Computational grid provides virtual powerful computer for solving any large-scale scientific, commercial, and engineering problems by aggregating and sharing heterogeneous resources. However, designing an effective resource allocation method is a complicated task. To undertake the task, several economic-based resource management models have been studied. Bargaining (or negotiation) is one of the most used and effective models even though it has several main defects such as high communication demand and the risk of losing a deal. Instead of repeating negotiation for every generated grid job (GJ), the novel contribution of threshold-based negotiation framework is to determine whether a new negotiation is needed or not, according to the current market conditions which depend on supply of resource providers and demand of consumers. In the proposed framework, agents request for renewing negotiation, provided that the amount of change in market conditions exceeds pre-defined threshold. Therefore, communication overhead and the risk of losing the deal are minimised by avoiding unnecessary negotiations.