Why Last-Price Forecasts Can Perform Well in Efficient Markets

Summary

The document raises a forecasting question: why a naive model that uses the latest observed spot price may outperform more elaborate forecasts. It proposes two possible explanations for consideration: market efficiency, linked in the question to forward prices as unbiased estimates of future spot prices, and a martingale model, in which the conditional expected future price equals the current price.

The central learning point is that these are related but distinct ideas. Market efficiency concerns whether available information is reflected in prices, while a martingale is a property of a stochastic process under a specified information set and probability measure. The text contains no answer or empirical analysis, so it does not establish that either explanation holds for the market in question, or that the proposed link from forward prices to the latest spot price is valid. Forecast comparisons also require care about the target, horizon, data, and evaluation metric.

Key ideas

• A last-observation forecast can be a strong baseline for spot prices.
• The question presents market efficiency and martingale behavior as candidate explanations.
• Market efficiency and the martingale property are distinct concepts, even when they are connected in a model.
• The document offers hypotheses but no evidence that either applies to a particular market.

Tags

# Good performance of naive forecasting in efficient markets


# Good performance of naive forecasting in efficient markets












I am doing spot price forecasting for a market, and so far, the naive forecasting model, which forecasts with the last observed prices, is the best forecasting model. I know that it might be because of my poor forecasting skills; however, how can this situation be explained if it is not the case?

So far, I have come up with two potential explanations:

1-The market is efficient. Therefore, forward contract prices are unbiased estimators of future spot prices. Since forward contract prices are distributed symmetrically around the spot prices under rational expectations, we can say that the last observed spot price is a good predictor for future spot prices.

2-Spot price process behaves like a Martingale. In Martingale, expected future spot prices are equal to the last observed spot price.

I am new to these topics; therefore, I am confused about differentiating the efficient market hypothesis and martingale process. Are these explanations meaningful? Is it possible that both explanations are correct, or only one of them can be right?

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.