A consensus price forecast strives to assess the wisdom OF the crowd. However, LBMA's Annual Precious Metals Forecast Survey seeks the individual analyst who makes the most accurate market price forecast for that given year — identifying the wisdom IN the crowd.
How effectively do the surveys serve as consensus metal price forecasts?
Consensus Forecasting Accuracy
Examining forecast survey data from 1997 to 2024, the aggregated consensus estimates of analysts' market price forecasts align favourably with the average LBMA market prices for the respective metals.
The difference between the consensus median for all years of the LBMA Forecast Survey and the comparable average market prices for all metals was approximately 1%, an incredibly close match, with similar results for the consensus mean. The respective average market price and the consensus median for all the years were:
The aggregated comparison would, notionally, support consensus estimates as reliable forecasts of future LBMA market prices.
However, when considering the distribution of forecasting errors over the individual years, the results are less compelling. The following graphs show the Absolute Percentage Error (APE) of the consensus median and the year-on-year market price changes, with the APEs averaging around 9%, emphasising the difficulty in predicting future market prices:
Reliability of Consensus Estimates
When we recommend using a central estimate, such as the median or mean, we assume that forecasting errors will be roughly balanced — some predictions will be too high, while others will be too low, but they'll mostly cancel each other out. However, this balanced pattern didn't happen every year. In many cases, the errors were skewed in one direction, either predominantly above or predominantly below the actual market price.
Market prices typically fluctuate, although they occasionally remain relatively stable. To help classify these changes, a simple 3% threshold is used:
The skewed distribution of the analysts' forecasts relative to the market price changes is tabulated below:
The clustering around the market average only accounted for ~8% of the years for which analysts submitted forecasts for LBMA's survey.
Most forecasts were either mostly above or mostly below the actual market average. Generally, when the market price declined during the year, analysts tended to overestimate the change, forecasting prices that were too high. On the other hand, when the market price rose significantly, they tended to underestimate the change, forecasting prices that were too low.
The way analysts' past forecasts have been skewed compared to actual market changes highlights the importance of verifying which type of consensus estimate is most accurate. The table below shows which consensus estimate has been the most reliable during Bearish/Neutral, and Bullish market conditions:
In years when the market was Bearish or Neutral, the lower quartile provided the most accurate forecast about 71% of the time. In Bullish years, the upper quartile was the most accurate in about 66% of cases. In comparison, the median (approximately 6%) and the average (approximately 9%) were the most accurate, significantly less frequently.
Based on the observed skewness of the analysts' forecasts relative to the subsequent market price changes, a generic rule of thumb for choosing the most reliable consensus estimate is proposed as follows:
- For expected market price increases of less than 3% (Bearish/Neutral), the lower quartile is likely to be the most reliable consensus estimate.
- For expected market price increases above 3% (Bullish): the upper quartile will likely be the most reliable consensus estimate.
The higher accuracy of using the lower or upper quartiles depends on correctly guessing which way the market will move. When most analysts expected a Bullish year, they were right about 89% of the time. For Bearish or Neutral years, their accuracy dropped to about 57%. Overall, about 81% of analysts correctly predicted rising prices and about 69% correctly predicted falling prices.
Conclusion
Overall, about 34% of survey analysts were more accurate than the median forecast. A smaller group, around 18%, were more accurate than the lower or upper quartile forecasts, based on actual market price changes. Participating in more surveys opens up the possibility of outperforming consensus measures. However, for those analysts who participated for more than ten years, ~17% were, on average, more accurate than the median, and a lesser 11% were more accurate than the lower or upper quartile forecasts, based on actual market price changes. The research showed that while individual analysts can be accurate, the "wisdom of the crowd" generally provides a more reliable estimate. Arguably, the best consensus isn't always the median or mean — it can be the lower or upper quartile, depending on whether market prices are expected to rise or fall.
Postscript
Based on the 2025 year-to-date (June 2025) average metal prices, the proposed rule of thumb — 2025 LBMA Annual Precious Metals Forecast Survey, most reliable consensus estimate:
