I believe it's because used cars aren't commodities, and it's hard to have efficient markets with non-commodities. That's what I think, I didn't do economics. I asked gpt and it (unsurprisingly) agreed:

"Great question. Yes, you can have an efficient market with non-commodities like used cars, but only to a limited extent and under specific conditions. Efficient Market Basics

An efficient market is one where prices fully reflect all available information. This works well with commodities like gold, wheat, or publicly traded stocks because:

    The products are homogeneous (one unit is the same as another).

    Information is widely available and trustworthy.

    There are many buyers and sellers.

Used Cars: A Non-Commodity Example

Used cars are the opposite of commodities in many ways:

    They are heterogeneous (different mileage, wear, accident history, etc.).

    Information is asymmetric — sellers know more than buyers.

    The market is often fragmented.

These traits violate key assumptions of efficient markets, especially the ones about perfect information and substitutability. Can It Approach Efficiency?

Yes, under some conditions:

    Platforms like Carfax, Autotrader, or Carvana reduce information asymmetry by providing standardized inspection reports and history.

    Ratings, reviews, and guarantees help build trust.

    Large-scale data and AI tools can help estimate "fair value" more accurately.

These innovations improve transparency and comparability, pushing the used car market closer to efficiency — but it still won't match the efficiency of stock or commodity markets. Summary

So, an efficient market with non-commodities like used cars is possible in degree, not in kind. The more the market reduces heterogeneity and information gaps, the more efficient it becomes — but perfect efficiency remains out of reach.

Would you like to dive into this from a behavioral, economic theory, or marketplace design angle?"