Economy & Loans Mentor

A less obvious partnership would be in the development of new products, services, or technologies designed to meet future customer needs. New partnerships offer the opportunity to be partners by design. These new partnerships offer more options because the organization is limited only by its imagination. But selecting the right partner can be a daunting experience if key decisions about the purpose of the partnership are not clear from the beginning.

A systematic approach to finding a partner works best. Generally my clients have some concept of who they might want for a partner. From our previous example, Bank of America knew they wanted to outsource the transactions in their human resources department. Businesses are familiar with the market, the suppliers, and the vendors. But first they must identify the need. Once the need is clearly identified, the assessment team should brainstorm a list of potential partnering candidates to fill it.

Robert called Peter at the city’s Building and Construction Trades Council. Peter is responsible for negotiating partnerships between the local union building trades and building contractors. Having worked with Peter in the past, Robert knew that he and Peter could figure out a way to meet each other’s needs. Robert set up a meeting. In most cases, an organization would conduct a needs assessment to figure out what they needed in a partnership. But in this case, both Robert and Peter knew exactly what they needed from each other. They had partnered before on many successful projects. Robert’s primary concern was to have the building completed on time. Second to this he needed to meet the tight budgetary guidelines given to him by the client. Naturally the budget proposal sent out requests for pricing to many different subcontractors.

Since pricing was a concern, he knew he needed something more from Peter than just a time commitment. He needed to be sure the bids were competitive. He wanted to use union labor, but the price had to be right. At the meeting, Robert began to express his needs. If he and the unions were to be partners, he needed a no-strike agreement and a commitment to provide the skilled and licensed labor when required— and the bids needed to be competitive with nonunion labor. Peter’s top priority was simple: a union work site with all-union labor. He also needed to be sure that negotiated wages and benefits were paid to the trade members and that management followed union work and safety rules. Based on their discussion, both Robert and Peter determined that they could satisfy each other’s needs. They agreed to explore what it would take to partner on this project.

The city is in the midst of a building boom.All over downtown, the landscape is changing as office space gets tighter and towers sprout up on once-empty street-level parking lots. One of the city’s largest,most prestigious contractors has just completed negotiations with a legal firm to build a new office tower. The location was selected, and preliminary architectural design work was completed.With the tight budget and time constraints, how could Robert, the contractor, get the job completed on time and on budget? The legal firm was currently leasing fifteen stories, and their lease was up at the end of the following year. They needed to have their new building finished by then.

Once the deal was signed, Robert went to work scoping out the project plan and timeline.Although he had completed many projects in the past, this one was special. The magnitude of the project, the prestigious location, and the scale of the building put additional pressure on him to make sure the project went off without a hitch. Constructing a thirtyseven- story building is a complex task. Coordinating the numerous variables can strain the best project manager’s Partnering Intelligence. But nothing is quite so critical, or as potentially stressful, as the labor issue. Not only is labor one of the biggest expenses, it could be the component most easily mismanaged. Robert remembered one project when an electrical union had walked off the job.Not only did the strike shut down the entire project for over a month, but once it was settled, it took three weeks just to reorganize the project timeline—think of the headache caused by all the building materials arriving with no place to store them. It cost everyone a bundle in lost time, money, and goodwill.

What conclusions can be drawn from these observations? Portfolios that are optimized with regard to risk seem to do better than those that additionally rely on return forecasts. Sample means are not a good predictor for future returns of bond portfolios. In this context Jorion (1986) argues, that the sample mean is exposed to considerable estimation risk whereas variances and correlations are generally more stable over time. The high influence of the return vector on the weights of the tangency and equal risk portfolios can lead to extreme and volatile portfolio returns. Kallberg and Ziemba (1984) point out that errors in estimating future returns have a ten times higher impact on the out-of-sample performance of the optimized portfolio than errors in estimating the variance–covariance matrix. Similar conclusions apply when using adjusted value at risk or lower partial moments as a risk measure.

To simulate a realistic decision-making process we apply a backtesting methodology. Sample means and risk measures are estimated for the subperiod January 1987 to September 1998. In those cases where the return series is adjusted for autocorrelation the estimation of the desmoothing parameters is also based on the shortened sample period. Then the portfolios are optimized. With the obtained weights the series of monthly out-ofsample returns for each portfolio is calculated for the period October 1998 to September 2003. Based on this out-of-sample period the risk-return characteristics of the optimized portfolios and the benefits of the optimization approaches are examined.

