An effective PMS requires selection of efficient pavement key performance indices (KPIs) and precise prediction of actual pavement condition. A cost-effective PMS is possible only when maintenance requirements are identified at right time with the realistic prediction of overall pavement condition. Generally, KPI models are developed as the dependent variable are expressed as a function of independent variables. For Ontario highways, several KPIs are used in order for management decisions, such as such as Pavement Condition index (PCI), Distress Manifestation Index (DMI), International Roughness Index (IRI), Riding Comfort Index (RCI) etc. The prediction models are estimated by using ordinary least square (OLS) approach. Since these KPI models are highly correlated, estimation of these correlated models by using OLS approach might not be adequate. An approach is required which presents a joint method of estimating coefficients in generally encountered sets of regression equations. This study estimates these KPI models considering all available variables effecting pavement performance by using ‘Seemingly unrelated regression (SUR)’ method. Historical pavement performance data recorded by the Ministry of Transportation of Ontario (MTO) are used in this study. A total of sixty one road segments from Ontario’s major highway networks consisting of one hundred fifty one pavement treatment cycles are selected for empirical investigations. The KPI models are estimated and significance of the parameters are also tested statistically. After comparing these models, it is found that PCI model is highly correlated with DMI and RCI model. However, IRI model is not found highly correlated to other models.
