. 331-352.
Evans, William N. and Ioannis Kessides (1994), “Living by the ‘Golden Rule’:
Multimarket Contact in the U.S. Airline Industry,” The Quarterly Journal
of Economics 109, pp. 341-366.
Federal Communications Commission (1996), Memorandum Opinion and Order,
Transfer of Control of Citicasters to Jacor, September 17, 1996.
Gelfand, Matthew D. and Pablo T. Spiller (1987), “Entry Barriers and
Multiproduct Oligopolies: Do They Forebear or Spoil?” International
Journal of Industrial Organization 5, pp. 101-113.
Contributions to Economic Analysis & Policy
16
Vol. 5 [2006], No. 1, Article 17
Gimeno, Javier and Carolyn Y. Woo (1996), “Hypercompetition in a Multimarket
Environment: The Role of Strategic Similarity and Multimarket Contact
on Competitive De-Escalation,” Organization Science 7, pp. 322-341.
________ and ________ (1999), “Multimarket Contact, Economies of Scope, and
Firm Performance,” Academy of Management Journal 42(3), pp. 239-259.
Heggestad, Arnold A. and Stephen A. Rhoades (1978), “Multimarket
Interdependence and Local Market Competition in Banking,” The Review
of Economics and Statistics 60, pp. 523-532.
Jayachandran, Satish, Javier Gimeno and P. Rajan Varadarajan (1999), “The
Theory of Multimarket Competition: A Synthesis and Implications for
Marketing Strategy,” Journal of Marketing 63, pp. 49-66.
Korn, Helaine J. and Joel A.C. Baum (1999), “Chance, Imitative, and Strategic
Antecedents to Multimarket Contact.” Academy of Management Journal
42(2), pp. 171-194.
Mester, Loretta J. (1987), “Multiple Market Contact Between Savings and Loan,”
Journal of Money, Credit, and Banking 19, pp. 538-549.
Morgan Stanley Dean Witter, Micheal Russell (1999), “Ad Hoc: Media Pricing
Trends—A Look Under the Hood,” March 3, 1999, Table 1, p. 3.
Noam, Eli M. (2004), “Media Concentration in the United States: Industry Trends
and Regulatory Responses,” Virtual Institute of Information—Industry
Report, Columbia Business School.
Rhoades, Stephen A. and Arnold A. Heggestad (1985), “Multimarket
Interdependence and Performance in Banking: Two Tests,” The Antitrust
Bulleting 30, pp. 975-995.
Scherer, F. M. (1980), Industrial Market Structure and Economic Performance,
Houghton Mifflin, Boston, MA.
Scott, John T. (1982), “Multimarket Contact and Economic Performance,” The
Review of Economics and Statistics 64, pp. 368-375.
Silk, A.J., L.R. Klein and E.R. Berndt (2001), “Intermedia Substitutibility and
Market Demand by National Advertisers,” NBER working paper 8624.
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
17
TABLE 1
DESCRIPTIVE STATISTICS
Panel A: Sample Statistics
1995
1996
1997
1998
% Change
No. of Stations
5794
5846
5771
5888
1.6
No. of Owners
3040
2773
2448
2268
<25.4>
Avg. No. of Stations Owned
1.8
2.1
2.4
2.6
44.4
Panel B: Averages by City Market (248 Markets)
Stations (Avg.)
Owners (Avg.)
Avg. Revenue/listener
Avg. Revenue/listener (101
audited markets)
Avg. Spot Prices
Concentration
2 Firm
4 Firm
Herfindahl (hhi)
Multimarket Contact
2 Firm
4 Firm
1995
1996
1997
1998
% Change
21.7
11.3
393
371
21.8
10.3
415
400
21.5
9.1
445
425
22.0
8.5
491
472
1.4
<24.7>
24.9
27.2
6.02
6.49
7.12
--
18.3 (95-97)
0.51
0.75
0.22
0.56
0.80
0.25
0.61
0.83
0.28
0.63
0.86
0.29
23.5
14.7
31.8
1.27
1.24
1.61
1.39
2.34
1.73
3.17
2.08
149.6
52.8
Contributions to Economic Analysis & Policy
18
Vol. 5 [2006], No. 1, Article 17
TABLE 2
MULTIMARKET CONTACT (MMC) FOR TEN LARGEST OWNERS (BY NUMBER OF STATIONS IN 1998)
Clear
Channel
Comm.
