Marsoft

Marsoft provides forecasts for major market segments in the dry bulk, tanker and containership markets. These forecasts span a period of five years (on a quarterly basis) and cover a specific spot rate for each market segment as well as one-year time charters, newbuilding and secondhand prices. We have published forecasts since 1985, although the statistics herein refer to the period from 1990 for our dry bulk and tanker forecasts and from 1999 for our container forecasts. In addition to our Base Case (or most likely scenario), we also provide High and Low Case scenarios to capture selected elements of market risk.

The statistical models we use to forecast rates and prices based on input supply and demand factors generally explain 80% - 95% of the historical variation in rates and prices (given the supply and demand inputs). The models are re-estimated every few years to account for changes in market structure or the availability of new data. The model details are described in the technical documentation available under separate cover.

Providing a single simple measure of forecast accuracy covering all those markets, at all horizons, and for different forecast variables is challenging. We believe the attached summary of the performance of our Base Case forecast of one-year time charter rates, focusing on the accuracy six months after the forecast is published, is representative of Marsoft’s track record.

A key issue with regard to interpreting forecast accuracy is to use the right benchmark. While it is important to compare the forecasts to actual market developments, it is also worthwhile to consider how the accuracy of the Marsoft Base Case forecast compares to the accuracy of alternative forecasts available at the time.

Of course this comparison is made difficult by the fact that we do not have access to a historical database of alternative forecasts. One alternative benchmark that is available, however, is to assume that rates are simply equal to the historical average of rates prior to the date of the forecast. This kind of moving average outlook does not account for any fundamental analysis, but it can be interpreted as a kind of “judgment free” forecast.

The attached figures compare the performance of Marsoft’s Base Case forecast to the performance of a forecast that sets future rates equal to the ten-year average of rates prior to the date of the forecast. The following page evaluates the accuracy of each forecast across the dry bulk, tanker, and container markets combined, and the subsequent pages examine each market individually. We compare each forecast on two different measures of accuracy; distribution of forecast errors and turning point accuracy. The first measures the percentage deviation of each forecast from the actual TC rate six months later, while the latter is a measure of how precisely each forecast is able to predict the direction of the market on a 6-month ahead basis.

On both of these measures, the accuracy of Marsoft’s Base Case forecast clearly dominates that of alternative technical forecasts based on historical rates. This is particularly relevant for credit rating purposes, as typical rating exercises rely upon an assumption that markets in the future will, on average, match markets in the past.

Marsoft Track Record for Dry Bulk, Tanker and Containership Markets

November 2006

The graph below plots the distribution of errors of Marsoft’s 6-month ahead TC rate forecasts against that of the 10-year historical average. Over the period 1990 – Q3 2006, our forecasts fell within 10% of the actual TC rate 55% of the time, compared with just 18% of the historical average forecasts. Similarly, in 78% of our forecasts we predicted the TC rate to within 20% of the actual value. The historical average achieved that level of accuracy just 35% of the time. In addition, our forecasts are much more powerful in times of extreme rates, such as the last couple of years. As the graph below demonstrates, the historical average is much more likely to substantially underestimate rates than are our forecasts. In fact, 23% of the historical average forecasts underestimated future rates by more than 50%. This figure falls to just 3% for Marsoft’s forecasts.

Table 2 illustrates the accuracy with which each forecast was able to predict rising or falling rates over a 6-month forecast horizon. Marsoft accurately anticipated the direction of the market 78% of the time, while the historical average correctly predicted the direction of the market just 47% of the time. Remarkably, increasing the forecast period does not diminish the ability of our forecasts to predict future market trends. On a 1-year ahead basis, our forecasts achieved a turning point accuracy of 78%, while the historical average correctly projected the direction of the market 50% of the time.

Distribution of 6-mo. Ahead TC Rate Forecast Errors for Dry Bulk, Tanker and Container Markets Marsoft Base Case vs. 10-Yr Historical Average

Percentage of Errors Within +/- 10% of Actuals

Marsoft Base Case vs. 10-Yr Historical Average

	Marsoft	Historical Average
6 Mo.	55%	18%
1 Yr.	29%	18%
2 Yr.	22%	18%

Marsoft Forecast Track Record