I dont think the CAPM model is dead. The model still seems to be most logical in explaining to a novice about the working of capital markets. It helps him/her in explaining why some stocks are priced higher in comparison to other. I agree that in practical sense there is little correlation of CAPM with the real market. But since most people agree that markets cannot be mathematically modeled very accurately, therefore considering this CAPM does not perform so poorly. Personally I have found that students after understand the model and after learning about its shortcoming become motivated to carry further research in the area. This is thus another vital importance of  CAPM as it acts as first step to understanding markets.

These articles might be of interest to you.

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2505597

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2523870

I don't think the models die. Sometimes models are used for understanding how things work in the real world, and simple models do this job better than concise yet complicated ones. CAPM is one of such simple models.

I thank each of you for the interesting considerations, pro and against the CAPM

And one delayed comment from me: Indeed, theoretically the CAPM is robust but empirically week. One of the most acclaimed critiques is perhaps the Roll's on the impossibility of creating a trully diversified market portfolio - one of the main independent variables in the model.

The other alternaitves models to the CAPM are the Fama-French three-factor model (1992), the APT, that resembles a more general version of the FF model. Basically, we all  know a model is as good as it assumptions are. 

Currently, I am working on a paper in which I try to evaluate the cost of equity and WACC for 5 listed banks in the periferial EU economies and estimate how CoE and WACC relate to the change in accounting capital ratios (I test the effect of another major financial model -- MM theorem). I use CAPM to get the past returns on equity. 

Merry Christmas to you all.

I mostly agree with the preceding contributions. In my opinion CAPM is an excellent mathematical development based on weak assumptions. The crucial point, I believe, rests in the "expectations" characteristics of the quantities in  play. Expectations cannot be treated as mere averages of the past - otherwise risky returns less than the riskless rate could hardly be explained. Every agent has his/her own expectations, this prevents any attempt to develop a positive theory of rational market returns.

In my advice stochastic processes are excellent in simulating the motion of particles in fluids; when it comes to represent human actions such as buying/selling goods or assets, they simply fail. Think e.g. of the 2007-08 panic: how could one express any reasonable expectation for the future? And also Behavioural Finance doesn't provide any solution to this puzzle - at  most, pinpoint "anomalies".

If I could answer I would deserve a Nobel prize!

CAPM is theoretically strong and it provides a simple explanation of risk-return relationship. But due to its simplicity (one factor model), CAPM may not explain how the real capital market works.

Any model that allows investors to accurately forecast the value of financial assets will either crash the market or be crushed by the markets. That's because if all investors would start using the model, the market would bubble and crash.

Surely the CAPM is a robust "theoretical framework" that has expanded the horizons of research, and knowledge, on the asset pricing. Despite the weakness on the empirical level, it has inspired the development of empirical models multifactorial. Without the CAPM perhaps would not exist the strand of behavioral finance.

Any theory / model has its deficiencies and over time, and will motivate other researchers to build and improve on it. CAPM is no exception. However, as others have suggested, it is a good starting point to explain risk return behaviour.

Well, CAPM is so easy and useful, if you are looking for a more robust model you can use APT, we can go beyond discussing about symmetry or efficiency into the market and it is true CAPM has some weaknesses, but is easy, fast and cheap to calculate it, so keep that in mind. I can recommend you to check the Downside Capital Assets Pricing Model D-CAPM, it works better for  emerging markets  

CAPM and APT rely on second order statistics. When contrasted with other valuation models which assume complete knowledge of underlying Probability spaces, CAPM and APT surely lack completness in this sense, and one can easiliy provide flaws within  CAPM pertaining to no-arbitrage as well as to the alleged positivity of the implied 'market weights' implied by it. However, for  empirical work correlations linked to second order statistics are definitely a PLUS to get some educated guesses as to where the market is heading, in particular when linked to some judiciously-chosen windows to produce their estimates as can be seen from the old Riskmetrics(TM) document(s). Besides, their teaching, even without normality,  is a reasonably simple and certainly useful tool to understand at an elementary level (second-order properties only) the risk-return relationship, and as a prelude to higher and more complete treatments with assumed complete probability spaces that are MUCH MORE difficult to calibrate in practice and which raise further complications in empirical work.  I do agree with all previous judicious comments . In particular with Juehi Shi, in that we simply cannot predict the future. The great Physicist Niels Bohr is known to have said (this is intended as a pun) 'It is diffcult to predict. In particular, the future'.

I think that this paper provides some interesting viewpoints regarding this issue:

David Nawrocki, Fred Viole- "Behavioral finance in financial market theory, utility theory, portfolio theory and the necessary statistics: A review", Journal of Behavioral and Experimental Finance, 2 (2014) 10–17.

We have studied CAPM whether it's linearity holds in extreme conditions or not. We found that in come cases the nullhypothesis of CAPM linearity should be rejected. You can find more details here: https://www.researchgate.net/publication/227358232_Non-parametric_and_semi-parametric_asset_pricing

We also studied the capability of capturing risk in one meausre and predicting future returns for securities. We compared one factor models, like standard deviation, CAPM beta, and additional two entropy based methods. Based on the results we found both advantages and disadvantages. CAPM beta is good for explaining returns in short term, but it less effective predict them for the next period because the value of CAPM beta for each securities is changing with a high variance over the time. More details here: https://www.researchgate.net/publication/270221149_Entropy-based_financial_asset_pricing

Overally I think the theory is usable, however it has weaknesses. It can be a good basis for a more efficient multi-factored models, like Fama-French model (SMB-HML), Carhart model (SMB-HML-Moment) or additional factors like liquidity or entropy.

Article Non-parametric and semi-parametric asset pricing

Article Entropy-Based Financial Asset Pricing