What can we learn from economic models even if they are constantly proved wrong?

Economic models usually fail to predict the future accurately as it should.  What are the examples and why do they fail?  What are the usefulness of economic models despite they are constantly proved wrong?

You can find hundreds of models that have failed to forecast the future correctly.  One of the biggest failure of all time is the recession in 2008.  The diagram below from Bloomberg shows how badly economic models are in predicting future.  Even at Q3 of 2008 when global economies started to decline, all models failed to recognise the upcoming big recession.

What are economics models?  Economic models are just like other science models, they try to simulate the reality using assumptions and constraints to simplify the complexity of reality.  They are 'designed to yield hypothesis about economic behaviour that can be tested' (IMF).  Mathematics equations are often include in economic models about how economic agents behave or how economic works.  Economic models under climate change mitigation are empirical models.  They verify the prediction of theoretical models and convert these prediction to precise outcomes.

Stern Review mentioned some existing economic models .  For example, the 'Tol' model (Tol, 2002) estimates the impacts for different market and non-market sectors e.g. agriculture and ecosystems.  It predicts global impacts become negative beyond 1'C beyond increase in global mean temperature and are up to 0.5-2% of global GDP.  However Stern criticised that these 'old' models did not capture the full cost of climate change, including parameters such as extreme weather events and political factors.  

The model Stern uses, PAGE2002 IAM (Hope, 2003), 'take account of risks of very damaging impacts as well as uncertain changes over very long period' (Stern, 2006), to calculate the overall welfare cost of climate change  (Although the studies are somewhat outdated, they are cited as authoritative).  The estimates differences in cost are approximately a quarter higher than those old models.

No models are perfect.  This is simply because the reality is too complex to be simplified and predicted perfectly.  Unavoidably there are critics questioning the accuracy of the IAM models.  Rosen (2014) says we should 'stop trying to asses the long-run economics of mitigating climate change since that is unknowable'.  This is due to the fact that the way that firms minimise cost, the well-being of people and the advance of technology changes over short to medium term and therefore no point in predicting the economy in 100 years time.  On the other hand, Ackerman et al. (2008) suggested that Stern underestimated, even that is what Stern emphasised, the potential costs of climate change damage.

Why do we still care about the predictions?  Even if these models are proven wrong, we could still take some positive away from them.  At least we know we have to act now, despite not knowing the scale of the impact, to combat climate change.  Or else the cost of it will exaggerate.  Moreover, as Rosen says, instead of trying to predict the future, it is better to 'focus on the details of how to mitigate climate change'.