One of the talking points that has come up as part of the NBA lockout is the disparity in spending across teams. The league (as represented by David Stern) like to mention that the Lakers spend twice as much as the Kings. This can’t possibly be fair; the suggestion is that the Kings can’t compete when the Lakers can spend so much (ignoring, of course, that the Kings had perfectly average salaries the previous couple seasons and spend 25% more than average for a couple years in the early 2000s, or that the Lakers were pretty average spenders for some seasons sprinkled through the last 10). The Union has countered that salary has little to do with winning (for example), a point I have made before. Yes, teams that spend more tend to win more, but it’s to a limited extent. There are barriers to teams simply buying a title in the current NBA system, like the luxury tax, rookie salary scale, and inability for teams to identify the players that really help them win. The guys at the Wages of Wins have mentioned that changes in cap situations in other leagues has demonstrated that they have little effect on parity. I thought I would turn to the NHL, home of my beloved Red Wings, to see what I could get from the numbers.
I found this wikipedia page with a list of league average salaries and team salaries for a spread of years. Since the numbers end in 07-08, that’s as far forward as I went. Since ESPN’s standings go back to 01-02, that’s when I started. That isn’t a lot of seasons, but it does happen to coincidentally give me three pre-lockout and three post-lockout seasons to compare. So I gathered the salary for each team in those six seasons (the seventh year being the absent lockout year), how many points they scored (meaning record, not goals), if they made the playoffs, and how far they got in the playoffs (measured as 1 point for losing in the first round, 2 points for winning in the first round, 4 points for winning the second round, 7 for winning the conference finals, and 11 for winning the title). I converted salary to a proportion relative to league average as well as a z score, and team points to a proportion relative to league average (teams scored more points after the lockout because ties were eliminated). I’ll only mention salary as a proportion though, as the correlation between that and as a Z-score is over .9. As a quick summary, the NHL went from a league with no revenue sharing or caps of any kind to a league with a hard cap ceiling and floor as well as individual player caps.
There are a lot of numbers we could look at. Stacey Brook already covered one aspect, which is the Noll-Scully measure of competitive balance. Balance did increase from pre to post-lockout, but that was part of a decades-long increase in balance in the NHL. It isn’t clear that the salary cap was the reason there. But it is certainly true that spending balance has increased; the standard deviations listed in the wikipedia page dropped dramatically after the lockout, as did the coefficient of variation (standard deviation divided by the mean, which helps account for the fact that variables with higher means also tend to have higher standard deviations).
The discussion in the NBA has been about the correlation between salary and wins. In the entire six years, treating each team-season as a separate data point, the NHL correlation (using salary proportion and points proportion) is .463. But what we really want to know is if that correlation changed with the lockout. I ran a regression predicting points from salary and if the season was capped or not. Having a salary cap did not significantly interact with salary. If we split out the three pre and post seasons and run the correlation just in those seasons, the correlation pre-lockout is .52 and the correlation post-lockout is .41. Obviously a numeric difference, but it doesn’t pan out statistically.
How about getting to the playoffs? Again, spending more is significantly associated with making the playoffs. A team that spends an average amount has about a 55% chance of making the playoffs (the NHL sends 16 out of 30 teams); a team that spends 120% of average has closer to a 70% chance. How was this affected by the cap? There’s a trend for a positive interaction (spending more gives you a better chance in capped years) that is countered by a negative main effect of having a cap. But into numbers: an average spending team in uncapped years had a 55% chance of making the playoffs and a 54% chance in capped years. A 120% spending team had a 67% chance in uncapped years and a 80% chance in capped years. Thus it looks like the cap may have had the reverse effect; teams that spend more are more likely to make the playoffs under the cap than they were without a cap. If owners had the intuition that they could spend their way to the playoffs, they would be more rewarded with the current cap structure.
The NHL playoffs are a crapshoot, at least compared to the NBA. There isn’t even a home-ice advantage that I can find. But does spending get you further in the playoffs? I regressed playoff points by salary. You get a significant effect of salary, but it was not affected by the cap. An average spending team in either period could expect 1.4 points (getting into the playoffs and maybe winning a round); spending more does numerically lead to bigger expected points, but the effect is very small. A team spending 180% of league average could expect 2.5 points pre-lockout (winning the first round with a small chance of moving further) but 3.5 points post-lockout (close to winning in the second round). But again, the benefit to spending more is in the post-lockout period.
Is this really about spending? Let’s assume that team get points by having good players. Does spending more explain anything beyond team quality as measured by points? I ran the previous two regressions (the playoff binomial and playoff points) again but included team points. Salary now has no ability to predict making the playoffs; if you use scaled values points is dramatically more important than salary (20 times post-lockout, 60 times pre, and 100 times in the entire data set). The same story is true for predicting playoff points.
There are only six champions in the data set, so running a regression to predict winning the Stanley Cup is a little sparse. But you do get significant results using points, but not salary. In fact, salary isn’t predictive even if used alone in the regression; team points is.
As a final step, I summed across seasons for a team to see if team consistency has changed with the cap. Across the full set, the correlation between team total points and team total spending is .586, but the confidence interval is obviously pretty wide (there are only 30 data points). If you split it into pre and post, the correlation is numerically lower post-lockout but not significantly (.587 down to .457; I assume not a significant difference because the confidence intervals overlap to a large degree).
So let’s sum up. Before the lockout in 04-05, the NHL had no cap and little (if any) restriction on player salary. There was a correlation between spending and points, the probability of making the playoffs, and how far a team would make it in the playoffs. This was obviously unfair to the teams unwilling or unable to spend as much as, say, the Red Wings, who averaged about 170% of league-average spending in the three prior years. However, after a variety of restrictions were put on spending after the lockout, there was still a correlation between salary and all those other measures. And the correlation didn’t really change; if anything, it appears to reward spending more than before. So instituting a hard cap and whatnot did nothing to improve competitive balance, and it may have made it easier for teams to buy their way to titles (spending 120% of average gets you more than it used to).
But all of this is really moot; once we account for team quality spending doesn’t matter. That was true before the lockout and it’s true after the lockout. So I’m going to keep repeating this: if owners would spend their money wisely, they could likely be very competitive even without spending a lot of money (the Red Wings were average spenders after the lockout and put up virtually equivalent results to the prior three years). Would spending more help them? Only if they spend the extra money on good players. Runaway spending in the pre-lockout NHL, as well as the NBA now, is because owners can’t keep from throwing max contracts at non-max players. It isn’t the players’ fault; if someone offered me $120 million I would take it. If the owners are losing as much money as they claim, it’s because they couldn’t keep their wallet in their pants.