You. I like you. Good questions.
do we know with any degree of certainty that NPD is a representative sample of global (or even just national) sales?
NPD forecasts the US video game Packaged goods market on an over 90% sample. The extrapolation rates for the remaining percentage are quite solid. Therefore, yes, there is a high degree of reliability for the market.
For instance, I've heard Gamestop isn't counted. If that is correct, then it might matter whether buying habits of Gamestop customers differ from other retailers that are counted in regards to whether the sample is representative.
You heard incorrectly. NPD has arrangements with all major US retailers including Walmart, GameStop, Best Buy, Amazon, and many, many others.
I guess what I am asking is how accurate these numbers are at predicting whatever the "real" sales data is globally or nationally, physical and digital. Assuming we ever even get to know that real data.
The figures NPD publishes are quite good for estimating the size of the packaged goods, physical, retail, video game market. However, the figures do not include sales of digital software on PSN or XBL, for example. Therefore, the slice NPD represents is a quite poor representation of the complete market for video game spending.
The packaged goods market has been in decline for some time. However, these declines are more than made up for by growth in Digital, Mobile, Tablet, PC, etc. The overall gaming pie is growing at a fantastic rate. However, the portion of the market NPD covers is in decline.
NPD has cracked down on the type of data it releases publicly because, as this slice is the only data available, it is quite misleading for the health of the industry as a whole.