To determine the set of efficient portfolios, pairwise tests for secondorder stochastic dominance are conducted. When borrowing and lending are allowed and considered in the stochastic dominance algorithm the efficient set contains only one portfolio. It was obtained by minimizing portfolio risk in a shortfall risk framework. This portfolio does not only dominate the performance of the other optimized portfolios but also of every single asset class. Obviously the approaches that take into account skewness and kurtosis of the return distributions yield portfolios with a superior risk-return profile. Three more portfolios that result from using the shortfall risk or Corning–Fisher approach are contained in the efficient set when borrowing and lending constraints are imposed.

Furthermore it should be noted that adjusting for autocorrelation during the process of portfolio construction has a positive effect on the out-of-sample performance of the chosen portfolios.

Our study demonstrates that the increase of high-yield volatility effected by the desmoothing of the monthly return series moves the asset classes and the efficient frontier in a risk/return chart to the right. Evidence shows that consequently the composition and the risk-return profile of the optimal portfolios also change. While the weight of high yield reduces by 2–13 percentage points, mortgage-backed securities and investment grade corporate bonds become more important in a portfolio context. In the ERPs diversified corporate bond sub-portfolios receive a weighting of about 75 percent. Except for the portfolio based on the mean–variance framework, investment grade corporate bonds are also part of the TP. Investors looping for a particularly attractive risk/return profile or willing to maximize their return for a given level of risk, seriously should consider corporate bonds as an alternative. Even for investors who are averse to negative skewness and leptokurtosis – as modeled by the Corning–Fisher approach – the admixture of a small high-yield portfolio can improve the risk/reward profile of their fixed income portfolio.

Again, the risk/return characteristics of the optimized portfolios are summarized in our study. We noted before that in comparison to a pure government portfolio the presented portfolios offer a more attractive investment opportunity. This observation is supported by the results of the stochastic dominance algorithm. As in the case without adjustment for serial correlation the government portfolio is dominated by all optimized portfolios. The results of the stochastic dominance algorithm also indicate that after the desmoothing process only one portfolio is excluded from the second-order stochastic dominance set: the minimum risk portfolio based on the Corning–Fisher approach.

So far three methodologies for a quantitatively driven process of portfolio construction have been examined. The results of the empirical study show that the way of accounting for skewness, kurtosis and autocorrelation in return series has significant impact on portfolio weights. Measured by riskadjusted performance numbers as well as stochastic dominance criteria, the identified differences between the risk/return profiles of the optimized portfolios seem marginal. However, it has to be considered that so far the process of portfolio optimization and performance measurement has been based on the same sample period. Implicitly perfect foresight has been assumed. There is no doubt that this assumption does not reflect a realistic investment decision-making process under uncertainty. In practice, first a decision about asset allocation is made and implemented. And afterwards the risk and return characteristics of the portfolio are measured either in absolute terms or relative to a benchmark index.

The implementation of the desmoothing methodology increases portfolio risk. We already mentioned that there is a permanent and relatively stable component in bond (index) returns resulting from interest accrual, roll down and yield-curve effects. Therefore we apply the desmoothing technique only to the series of monthly changes in the price indices, and afterwards add the other three components to derive the adjusted return series.

It can be stated that the higher the degree of first-order serial correlation in monthly changes of the price indices the more portfolio risk surges. Our study shows that the high-yield sector is most exposed to this effect. Interestingly though, in mean–variance space the risk/return profile of government bonds is dominated by mortgage-backed securities and agencies before as well as after adjustment for serial correlation. That said, it is no surprise that government bonds do not receive a significant weighting in any of the optimal portfolios.

For the sample period the first-order serial correlation coefficient for the Merrill Lynch US High Yield Index is 0.26. Brown (1985) notes that 0.25 represents the upper bound for the serial correlation that may be introduced by a valuation process characterized by nonsynchronous trading. Even after considering the effect of relatively stable components in index returns such as interest accrual, roll down and yield-curve effects it can be claimed that the US high-yield market suffers seriously from illiquidity. The desmoothing process described above increases the volatility of the high-yield index by considerable 63 bp per month.

The VIX Index reflects the equity and options markets’ expectation of earnings volatility. Companies with deteriorating credit statistics are more likely to experience high equity price volatility than companies with a stable credit trend. As financial profiles of companies improve and the uncertainty about their future earnings declines the hedging costs to invest in those companies will come down and the VIX Index will decline.

High-yield market sensitivity to changes in VIX varies over time and it does not vary simply by ratings. The correlation will always increase when an increased leverage in the whole credit market can be observed. This is usually the case when CCC-rated companies have a high share of the credit market. So it can be stated that the CCC portion of the market drives the sensitivity between high yield and the VIX Index. A decline in implied equity market volatility results in lower risk premiums in the high-yield market.

The correlation between the VIX Index and high-yield spreads was only moderate at 0.45 during the period Dec. 1996–Nov. 2003. Ahigh correlation of 0.83 can be measured for the period Aug. 2000–Nov. 2003.