Side note first: I don't think that physicists always agree on their theories and predictions. As chemist I can say that chemists certainly do not always agree. And, also for the hard sciences, finding the truth is not so easy  .
But to your question: I think you may be interested in Tomáš Sedláček's "Economics of Good and Evil". The book follows the idea that economic world views are quite fundamental beliefs: he speaks about the world views (myths) and says that also the statistical models employed in econometrics incorporate these world views.
The other thing is that while the "hard" sciences rely very much on controlled experiments, this is much less the case in macro-econometrics. And IMHO it would be either unethical to perform macro-economic controlled trials (you may count communism/socialism vs. capitalism/free market economy in the 21st century as such an experiment, but how free were the participants to participate? And how well randomized was the trial?) or unpractical (who'd sign up for a controlled, randomized makro-economical trial? Where should it take place (buy half of antarctica, maybe?). So the starting conditions for this science are that you have your set of beliefs (of which you are more or less aware), you have noisy observational data (and no controlled experiment, hardly any prospective study, you need to do mostly with retrospective analyses), and you have a possibly mixed set of "believers" (in different [economic] world views) in the society you study that act more or less accordingly to their world view.
One point where the hard sciences are at a huge advantage is that they can (with a few precautions) rely on having a much lower feedback (i.e. the "experimenter" influencing the outcome of the experiment). Predictions and independent testing are possible. Even the weather-forecasters can have enough tests to judge whether they are doing well or not at forecasting. However, with macro-economics, things don't work out that way. If a macro-economist makes a prediction, it can in theory be right or wrong, but in addition, the society may take it into account and act accordingly (say, to avoid the predicted problem). So you end up with situations where you can either argue that the prediction was wrong - or that the prediction was right, but as society acted accordingly, the problem was avoided. For the science part that means that no conclusions about the correctness of the prediction can be drawn.
Yes, it would in priciple be possible to make a prospective macro-economic study: people should hand in their predictions in a covered way, never talk about them, and after a previously specified period of time the envelopes with the predictions are opened and compared to what really happened. One could then have a look how many predictions were correct. However, even that would not solve the problem of concurring macro-economic world-views, as societies that implement the policy in question need to be "available" to check the predictions against.
In addition, macro-economics usually doesn't change that fast. How many possibilities has even a perfect macro-economist to make interesting (≈ information gain by the prediction being fulfilled) predictions during a human lifetime? If the prediction period is too short, the prediction is not very interesting (obvious). If the period is too long, the event may occur for purely random reasons. If the event didn't occur till now, who guarantees it won't in the future? And would that be far-sighted or random? Now, add to this situation a multitude of persons who do predict this or that, so we'd need to filter out those predictions that truly bring information gain for the science of macro-economics from the noise of accidentally correct predictions.
All together I think that we're looking at a very noisy system here, which means that "convergence" to the "truth" should be expected to be very slow. And by the way, I don't know whether there is only one macro-economic truth: I can very well imagine that there exist different sets of economic rules that lead to "solutions"*.
Which is why I think that this is a politics question.
* Solutions being e.g. prosperous/pleasant to live in/... economic systems - put in other characteristics - which underlines the ethic dimension Sedláček insists on: the "solution" would be a "good" ("the right") economic system, and good (with the superlative "the right") is an ethic category.