Why performance pay may underperform

Tim Harford wrote an interesting piece last month about performance pay. He wrote:

Here’s an age-old management conundrum: who should be rewarded for high performance, and how? As Diane Coyle, the economist and former adviser to the UK Treasury, recently observed in this newspaper, the answer to the question is usually self-serving. Simple and easily monitored jobs, such as flipping burgers, are natural candidates for performance incentives. Yet somehow it’s the inhabitants of the C-suite who tend to pick up bonuses, despite the fact that their complex, hard-to-measure jobs are poorly suited to the crude nature of performance-related pay.

Harford (as always) does a great job of summarising the state of research in this area. He correctly identifies that financial rewards don't work in all situations. I'm going to use this post to highlight the factors that must be in place for a performance pay scheme to work well - the absence of one or more of these factors will lead to a performance pay scheme that won't work so well.

First though, some background. Why have performance pay? Employers face a problem we refer to as the principal-agent problem (a specific type of moral hazard). The employer (the principal) engages an employee (the agent) to work on their behalf. However, the goals of the employee are rarely perfectly aligned with the goals of the employer. So, because the employer cannot easily monitor all of the employee's activities, the employee can engage in activities that are not necessarily in the best interests of the employer (like goofing off). Rewarding employees for meeting set targets (and performance pay more generally) is one way of re-aligning the interests of the employee with those of the employer (other options include closer monitoring of employees, and paying efficiency wages, which I have previously discussed here). However, for performance pay to work well a number of things factors need to be present.

First, employees must respond to incentives. This seems like a given, since one of the assumptions economists make is that people respond to incentives. However, it might not always be true. If employees are already highly paid, and the performance payment is small, it might not be enough incentive to encourage greater work effort. So, if you are selling farm machinery, for performance pay to work well your salespeople must be willing to try to earn more sales to capture the performance pay (e.g. commissions).

Second, employees' output must be sensitive to their effort. If employees work harder but the additional effort does not enable them to produce (or sell) more, then rewarding them with performance pay simply won't work well. Why waste your time working harder if it doesn't lead to more output (and higher pay)? So, for performance pay to work well your farm machinery salespeople must be able to sell more tractors if they work harder (at trade shows, visiting potential clients, etc.).

Third, employees' output must be measured easily. If you are going to reward employees for their performance, you must be able to objectively measure their performance. On top of that, employees must be able to believe that the measurement of their performance is accurate. Measurements that are not credible will not encourage employees to work harder. So, for performance pay to work you must be able to know how many tractors each salesperson is selling.

Fourth, employees must not be too risk averse. Risk averse people prefer a higher degree of certainty. Performance pay reduces the certainty of employees' incomes, so if they are highly risk averse they will prefer to work elsewhere (this may or may not be a good thing!). So, your farm machinery salespeople must be willing to accept some likely fluctuations in their salary (as their sales increase or decrease from month to month).

Finally, the level of risk that is beyond the employees' control must be low. If employees' performance depends on their own effort, but also on other factors that are beyond their control, then rewarding high performance may not necessarily increase work effort. This will be a greater problem the greater the share of employee performance that is driven by the external factors. So, if farm machinery sales are driven more by the weather and how farm profits are going, and less by the efforts of the individual salespeople themselves, it will not work so well.

On that last point though, one work-around is to reward employees for their relative performance, i.e. their performance relative to their peers, to industry benchmarks, or to agreed targets. However, it is possible to take performance pay too far - to the extent where competition between different employees is encouraged to an unhealthy extent. In any case, performance pay is one method of dealing with underperforming employees - but if the factors noted above are not present the performance pay scheme itself may well underperform.

The economic non-impact of stadiums and arenas

Peddlers of economic impact studies are modern day snake oil salesmen. Most economic impact studies would not be less credible if they were printed on toilet paper (but they would be more useful!). Exhibit A is the cost-benefit analysis used to justify the upgrade of the white elephant Claudelands Events Centre in Hamilton (see here and here, and note that things are not getting any better). Fudging the figures on cost-benefit analysis to justify public spending on sport and recreation facilities is nothing special to Hamilton either - see this recent post by Tim Harford on measuring (or should that be mis-measuring?) the benefits of the Olympic Games.

A new paper by Sam Richardson, published in New Zealand Economic Papers (sorry I don't see an ungated version anywhere), makes the case against any substantial economic impact of stadiums and arenas, based on New Zealand data. And New Zealand is not an outlier - Richardson's work simply adds to a mountain of existing research that says the same, not just for stadiums and arenas, but also conference centres and large events (on this topic I'm particularly looking forward to reading this book by Andrew Zimbalist, which has been on my to-be-read pile for too long, and which I will review here when done).

