Lost in many of the post-crisis financial reform proposals to rein in destructive financial innovation---such as calls to ban naked CDS, establish centralized clearinghouses for derivatives, and eliminate high-frequency trading---is the broader issue of whether these innovations could actually enhance welfare. The Economist recently took up the challenge by hosting a debate between two academic bigwigs: Brown's Ross Levine, an empirically-oriented financial economist who has spent much of his career studying the relationship between financial development and growth, and Columbia's Joe Stiglitz, an economic theorist who won the prize for his pioneering work on asymmetric information, including seminal work on credit and financial markets that was cited by the Nobel committee in its award of the prize to Stiglitz.
Much of the disagreement between the two, where it exists, boils down to their respective views of whether these innovations spur an expansion of the real economic pie. The tradeoff appears to be between the gains from that result when financial innovations help establish the efficient market-clearing price level, which in turn raises welfare (by the second fundamental theorem), versus the social value of these innovations, many of which appear to have only served to contribute to financial sector bloat and increase banker compensation. This is the efficiency versus equity debate all over again, recast for modern finance.
I have little to add to their excellent arguments on each side, although I'd note that the debate did drift a little from the motion ("This house believes that financial innovation boosts economic growth") to a debate over whether recent financial innovations are good for economic welfare. To which I would point out that this raises three slightly more subtle issues about the nature of recent financial innovations.
First, one needs to concede that it is virtually impossible to figure out, ex ante, which innovations would actually generate real economic growth, versus those that mainly serve to transfer resources between agents (and so are neutral for the aggregate economy). Given this fundamental uncertainty, many would argue that it is perhaps best to leave Mr. Market to figure out the innovations are genuinely beneficial---such as venture capital financing---and weed out those---such as CDO-squareds---that are essentially not (an analogous argument to the case against "picking winners" in industrial and trade policy).
Second, this market-driven selection of beneficial financial innovations does not really require the best and the brightest to work in finance. After all, hyperintelligent agents are neither necessary nor sufficient to guarantee financial market efficiency. They are not necessary, since even zero intelligence traders have been shown in laboratory markets to be able to achieve (allocative) market efficiency. Nor are they sufficient, since complex financial products are stupendously difficult to price, and can outwit even the smartest traders equipped with the most powerful computers (one could also make a snarky remark here that the cluelessness of some bankers as to the operations of their own institutions prior to the crisis probably casts doubt that those working in the financial sector are, in fact, the best and brightest).
Third---and most important---even if it were the case that increased efficiency does enhance welfare in a hypothetical frictionless world, there may be complications in reality that call the efficiency gains into question. Consider, if you will, the idea that hyperefficient pricing (with millisecond speed) is desirable because it enhances price discovery and contributes to overall liquidity. Although clearly beneficial from the point of view of improving the efficiency of the financial market in question, it is less clear whether this is immediately good for complementary markets, such as markets in the real economy that utilize financing for production purposes but sell their output in slower-moving goods markets (or hire workers from even more slowly-adjusting labor markets).
More generally, welfare gains to allocative efficiency are only realized when resource reallocations are complete. But if sticky prices and wages mean that these reallocations take place relatively slowly, then having allocations "outrun" resource transfers could raise the short-run welfare losses that result from unemployed resources in the real economy. While gains from intertemporal trade are ultimately obtained from a well-functioning financial market, the relevant welfare criterion would appear to be an evaluation of the long-run gains from the efficient allocation of resources, versus the aggregate short-run losses from temporarily unemployed resources. Throw in the fact that increased volatility is likely to be detrimental to welfare, and the case for hyper-efficient financial markets becomes even harder to make.
At the end of the day, a large body of evidence does in fact tell us that financial development is good for economic growth and welfare. However, the optimal level of financial development in any given economy may well be endogenous. Developing countries with immature financial sectors may benefit from a more measured, tiered approach as they move toward greater financial liberalization. A full-speed ahead strategy is likely to impose unnecessary adjustment costs on countries that are neither ready nor require the most sophisticated financial instruments. Developing countries can also take advantage of second-mover advantages and benefit from demonstration effects in deciding whether to approve or adopt a particular class of financial products.