Capstar
Broadcast.
Partners
Cumulus
Media
LLC
Salem
Comm.
Corp.
CBS
Corp.
Chancellor
Media
Corp.
Citadel
Comm.
Corp.
Cox
Radio
Inc.
Sinclair
Comm.
Inc.
Entercom
Clear Channel
Comm.
80
16
3
18
17
17
8
9
7
2
Capstar Broadcast.
Partners
16
62
3
1
3
0
6
1
3
1
Cumulus Media LLC
3
3
30
0
0
0
0
0
0
0
Salem Comm. Corp.
18
1
0
27
13
18
1
2
1
4
CBS Corp.
17
3
0
13
26
10
0
2
3
6
Chancellor Media
Corp.
17
0
0
18
10
25
0
5
0
2
Citadel Comm. Corp.
8
6
0
1
0
0
22
0
1
0
Cox Radio Inc.
9
1
0
2
2
5
0
13
0
0
Sinclair Comm. Inc.
7
3
0
1
3
0
1
0
10
1
Entercom
2
1
0
4
6
2
0
0
1
8
Note: The main diagonal represents the number of city-markets in which the firm operates. The off-diagonal represents the number of citymarkets in which the firm-pair interact. A number greater than one in the off-diagonal represents multimarket contact for that firm-pair.
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
19
TABLE 3
NUMBER OF STATIONS AND NUMBER OF CITY-MARKETS
FOR TEN LARGEST FIRMS (BY NUMBER OF STATIONS IN 1998), 1995 - 1998
1995
Owner
1998
No. of
Stations
No. of
Owner
City-markets
No. of
Stations
No. of
City-markets
409
80
228
62
Jacor Communications
Incorporated
American Radio Systems
License Corp
SFX Broadcasting Inc
52
23
45
18
Clear Channel
Communications
Capstar Broadcasting Partners
37
17
Cumulus Media LLC
142
30
Evergreen Media Corp
35
14
Chancellor Media Corp.
124
25
Clear Channel
Communications
Salem Communications
Corporation
Benchmark
Communications
Group W Radio
33
11
104
22
29
19
Citadel Communications
Corporation
CBS Corporation
89
26
25
9
Cox Radio Inc
55
13
23
12
Sinclair Communications Inc.
50
10
Triathlon Broadcasting
23
7
44
27
River City Broadcasting
23
8
Salem Communications
Corporation
Entercom
37
8
Contributions to Economic Analysis & Policy
20
Vol. 5 [2006], No. 1, Article 17
TABLE 4
MULTIMARKET CONTACT (MMC) AND CONCENTRATION BY MARKET SIZE (1998 POPULATION QUARTILE)
CITY-MARKETS, 1995-1998
City Markets (248)
MMC
(Top 2-firms)
by 1998 Population
Quartile
(Avg. Pop. 1000’s)
Concentration
(Top 2-Firms)
1995
1998
% Change
All Markets
1.27
3.17
149.6
Quartile 1
(109.9)
1.02
1.26
23.5
Quartile 2
(210.2)
1.02
1.62
58.8
Quartile 3
(404.5)
1.09
2.28
109.2
Quartile 4
(1980.2)
1.94
7.08
264.9
All Markets
0.51
0.63
23.5
Quartile 1
0.60
0.67
11.7
Quartile 2
0.54
0.65
20.4
Quartile 3
0.53
0.64
20.8
Quartile 4
0.38
0.55
44.7
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
21
TABLE 5
FCC RESTRICTIONS ON LOCAL JOINT OWNERSHIP OF RADIO STATIONS AND MEASURES OF MULTIMARKET CONTACT
AND CONCENTRATION BY POLICY BANDS USING MARKET SIZE
Restrictions on Local Joint Ownership of Radio
Stations Pre- & Post-1996 Telecommunications Act
Max. No. of Jointly
Limit on No. in Same
Owned Stations
Service (AM or FM)
Size of
Market
(No. of
Stations)
Pre- 1996
Post- 1996
Pre- 1996
45+
4(a)
8
2
30-44
4(a)
7
15-29
4(a)
0-14 (b)
Nationwide
Notes:
(a)
(b)
Post- 1996
Measures of Multimarket Contact and Concentration by
City-Market, 1995-1998
Multimarket Contact (MMC)
Concentration
(Top 2-Firms)
(Top 2-Firms)
95
98
95
98
5
2.73
8.00
193.0
0.27
0.43
59.3
2
4
1.60
6.11
281.9
0.39
0.57
46.2
6
2
4
1.11
2.43
118.9
0.49
0.62
26.5
3
5
2
3
1.01
1.24
22.8
0.67
0.71
6.0
40
No limit
20
No limit
1.27
3.17
149.6
0.51
0.63
23.5
In any case, no one may own more than 25% of the stations.
