Even more importantly, the notion of a "positive feedback" which underlies the concept of path dependency put the neo-classical dogma in the dust bin (kind of). The neo-classical theory depends on the concept of a negative feedback i.e. the more of something you get, the less effect it has - which underlies its concept of diminishing returns, which in turn makes the concept of self-regulating markets possible.
For example, in the neo-classical view, the more widgets are being produced the less profitable the production is, because the price drops as the supply increases. This negative relationship between supply and price (and profitability) creates a "feedback" or causal relationship that creates natural limits for widget production, defined by the cost and benefits balancing each other out.
By contrast, in the path dependency approach, transaction costs (which include anything from transportation infrastructure, to supply of skilled labor, to technical standards, and to contract enforcement) is a key component of profitability. In other words, high transaction costs may render production totally unprofitable. Lowering those transaction costs, in turn, may change things around. Better yet, selective transaction cost lowering can make otherwise costly and cumbersome products more "economical" than their less costly and better designed counterparts.
For example, the total cost of car-based transportation by far exceeds that of the next available alternative - rail. For every mile of freeway, four or so miles of tracks can be build. Rail traffic flow is much easier to manage than highway traffic flow, which results in less congestion. The per passenger cost of rail transit fuel and maintenance is lower than that of car based transit - not to mention their environmental impact.
In short, the total cost of car based transit is several times higher than that of the rail-based transit, but the product in both cases is essentially the same, moving from point A to point B. Therefore a rational actor would certainly choose rail over car. However, that situation can change quite dramatically when some of the car-related transaction costs are reduced, say by government subsidies. If the government builds the roads and lets them use for free, if it pays for the cops to enforce traffic regulations, and subsidizes the cost of fuel - the direct cost to the individual car user drops quite dramatically to the point of making it more "economical" than otherwise more economical rail.
What is more, selective reduction of transaction costs does not have to be achieved by a purposive action of the government, it can often result from the economies of scale and externalities. For example, if a mfg plant opens in a certain location, this may spur certain externalities, such as the development of transportation and communication systems, education systems to train labor, public services resulting from population increase, etc. Those externalities lower the transaction cost for the subsequent plants in that area, and thus induce the new plant owners to locate in this area rather than somewhere else. That lowers the transaction costs even further, attracting even more industries of the same kind. This is called "path dependence" created by a positive feedback resulting from lowering transaction costs.
Or take another example. The Internet Explorer was an inferior product to Netscape, but was "bundled" with the Windows OS i.e. the transaction of cost of acquiring and running this utility was close to zero (as opposed to Netscape, which had relatively higher transaction cost in the form of downloading, installing and configuring the OS). This factor alone was sufficient to establish Micro$oft monopoly in the browser market and kill an otherwise better product.
The bottom line is that selective changes in transaction cost structure remove the "natural" (in neoclassical approach) barriers to the supply of certain products set by their prices and profitability and allow the supply of these products to grow virtually unchecked to dominate the market. They dominate not because of their superior design, rationality or utility but because the transaction cost of their manufacture and distribution was selectively lowered comparing to that of their alternatives. In other words, they dominate because of the benefit of path dependence (or "head start") not because of their superior efficiency or utility. Ditto for policy decisions.
It is important to underscore that path dependencies are the norm not an aberration in human and organizational behavior. Therefore, a lion share of that behavior can be explained by "path dependencies" i.e. the factors affecting the decision making process (such as transaction cost structure or ease to reach an agreement) rather than the resultant utility of the outcome - a fact that eludes assorted conspiracy theorists who see evil intentions each time somebody farts.
Wojtek