【正文】 in this study to describe the change in overall construction costs in the future. The formulation of the index is described later in the paper. Past Increases in Construction Costs When the change in overall construction costs in the past is observed（ as measured by popular construction cost indices） , it is apparent that they change significantly from year to year and that the changes can sometimes be quite erratic. The mon assumption that construction costs change with the rate of inflation can lead to poor estimates of future construction cost. To illustrate, the Federal Highway Administration’s Composite Bid Price Index, an index of overall highway construction costs, is plotted in Fig. 1 together with the Consumer Price Index （ CPI） , a mon expression of general inflation. The FHWA CBPI for the entire nation and for Louisiana alone is plotted in the diagram. All indices have been normalized to a value of 100 in 1987 for parison purposes. From the diagram, it is clear that highway construction costs change erratically and even display different short and longterm trends from to those of the CPI. It is also apparent that construction cost changes are different in Louisiana from those in the nation as a whole. While not shown here, review of the FHWA CBPI from other states shows that many of them show a deviation from national values. Past Methods of Forecasting Highway Construction Cost Forecasting future highway construction costs has been achieved in basically three ways in the past. First, unit rates of construction such as dollars per mile by highway type have been used to estimate construction costs in the short term. However, this method has generally been found to be unreliable, because site conditions such as topography, in situ soil, land prices, environment, and traffic loads vary sufficiently from location to location to make average prices inaccurate estimates of the price of individual projects or even of all projects in a particular year. Second, extrapolation of past trends, or timeseries analysis, has been used to forecast future overall construction costs （ Koppula 1981。 Hartgen et al. 1997） . Typically, construction costs have been collapsed in these analyses to a single overall expression of constructioncost such as the FHWA CBPI or the Engineering News Record’s Building Construction Index ~ENR BCI! or Construction Cost Index ~ENR CCI!. However, these types of models are usually only used for shortterm forecasting due to their reliance on the notion that past conditions are maintained in the future. Third, models have been established that describe construction costs as a function of factors believed to influence construction costs. The relationship between construction costs and these factors have been established from past records of construction costs. Typically, the models established in this manner have been used to estimate the cost of individual contracts. These models, with their relational structure, are the only models expected to provide reliable longterm estimates. The model developed in this study is of this type. Proposed Construction Cost Model It is clear that there are numerous factors that affect construction costs. However, it is striking that most construction cost models developed in the past have used only a few of the many influential factors identified above. One reason for this is that information is generally not available on many factors in data sets used to estimate models. Another reason is that information on the qualitative conditions surrounding each contract is difficult to obtain. These are problems that prevail in most circumstances and are difficult to overe. To mitigate against the effect of an inplete set of factors, two strategies can be employed. First, it may be possible to represent some of the absent factors by surrogate variables that are in the data set. For example, as mentioned earlier, annual bid volume has been used in the past as an inverse measure of the level of petition prevailing in the construction industry at that time （ Herbsman 1986） . Similarly, the number of plan changes each year can serve as a measure of design quality. Second, if the modeling of construction cost is changed from estimating the cost of individual projects to estimating overall construction costs each year, the modeling task is simplified. This is because it is no longer necessary to try to model individual projects in which conditions inflate the price in one case and deflate it in another, since such conditions would tend to cancel themselves out among projects in the same year. For example, firms that reduce their bid prices in an effort to win a particular contract could be balanced out within the same fiscal year by those that increase their prices because they already have enough work and are not particularly interested in winning the contract. Similarly, those firms with expertise in the type of construction required will be balanced out by those with low levels of expertise in that area. Thus,