A big part of the reason that Robert Shiller is one of the best writers and thinkers on the subject of housing is that he's one of the few people to consistently and rigorously differentiate between the price of a house and the price of land. And in a recent op-ed, he makes the excellent point that since most of the price of the typical home purchase in America is the price of the house, you should expect house-buying to generally be a bad investment strategy. A house is like a boat. It's a durable asset, but it's going to depreciate.
But as Adam Ozimek argues, Shiller seems to go too far by arguing that land never sees sustained systematic price increases. Ozimek pulls a chart from an Econsult report on the Philadelphia housing market that pretty clearly illustrates the existence of sustained changes in the relative price of different neighborhoods in the Philadelphia area. And it's hard to see how this could be any other way. The opening of Green Line Metro stations in Petworth, Columbia Heights, U Street, and Shaw objectively increased the desirability of nearby parcels of land in Washington, D.C. The invention of the automobile objectively decreased the desirability of living within walking distance of downtown Cleveland.
America, as a whole, is not a particularly crowded country, and land in general is not in short supply. But that doesn't mean that land in particular places can't be in short supply in economically relevant ways.