punatang - One month out of 12 in 2023 which is the subject of the post above. Those who have studied statistics are aware of how improbable that dataset is
But it's not improbable at all, and there are several other years with similar distributions: 2021 had 11 of 12 months as an undercount, that had to be revised upwards to the tune of ~1.9 million jobs that were missing from the month-to-month survey data (seasonally adjusted, 3rd-to-1st revision totals). 2014 also had 11 of 12 months undercount, with ~440,000 jobs not initially counted. 2010 had 11 of 12 months undercounted, 2008 had 11 of 12 months overcounted, and so forth.
In long enough datasets there are all sorts of "runs" that ultimately don't mean anything, as the underlying data is random. BLS has no control who replies in the month-to-month survey and what numbers they provide, which is why overall the BLS tends to overestimate roughly as often as they underestimate. This is similar to how when flipping a coin you getting 10 heads in a row isn't any more rare than 10 tails in a row, or perfectly alternating heads and tails, or any other combination. The human brain is prone for ascribing meaning to the patterns, but the math makes clear it's all the same statistically.
But please, punatang, do share your regression analysis to show how 2023's data is some sort of outlier in its "fakeness".
When the deeply conservative William F(ucking) Buckley Jr. National Review has to point out "No, the Biden Administration Is Not Manipulating Jobs Data" it should be clear how far out into nonsense interpretations of the world one has strayed - Dunning-Kruger effect run wild.
Umm, keep it Hawaii related, right... HPP is doomed, the data is pointing right at your property, get out while you can! Only you can save the Hawaii housing market's year-over-year sales figures! Won't someone think of the local realtors! and so forth.
But it's not improbable at all, and there are several other years with similar distributions: 2021 had 11 of 12 months as an undercount, that had to be revised upwards to the tune of ~1.9 million jobs that were missing from the month-to-month survey data (seasonally adjusted, 3rd-to-1st revision totals). 2014 also had 11 of 12 months undercount, with ~440,000 jobs not initially counted. 2010 had 11 of 12 months undercounted, 2008 had 11 of 12 months overcounted, and so forth.
In long enough datasets there are all sorts of "runs" that ultimately don't mean anything, as the underlying data is random. BLS has no control who replies in the month-to-month survey and what numbers they provide, which is why overall the BLS tends to overestimate roughly as often as they underestimate. This is similar to how when flipping a coin you getting 10 heads in a row isn't any more rare than 10 tails in a row, or perfectly alternating heads and tails, or any other combination. The human brain is prone for ascribing meaning to the patterns, but the math makes clear it's all the same statistically.
But please, punatang, do share your regression analysis to show how 2023's data is some sort of outlier in its "fakeness".
When the deeply conservative William F(ucking) Buckley Jr. National Review has to point out "No, the Biden Administration Is Not Manipulating Jobs Data" it should be clear how far out into nonsense interpretations of the world one has strayed - Dunning-Kruger effect run wild.
Umm, keep it Hawaii related, right... HPP is doomed, the data is pointing right at your property, get out while you can! Only you can save the Hawaii housing market's year-over-year sales figures! Won't someone think of the local realtors! and so forth.