Economic impact studies typically go wrong in one (or often more than one) of three ways: (1) they use the wrong counter-factual; (2) they don't account for leakages to outside the local economy; and (3) they fail to consider the opportunity costs. In the first of these, many economic impact studies assume that all spending that relates to stadium events would not have been spent otherwise. Clearly, this is incorrect for local people attending events at the stadium, who probably would have spent that money anyway. In the second, many economic impact studies fail to recognise that not all spending associated with a stadium stays in the local economy. For instance, if a hot dog seller comes from out-of-town to service the event, then their income goes out of the local economy. In the third, almost all economic impact studies fail to consider what the next best use of the money spent on the stadium is, and the net benefits that alternative use could have generated.

Anyway, back to Richardson's work. He looks at the impact of stadium construction across 13 territorial authorities that built or upgraded 24 sports facilities over the period 1997 to 2009. Specifically he looks at the impacts on employment in the construction sector, and local GDP, both during and after construction. He finds:

that there is a statistically significant (judged in terms of a 10% significance level) increase in quarterly employment growth of 1.033 percentage points (p-value D 0.003) for each quarter during facility construction...

Model 1(b) shows the impact of the post-construction period for the combined facility projects, and there is no statistically significant impact...

Results for model 2(a) indicate that the aggregated facility during-construction coefficient is not statistically significant, indicating that facility projects in general did not have any impact on quarterly real GDP during construction. Results from model 2(b) show that the post-construction aggregate facility coefficient is also statistically insignificant.

In other words, there is a small impact on construction sector employment growth during the period of construction (no surprises there), but that the effect does not persist after construction. Probably the construction employment is only temporary because the large construction firms that engage in building stadiums and arenas bring in construction staff, but then move them onto the site of the next large project when the current one is complete. Even worse, there is no effect on local real GDP, either during or after construction. Now, we might expect a lesser impact of

Richardson concludes:

Results from this paper strengthen the conclusions of the majority of research throughout the scholarly literature in this field that sports facilities should not be relied upon as economic stimuli. They do not generate increases in long-term employment, and they have no impact on local area incomes (as measured by real GDP).

For more context on Richardson's work, see his blog (although I note that he hasn't posted for over two years, but there is good stuff there nonetheless).

[HT: Eric Crampton at Offsetting Behaviour]

Winners and losers in population growth

I was quoted at length in a story by Michael Daly published in Stuff last week. My comments were based on my ongoing research programme (with many collaborators) on subnational population projections and migration:

While New Zealand's population was continuing to grow it was becoming much more concentrated in the main centres, Cameron said. "For a lot of regions it really is about managing the decline." 

Declining areas could have a reverse momentum. "You can get young people moving out of the area. You're going to get less natural increase, that's going to reinforce population decline," Cameron said.

Declining rural areas tended to have more older people, while the larger centres had tertiary education opportunities that drew in the young.

"Areas that have more job growth, better income availability, lower unemployment, those tend to be places that are attractive for people to live," he said.

Good amenities were also important. "There's quite a difference between the sorts of things you can do in Auckland from Taumarunui, for instance. People like to be able to be able to do things, and urban centres tend to have more of those opportunities."

Migration was one factor contributing to fast growth in some areas but so was natural increase - the difference between births and deaths.

Although some migrants were retirees, most tended to be younger than average. "Younger people have more babies so that reinforces itself."

Cameron did not expect there would be a tipping point where Auckland's high house prices and traffic congestion would lead to an avalanche of people moving out. "It's a trickle rather than a torrent," he said.

But Auckland's high property prices were benefiting Hamilton and the Waikato District. While some people were commuting north into Auckland, jobs were also spilling over from Auckland into Waikato, where land was much cheaper.

Waikato also had good road and rail links to the ports in Tauranga and Auckland, Cameron said. The dairy boom, although ending 18 months or so ago, had also brought considerable income into Waikato, as well as into Taranaki.

Hamilton did have a similarity with Dunedin that counted against the cities. "They have the university there (Dunedin), which brings in a lot of young people but once they finish they are all heading out of Dunedin. We have the same thing here in Hamilton."

Dunedin's slow growth was a long term trend. It had been New Zealand's largest city in the 19th century and it was hard to pull out the causal factors that had led to its decline in importance.

Queenstown had a booming tourist industry, which was labour intensive, Cameron said. "There's a lot of jobs available. Those jobs pull in people. The more people you have the more hairdressers and things you need. It gets a little bit of momentum going."

Nelson had the same sort of sunbelt migration that Tauranga did, including the arrival of many retired people. Gisborne was "so far away from everywhere. It's very isolated out there."

The notoriety of Wellington's weather didn't seem to be a massive disadvantage, Cameron said. He had looked into the effects of climate on migration, and while it had an effect it wasn't very large.

"People do tend to move to sunny, warmer, less wet places, but the actual size of that effect is pretty small."

Some work had been done on whether regions could arrest population decline by attracting migrants, he said. "But the amount of migration you would need to offset both the ageing population and the fact young people want to move out - it's unrealistic."

