By Jim Walker – Wealthy Nations (HK) –
There is not a data release that comes from China in any form that is not also accompanied by a discussion about the veracity of the statistics and what this means for any conclusions drawn from them. This is problematic on the macro level, as it means economists and analysts are stuck either using data that they feel is incorrect or trying to estimate what they believe to be a more accurate number based on other indicators.
Both of these methods have obvious drawbacks, the first as you’re admitting up front that whatever conclusions you draw are only valid if the data inputs are and the working assumption is that they aren’t. And the second because any use of indicators to guide an estimate not only introduces more uncertainty, they also assume that the data used for the indicators are for some reason free from error or manipulation. That seems like a stretch given the working assumption.
However, once we ‘clean’ the data using the blueflag Machine Learning algorithm to find examples that look like previous frauds we get a different picture.