It is rare that bank regulators confront each other in public, as happened recently when Andrew Bailey, the chief of the Bank of England’s Prudential Regulation Authority, criticised the European Banking Authority for a ‘misguided’ approach to the reform of bankers’ bonuses (Financial Times, 16 October, 2014). The EU capital requirements directive mandates that bonuses should not exceed 100% of the base salary – or 200% with shareholders’ approval. The British regulators are strongly opposed to this invariant cap on bonus pay, which would effectively limit the performance-sensitivity of banker compensation. Instead, the Bank of England puts its trust in deferred compensation and claw backs to strengthen the sensitivity of bonuses to long-term performance.
The conflict between the European Banking Authority and the Bank of England reveals that there is no consensus about the role of incentive pay in banking. While its advocates, in particular the banks themselves, highlight its positive effect on profitability, opponents warn against increased risk-taking incentives. Indeed, any purposeful reform of bonus payments in banking depends on the answers to the following key questions:
- Does higher performance sensitivity of banker compensation increase bank profitability?
- Does it increase the risk taking by bankers?
- And if the answer to both these questions is yes, can banker compensation become too performance sensitive either from the shareholder or the public policy perspective?
New employee-level payroll data
That we have so little empirical evidence regarding these three questions has much to do with the lack of information about bank compensation, where privacy concerns are very understandable. Yet, modern techniques to anonymise individual pay data can go a long way to address such concerns. Using such anonymised pay data on 1.2 million bank employee-years, our group of four researchers at the Universities of Geneva, Cologne, and Dresden were able to analyse the incentive pay in 66 banks in the Austrian, German, and Swiss banking sector collected over an eight year period from 2004 to 2011 (Efing et al. 2014).
The richness of the data allows us to identify the particular bank function and hierarchy level of each bank employee so that the bonus share – defined as a percentage of total annual compensation – can be calculated separately for the critical segments investment banking and capital market/treasury. Unsurprisingly, bonuses relative to total compensation are highest in investment banking and capital market related segments. Moreover, the size of the bonus share in these two bank segments correlates strongly with its variability – suggesting that bonuses are a truly variable compensation component.
Pre-crisis and post-crisis incentives in comparison
In a first step, we compute the average bonus share for all employees in the segments investment banking and capital market/treasury. Unlike existing research, we include employees at all hierarchical levels and not only bank executives. Figure 1 plots the average bonus share of a bank during the pre-crisis period 2004-7 (horizontal axis) and the post-crisis period 2008-11 (vertical axis). The data allow drawing two conclusions (see Figure 1):
Banks feature a large variation in their average bonus share even when we focus only on employees in the investment banking and capital market/treasury management functions.
For most banks in our sample, the bonus share decreases substantially from the pre-crisis to the post-crisis period.
Figure 1. Pre-crisis and post-crisis average bonus share
One may wonder whether the average decrease in the bonus share after 2008 may just reflect a decrease in profitability in the investment banking and capital market functions of the respective banks. However, Figure 2 shows evidence against this hypothesis. Each point in Figure 2 represents the average (logarithmic) trading income of a bank during the pre-crisis (horizontal axis) and the post-crisis period (vertical axis). Clearly, the points cluster around the 45-degree-line. Therefore, the drop of the bonus share after 2008 (Figure 1) cannot be explained by a corresponding decrease in trading income (Figure 2). We conjecture that external pressure on banks to reduce bonuses might be responsible for the decrease in incentive pay during the post-crisis period.
While absolute trading income is a suitable measure to compare the pre- and post-crisis profitability of a given bank, it is not well suited to compare trading profits across banks due to the large cross-sectional differences in the size of banks. In order to account for bank size, we use a bank’s interest income as a size measure and define the relative trading income as the ratio of the annual trading income and the annual interest income.
Figure 2. Pre- and post-crisis average trading income (in log)
Do large bonuses increase trading profits?
