Clusters - A necesary investment or a Waste of Money?

So much has been written about clusters that is rare to find not  a novel approach but an approach that excites our curiosity. 
The term was introduced and popularized by Porter in the 90's in The Competitive Advantage of Nations (1990) and since then has enjoyed tremendous success among governments and innovation agencies, specially in Europe and Asia.
Many of us suspect that this success is related its ability of being operationalized or put in other terms, the temptation to try to build artificial clusters.
Anyway, governments form France, Germany, Korea, Japan, Brazil and Spain have embraced the concept fiercely. And just to illustrate what fiercely means, France spent last year €1.5B in cluster development.
Therefore the question if this is a well spent money is clearly hugely relevant.
A recent article from Philippe Martin, Thierry Mayer and Florian Mayneris from Université Paris1 and Paris School of Economics, Natural Clusters: Why policies promoting agglomeration are unnecessary tries to shed some light on the subject.
We have to say first that the title is a little bit misleading. Industrial clusters as described by Porter, and exemplified with Silicon Valley are not the result of policy intervention but emergent result of a natural process. The article refers to the question of if there is a market failure in this process and therefore it should be supported by public money in order to compensate for it, and not to the question if natural clusters exist or not.
However, the findings are pretty interisting. They model cluster advantage as providing Marshall-Arrow-Romer (MAR) externalities, namely:

1. Input externalities. Concentration of offer increasing efficiency.
2. Labor externalities. Concentration of labor offer.
3. Knowledge externalities. Knowledge spillovers, etc.

Even if you may find Porter's analysis more complete because it introduces a dynamic vision, I am confident that in order to answer the question and asses to what extend this advantages are internalized by companies when making their location decisions, this is good enough.
The study finds, not surprisingly that companies do pretty well, very close to what their model finds as optimal.
This is not the first time that we find this result and to be fair, we must say that not all policy interventions in clusters are directed to localization. For example in Barcelona 22@ is creating a cluster inside the city instead of letting the market decide where, because of urban planning reasons and in my opinion this is pretty reasonable. Also you have many cluster interventions that in reality aim to build capacity (normally research) provide infrastructure, etc. not necessarily location based but supporting existing market decissions.
This same week we have in Finland the Assembly of Competitiveness where many institutions gather to work on these issues. 
Cluster policies are in the category of innovation policies acting on the offer, together with science parks, infrastructure, technological centers and collaborative projects. 
We all know how limited these interventions are, an example of this limitation is the so called European Paradox refering to the gap between Science and Innovation in Europe or put in more simple terms, our inability to produce innovation even if we are pretty good at science.
Revisiting the research like this one and aiming at the demand side could improve, in my opinion, our chances of success because as the article says, the market does pretty well in incorporating location decisions.