Overseas, where areas with declining populations had managed a resurgence, it was usually because of some sort of black swan event. For example, the only thing that turned the population change in North and South Dakota around had been the fracking boom. "It was really a one-off," Cameron said.

One small district that had done a good job of turning around declining fortunes was Otorohanga, which had been losing people for a long time before growing between the last two censuses.

"They managed to retain a lot of their young people," he said. Dale Williams, who was mayor from 2004-2013, had a compact with local employers to make jobs available for young people.

"Because young people could stay in Otorohanga and have a good job, many chose to stay. Then you have more natural increase in the population, as well."

Daly did a good job of collecting and summarising my comments. The key point is that the areas that are already growing fast (especially Auckland, Tauranga, and Hamilton - the so-called 'Golden Triangle' of the upper North Island) are doing so not solely because of migration. Migrants tend to be younger than non-migrants (even for Tauranga a lot of in-migration is young people), and younger people generate additional population growth because they have children. At the other extreme, rural and peripheral areas of the country are experiencing sustained out-migration of the younger population, which is a double-blow (again because there will be fewer children as a result). It could be (and may yet become) worse though - consider the situation in Japan.

The idea that there will be 'winners' and 'losers' in future population growth is nothing new. Consider the discussion of 'zombie towns' in New Zealand (which I discussed here). The Marsden funded project Tai Timu Tangata (led by Natalie Jackson, and including me) will begin producing some final outputs over the coming months. I look forward to outlining some of those outputs here.

Why you can't use taxes to restrict a monopoly's market power

I mark all my own exam papers (whereas many other lecturers out-source this to tutors). I think this gives me a better feel for where students are going wrong and usually I can figure out why. This helps me to improve my teaching for future semesters, especially where it is some passing comment that students have adopted in their answer. Sometimes though, I'm sure it isn't something I've said that has led the students astray. One case of the latter happened in the A Semester ECON100 exam, when I asked a question about the options government has to restrict the market power of monopolies. Many students answered that an appropriate way to restrict market power was to tax the monopoly. Here's why that's not a correct answer.

A monopoly is a sole seller of its product. This gives it market power - the ability to choose a price that will maximise its profit. Consider the diagram below (which assumes a constant-cost firm). The monopoly will operate at the profit-maximising quantity (where marginal revenue (MR) is equal to marginal cost (MC)), which is QM. To sell that profit-maximising quantity they will set a price of PM (because at the price PM consumers will demand exactly QM units of the product).

With that price and quantity, the consumer surplus (the difference between the amount that consumers are willing to pay (shown by the demand curve), and the amount they actually pay (the price)) is the triangle area ABF. The producer surplus (the difference between the amount the monopoly producer receives (the price), and their costs (which are shown by the marginal cost curve)) is the rectangle area FBDG. Total welfare is the combination of consumer and producer surplus, i.e. the area ABDG.

If this was a perfectly competitive market, the market would operate at the point where supply is equal to demand (and note that the supply curve is the marginal cost curve for these constant-cost firms). The perfectly competitive market would operate at the quantity QC and price PC. Consumer surplus would be the triangle AEG, while producer surplus would be zero (because the price PC is equal to the cost of every unit produced), so total welfare is also the triangle AEG.

Now, comparing the monopoly with perfect competition, we see that market power (exercised by the monopoly firm) leads to a higher price (PM) than under perfect competition (PC). That in itself isn't a problem though. What is a problem is the loss of total welfare, which is ABDG with a monopoly firm, but AEG with perfect competition. The difference is the area BED - the deadweight loss of monopoly. It is this deadweight loss that is the main reason why governments might prefer to restrict market power.

So, what happens if you tax the monopoly firm? Consider the diagram below. The tax would likely be levied on the seller since this is administratively easier, so this is like increasing their costs. We represent the tax as a new curve S + tax (or MC + tax in this case for a monopoly). These 'higher' costs for the monopoly move the profit-maximising quantity down to QT (the quantity where MC + tax is equal to marginal revenue), and lead to an even higher price PT.

Consumer surplus with the taxed monopoly is the triangle area AJH, and producer surplus is the area HJKL. The government receives tax revenue equal to the area LKNG (this is part of total welfare for society because the government can use that revenue to pay for roads, schools, etc.). So total welfare with the taxed monopoly is the area AJNG. Notice that the combination of monopoly plus tax has increased the deadweight loss from the area BED to the area JEN. Taxing the monopoly makes the problem of lost welfare worse not better.

And that is why you can't use taxes to restrict the market power of a monopoly. The monopoly can still use its market power to set the price, and in response to the tax it will set an even higher price, increasing the size of the deadweight loss. A better response is to use a price control (a price ceiling - a maximum price that is lower than PM), government ownership (so the government can choose any price it wants, including a price below PM), or using anti-trust laws to prevent firms from merging into larger firms with market power in the first place.