Figure 1 shows that the bonus payments constitute a large component of traders' payrolls and often reach levels of roughly 60% of total compensation. What motivates banks to pay such high bonuses? One possible explanation would be that banks promise bonuses to incentivise traders to work harder and to earn higher profits for the bank. Indeed, the data exhibit a strong positive correlation between the bonus share and the (relative) trading income of banks. However, while this observation is certainly consistent with a causal relationship between effort levels and profitability, the same positive correlation would also arise if bonuses were awarded for profits that had been generated purely by chance. Simply by looking at correlations, we cannot be sure whether performance-contingent pay in the form of bonuses is really causing trading income to increase or if causality runs in the reverse direction and bonuses are simply high because luck had been on the traders' side.
How can reverse causality be ruled out as an explanation? We identify two exogenous reasons why some traders receive particularly high bonuses that have nothing to do with trading profitability directly.
- First, we use the bonus share in retail, private, and corporate banking segments to proxy for the bonus share of traders.
Typically, employees in the retail segment do not receive larger bonuses because their colleagues in trading have generated high profits. Yet, banks that generally pay high bonuses to employees in retail banking do tend to pay high bonuses to their traders, too. Hence, we can use the bonus share in trading-unrelated bank divisions like retail as an instrument for the pay incentives in trading and be sure that this instrument is unaffected by reverse causality.
- Second, the share of employment outside the capital market divisions relative to total bank employment is used as an additional instrument.
A bank with a large retail, private, and corporate banking segment might monitor its traders with a different intensity than banks whose core business is investment banking.1
Figure 3a plots the annual relative trading profits (after subtracting the variation explained by control variables) against the predicted (instrumented) bonus share of the respective bank.
- Relative trading income shows a strong positive relationship with the instrumented bonus share.
This relationship has a causal interpretation if we are willing to assume that the bonus culture of banks (as proxied by the bonus share in unrelated bank segments) represents an exogenous influence on the bonus share of the investment banking and capital market segment.
We also note that the causal effect is economically large. An increase in the bonus share by one standard deviation implies an increase in the relative trading income by 1.1 standard deviations. One can also show that the effect is even stronger in a regression which puts more weight on the larger banks.
Does higher trading profitability come with more risk?
In financial markets, higher profitability typically comes with more risk. We measure risk-taking by the (log of the) standard deviation of trading income of each bank both for the pre-crisis and crisis period. The same set of instruments as before produces the scatter plot depicted in Figure 3b between the volatility of relative trading income and the instrumented bonus share.
- The strong positive relationship suggests that a stronger bonus culture implies a significant increase in the riskiness of trading profits.
As we dispose of only two volatility observations per bank (pre-crisis and post-crisis period), the scatter plot features much fewer observations compared to Figure 3a.
In conclusion, stronger bonus incentives increase the expected income of bank trading only at the cost of higher risk as measured by the standard deviation of the relative trading income.2 In a next step, we would like to explore if the increase in profitability is large enough to justify the increase in risk from a welfare point of view or whether risk-taking is excessive.
Figure 3. Annual relative trading profits and predicted bonus share
Are bank bonus incentives optimal?
To measure the risk-profitability trade-off in trading, we compute the ratio of trading income to its volatility, called the Sharpe ratio of trading. From a theoretical perspective, any incentive system that maximises this Sharpe ratio also maximises the value of the bank and is, in this sense, socially optimal.3
Figure 4 depicts the Sharpe ratio of trading against the instrumented bonus share. Crosses represent pre-crisis observations. They are scattered around a regression line with a steep negative slope.
- The ratio between trading profitability and risk improves when the (instrumented) bonus share decreases. We conclude from the negative marginal effect that pay incentives are too strong prior to the crisis from the viewpoint of bank value maximisation.
The post-crisis observations (represented by dots) show a regression line with a positive, albeit small slope. The bonus moderation after 2008 seems to bring pay incentives closer to their optimal levels for Austrian, German, and Swiss banks.
Figure 4. Sharpe ratio of trading and instrumented bonus share
Limitations of the analysis
A weakness of the above analysis is its reliance on the annual trading income of banks rather than banks’ individual asset positions. Unfortunately, public reporting requirements for banks are very lax compared to the requirements for publicly traded equity funds, which have to report every single stock position at regular intervals. Thus, we can only use a very noisy measure of bank risk taking which greatly limits the statistical significance of some of our results.