In any case, no one may own more than 50% of the stations.
Contributions to Economic Analysis & Policy
%
%
22
Vol. 5 [2006], No. 1, Article 17
TABLE 6
REGRESSIONS OF (LOG) AVERAGE PRICE ON MULTIMARKET CONTACT AND OWNERSHIP CONCENTRATION
2 AND 4 FIRM MEASURES
CITY-MARKET FIXED EFFECTS
(1)
FE
All Markets
MMC- 2 firm
Concentration- 2 firm
0.0037***
(0.0014)
0.0003
(0.0620)
(2)
FE
Audited Markets
0.0514***
(0.0098)
0.1158***
(0.0122)
0.2066***
(0.0169)
5.9235***
(0.0321)
0.0646***
(0.0119)
0.1184***
(0.0158)
0.2104***
(0.0325)
5.8696***
(0.0335)
0.0077
(0.0051)
-0.0299
(0.0718)
0.0529***
(0.0102)
0.1186***
(0.0128)
0.2104***
(0.0166)
5.9413***
(0.0516)
990
248
0.43
404
101
0.44
990
248
0.43
Concentration- 4 firm
1997
1998
Constant
Observations
Number of markets
R-squared
(4)
FE
Audited Markets
0.0033*
(0.0019)
0.0291
(0.0747)
MMC- 4 firm
1996
(3)
FE
All Markets
0.0117
(0.0092)
-0.0330
(0.0968)
0.0672***
(0.0113)
0.1228***
(0.0152)
0.2132***
(0.0286)
5.8932***
(0.0619)
404
101
0.44
Notes: All markets sample includes all city-markets for which all data are available. Audited markets sample includes markets
audited by an accounting firm. Regressions report standard errors clustered by market. ***/**/* represent significance at the
1%/5%/10% level.
http://www.bepress.com/bejeap/contributions/vol5/iss1/art17
Waldfogel and Wulf: Measuring the Effect of Multimarket Contact on Competition
23
TABLE 7
REGRESSIONS OF (LOG) AVERAGE PRICE ON MULTIMARKET CONTACT AND OWNERSHIP CONCENTRATION
2 FIRM MEASURES
CITY-MARKET FIXED EFFECTS AND IV REGRESSIONS
MMC
(1)
FE-AR(1)
(2)
FE
(3)
FE
1996-1998
1996-1998
1995-1998
0.0014
(0.0022)
0.0218
(0.0666)
-0.1112***
(0.0091)
-0.0836***
(0.0086)
0.0021
(0.0014)
0.0308
(0.0714)
-0.1559***
(0.0158)
-0.0916***
(0.0122)
0.0010
(0.0016)
-0.0558
(0.0823)
Constant
6.1264***
(0.0322)
6.1156***
(0.0476)
City-market*Trend
--
Observations
Number of markets
742
248
Concentration
1996
1997
(4)
FE-IV
Population
Instrument
1995-1998
(5)
FE-IV
Policy Band
Instrument
1995-1998
5.2388***
(0.0106)
0.0151
(0.0282)
0.0917
(1.7166)
0.0430
(0.0783)
0.0941
(0.1496)
0.1740
(0.1559)
5.8624***
(0.8369)
0.0365
(0.0436)
-1.2997
(1.6902)
0.1062
(0.0726)
0.2159*
(0.1311)
0.3013**
(0.1248)
6.5418***
(0.8039)
--
yes
--
--
742
248
990
248
990
248
990
248
1998
Notes: Sample includes all city-markets for which all data are available. Column (1) reports standard errors adjusted for serial
correlation (AR(1)). Column (2) is based on the same sample as in column (1), but reports standard errors clustered by market.
Regressions in columns (3) through (5) are based on the sample that includes all city-markets for which all data are available.
***/**/* represent significance at the 1%/5%/10% level.
Contributions to Economic Analysis & Policy