Here is a brief excerpt from one of the article:
"Of the three different approaches or methods that use the results from hedonic regression models to quality adjust price indexes, BLS employs the 'matched model' method in its official indexes.2 This method controls for quality changes based on the difference in product specifications or characteristics between two items when a substitute observation, or quote, occurs in the price index sample. It is important to note that under the 'matched model' approach only substitution price changes, or quotes, are eligible for hedonic quality adjustments."
"As previously announced, the Bureau of Labor Statistics (BLS) is extending the use of quality adjustments derived from hedonic models in the CPI. A hedonic model decomposes the price of a consumer product into implicit prices for each of its important features and components, thereby providing an estimate of the value for each price influencing feature and component."
Regarding the Federal Reserve's use of quality to adjust capacity, I was unable to locate anything quickly, but I was doing some analysis on recent changes in capacity and the numbers just did not make sense based on my observations of recent trends in capital equipment. I contacted the analyst that assembles the industry indices and he informed me that most of the growth in capacity had occurred in high tech and most of that growth was based on quality improvements of the output rather than investment in physical capital.
Since capacity utilization is simply the IP index divided by the capacity index, the utilization rate gets impacted indirectly. Also, the capacity index is derived by different methods depending on the industry and its data availability. Their benchmark is the Survey of Plant Capacity, for which data has a two year lag. Then they attempt to make adjustments to that data to make it current. So you are really using a denominator based on today's output over a numerator that is based on capacity that existed two years ago.
I work with all types of government and private data every day and when I start looking into how the data is constructed, it can be a little unsettling. It is useful for analyzing trends if one asumes that in the long run and in the aggregate the errors cancel each other out. But to draw specific conclusions based on a snapshot can cause problems. Some data is better than others. The trade data is horrible. I have had more than one official tell me that anyone that uses it to obtain any detailed information is a fool. Pretty encouraging, huh? NABE (National Associaton of Business Economics) has been lobbying extensively to make the data more accurate but it is not easy. The People who put the data together are all very capable and serious about their work but the data they put out can only be as good as what goes in. I guess the main problem is a lack of funding.