Secondly, we also highlight that the extreme leverage of most banks implies that shareholders might push for equity value maximisation rather than the socially desirable asset value maximisation. Under high leverage, the option-like character of equity creates a wedge between the socially desirable amount of risk taking and the much higher risk appetite of bank shareholders – particularly if banks enjoy implicit public guarantees. Socially excessive bonus compensation may therefore still be optimal from the shareholder’s perspective.
Finally, the conclusions drawn from our sample of Austrian, German, and Swiss banks need not necessarily carry over to Anglo-Saxon banks, which tend to pay higher bonuses.
A general conclusion of our findings is that performance-contingent bonus payments affect both the profitability as well as the riskiness of banking. A good incentive pay system needs to target the optimal trade-off between the two and it is an empirical question if a bank’s incentive pay achieves this goal. We highlight that in practice this trade-off is difficult to quantify and evaluate.
The policy of simply capping banker bonuses, as mandated by the European Banking Authority, is very questionable. Such a policy seems incompatible with the limited regulatory capacity to determine what constitutes optimal incentive pay. This is reflected in the arbitrariness of the proposed 100% bonus threshold.
The policy reforms by the Bank of England concentrate on deferred compensation and claw-backs for cash bonuses (Financial Times 29 July, 2014). While an alignment of variable compensation with long-term performance seems generally desirable, the issue of the optimal size of variable compensation is left to the banks and their shareholders. There are reasonable doubts whether this will lead to much reform. It certainly does not forestall a come-back of potentially excessive bonus pay.
In light of these problems with bank bonus reform, two alternative regulatory measures are much simpler and more promising:
- First, regulators could set higher bank capital requirements in order to better align the shareholders’ interests with those of the public.
The argument for a much larger equity share in bank funding has been forcefully made by Admati et al. (2011) and applies equally to the design of incentive systems; only if shareholders provide a sizeable proportion of the bank funding, can they be expected to pursue a socially desirable incentive system.
- Second, public policy could empower investors, shareholders, and bank boards by full public disclosure of all bank assets.
Transparency at the individual asset level is already the disclosure standard for equity funds and would allow a much more accurate public assessment and monitoring of bank risk. Here, the long-term benefits in terms of better risk pricing, better governance, and ultimately – better incentive systems – seem large. Why not go after the low-hanging fruits first?
Admati, A R, P M DeMarzo, M F Hellwig, and P C Pfleiderer (2011), “Fallacies, Irrelevant Facts, and Myths in the Discussion of Capital Regulation: Why Bank Equity is not Expensive”, Rock Center for Corporate Governance at Stanford University Working Paper No. 86, MPI Collective Goods Preprint, No. 2010/42.
Duffie, D (2011), How Big Banks Fail and What to Do about It, Princeton University Press, Princeton and Oxford.
Efing, M, H Hau, P Kampkötter, and J Steinbrecher (2014), “Incentive Pay and Bank Risk-Taking: Evidence from Austrian, German, and Swiss Banks”, CEPR Discussion Paper 10217.
Fahlenbrach, R (2009), “Shareholder Rights, Boards, and CEO Compensation”, Review of Finance, 13(1), 81-113.
Financial Stability Board (2009), “FSB Principles for Sound Compensation Practices - Implementation Standards”.
Financial Times (2014), "UK bankers face tough bonus claw backs", 29 July.
Financial Times (2014), "BoE lashes out at EU bonus cap rules", 16 October.
H Hau, and M Thum (2009), “Subprime Crisis and Board (in-)Competence: Private vs. Public Banks in Germany”, Economic Policy, 24(60), 701-751.
1 Previous research has indeed found weaker bank governance to be related to higher incentive pay (Fahlenbrach 2009) and bank risk (Hau and Thum 2009).
2 A more detailed analysis of tail risk is unfortunately not possible with the limited public data available to us.
3 Sharpe ratio maximisation is equivalent to the maximisation of the net present value of the bank if trading is highly leveraged: Under normal pre-crisis conditions, a dealer bank might have financed trading positions mostly with overnight repos with an average haircut of under 2%, thus allowing an effective leverage ratio of at least 50 (Duffie 2011, page 32). The optimal incentive system of a bank therefore needs to maximise the Sharpe ratio of trading income, which should imply a zero marginal effect of bonus incentives on the Sharpe ratio of trading income.