We all know what we mean by solving a model. We mean taking the unknowns and making them known. In other words, solving a model is expressing each and every endogenous variable in terms of the constants and exogenous variables of the model (because their values are known). The solution of a model is also called the prediction of the model. For example, if you design a model to analyze the behavior of national consumption (endogenous) in terms of national income (exogenous), then the solution of the model is what that model predicts will happen to consumption if income changes. In this particular case prediction is not really about the future. It is just about the solution of the model.
A good model is the one that can be repudiated. Someone should be able to take your model and confront it with the data and conclude that it is wrong, in the sense of giving incorrect or unrealistic predictions. If you cannot test a model against reality, how can you tell if it is useful or not?
