Barrie & Hibbert Blog
Understanding and measuring liquidity premia in asset markets
Posted on 17-12-2009 by Axel Kirchner | 0 comments
Earlier this month, leading academics and practitioners convened in London to discuss the latest thinking on liquidity premia: both its very existence and whether it can be reliably measured formed the focus for eager debate, with Yakov Amihud of the Stern Business School, NYU expressing amazement that its existence continues to be doubted. He has been studying liquidity premia since 1986 and is an absolute follower of the LP faith. For the most part, however, the main thrust of the day’s proceedings turned on issues of measurability and the concerns raised by regulators as to whether liquidity premia should be allowed for in discounting future cash flows. Axel Kirchner recounts the event.
Measurability
The body of theoretical work, strongly supported by empirical studies, shows that illiquid securities are priced at discounts to identical liquid securities irrespective of the time period studied or the methodology used. In other words, hard-to-trade assets sell at a different price to more liquid assets with otherwise equivalent characteristics indicating strong evidence for the existence of a liquidity premium.
Questions were raised throughout the day pertaining to whether there are reliable, objective and practical methods for estimating it at specific points in time. The methods currently in use for this purpose – the CDS negative basis, structural models and covered bond based approaches - all share a similar conceptual foundation and measure liquidity premium by comparing a relatively more liquid to a less liquid asset with otherwise identical characteristics. However, as liquidity premia are likely to exhibit strong pro-cyclical effects, questions remain as to how well estimation methods hold under stressed conditions. Is there a single superior method or should we instead combine estimates from multiple methods to arrive at a “consensus” estimate?
According to John Hibbert, founding director of Barrie & Hibbert, the main concerns of the insurance industry and regulatory bodies are squarely focused on how companies can “prudently, credibly, simply and reliably” measure liquidity premia. This view was echoed by Richard Olswang from Prudential: he also advocates simple and reliable LP estimates, but favours an estimation method which addresses a multi currency/economy environment, and provides reliable answers in periods of market stress.
Practical issues
As reliable liquidity premium estimates need to be available shortly after the reporting dates, and because liquidity premium models need to be calibrated and monitored on a regular basis, several practitioners cited timeliness as a key concern. Furthermore, it was felt that the choice of method of calculation itself should not lead to changes or disruptions in supply and demand in financial markets (for example where regulations favour one particular asset class with specific characteristics over another). The point was also made that insurance companies should be validate estimates themselves.
Uncertainty
An interesting behavioural finance take on how liquidity should be interpreted was offered by Mark Salmon of the Warwick Business School. He suggested interpreting it as aversion to uncertainty, with uncertainty being defined as the absence of knowledge about the appropriate probability distribution and by assuming a range of possible distributions. In other words, just because things have not happened before, doesn’t mean that they won’t happen in the future.
In this context, the explanatory power of VIX, used on its own, is probably too limited to be an effective proxy for LP. Salmon proposed that we should instead obtain a measure based on the “joint” distribution of all factors likely to affect the price of liquidity, not just volatility. He also highlighted that using “investor sentiment” as a variable removes that need for different regimes. Its measurability, however, remained subject to discussion.
On the regulators…
In response to a question regarding what we should be expecting from regulators’ proposals, Richard Olswang from Prudential pointed out that regulators should look for something simple in order to avoid a situation where the regulatory answer is, “there are too many different options – so we decide the consensus is to have zero LP”. Such a situation would be the least desirable option for the industry; not least because business with long-term illiquid liabilities and a long-term investment horizon, such as the annuity business, rely on realisation of liquidity premia as a source of yield and will be significantly affected should there be no regulatory allowance for a liquidity premium. Both regulatory and industry bodies should therefore strive instead to achieve a consistent view.
Sticking with the theme of regulators, the question regarding the ways in which insurance companies could present an easily auditable LP estimation method was explored. In the context of decomposing credit spreads, John Hibbert suggested combining different approaches such as CDS negative basis; structural model; covered bond approach and a regression based method in a scorecard. The regression based approach, which also combines different measures of liquidity, was initially proposed by David Lando from Copenhagen Business School. Richard Olswang was also positive about using that approach to come up with an estimate, while Yakov Amihud of the Stern Business School, NYU, pointed out that unobservable liquidity effects could be imputed from observable measures, for example the trading frequency of a fund. Funds which hold relatively liquid assets are more likely to trade compared to funds which hold relatively illiquid assets.
Conclusion: how should we discount liabilities or value assets?
Liquidity premia permeate the asset markets in different ways but, in general, delegates felt that it is hard to argue for one unique estimation method. A common industry view is that relatively illiquid assets should trade at a lower value compared to relatively liquid assets. Illiquid long-term liabilities should be discounted at higher rate when backed by illiquid assets as the long-term investor could earn the liquidity premium. Practitioners and academics alike share a view of vigilance towards absolute estimates of liquidity premia and highlight the requirement to benchmark LP measures against multiple models.
It was a lively discussion which arrived at a wide agreement that liquidity premia do exist and can be substantial. The academic panellists, however, did question how reliable different methods of estimating liquidity premia would be in providing a good assessment of the current level of liquidity premia.
About the conference
The one-day conference Understanding and Measuring Liquidity Premia in Asset Markets took place on 12 November 2009. Its aim was to review both the established consensus and the latest thinking of leading researchers on liquidity premia bringing together academics, practitioners and policy makers to discuss the complex matters of liquidity premium estimation. The conference was sponsored by Barrie & Hibbert, the European Insurance CFO Forum, MMF and Warwick University.
Presentation slides from the conference can be found here: http://www.barrhibb.com/documents/knowledge/Liquidity_Premium_Conference_Slides.zip
Setting the discount rate for defined benefit liabilities—what’s the right approach?
Posted on 18-11-2009 by Craig Turnbull | 0 comments
Our US office recently sponsored a seminar on public pension fund valuation. The discussion focused on the 'right' approach to setting the discount rate for defined benefit liabilities - a market-based approach (based on the Treasury curve at the valuation date), or a long-term expected return on the fund's assets. This led to the usual discussions of the pro's and con's of each approach with respect to economic transparency, stability of funding levels and spurious market volatility, the dubious track record of actuaries as subjective predictors of medium-term asset returns, etc.
One of the key themes highlighted in this discussion was that these measures are really answers to different questions: a market valuation for financial reporting purposes need not mean that contribution rates are calculated every day with reference to the yield curve. However, it was interesting to observe that many in the 'traditional' pension actuarial camp were firmly against the use of market-based liability valuation in any area of public pension fund management (incidentally, we were told that the aggregate size of US Federal, State and Municipal government employees' DB pension fund mark-to-market deficits are allegedly measured in the trillions).
There was particular emphasis on the government as a special case that was an ultra long-term entity, and hence should be valued on a going concern basis that capitalized long-term expected asset returns. So does this mean that if the Federal government were to issue Treasuries tomorrow and invest the proceeds in equity funds, the Federal deficit should suddenly be reduced? If not, why should it work when the equities are put inside a government pension fund?