reader rsm In response to my quotation of GDP and Human Development Index data:
Why not include the standard error?
Again, because they are so extensive, can you tell any story and use data to support it?
In addition to being an incredibly nihilistic statement, it is also a statement of universal ignorance.
The simple answer is that there is no such thing as standard errors in GDP estimates, at least not in the classical statistical sense. On the other hand, this does not mean that people should not try to convey the inaccuracy of GDP. In fact, every BEA report in GDP publishes the following table:
source: Bank of East Asia.
Therefore, BEA reported the inaccuracy of subsequent revisions. This is not standard error-I don’t even know what it means that standard error represents the standard deviation of the GDP sample population. On the contrary, the table conveys the degree of imprecision of subsequent revisions, and some people believe that subsequent years will more closely represent the “real” GDP.
There is a more extensive table in this document (“Comparison of GDP Revisions”, September 2021) Is specifically used to list the magnitude of the revision.
source: Bank of East Asia (2021).
in a Current business investigation article(Dennis J. Fixler, Eva de Francisco and Danit Kanal, “Revision of Gross Domestic Product, Gross Domestic Income and Its Main Components”, January 2021), provides comparable statistics on GDP, GDP components, GDI and GDI components.
But it is important to remember that these communication measures are not “standard errors” in the traditional sense (at least not to me!).
In his extensive discussion on how to report government statistics, he more strongly conveyed the uncertainty surrounding estimates, Mansky (Jere, 2015) Reference to the Bank of England’s uncertainty measures on UK GDP (“Pan Chart”).
Galvo and Mitchell (2019) Provide this example:
source: Galvo and Mitchell (2019).
Suppose someone really wants to solve the problem of inaccuracy comprehensively. what do you need? Mansky (Jere, 2015) Write
From the perspective of economic statistics users rather than from the perspective of statisticians, I think it is necessary to distinguish the measurement error of a well-defined concept from the uncertainty of the concept that should be measured. I also think it’s useful to distinguish between errors that decrease over time and errors that persist. To highlight these differences, I will discuss instantaneous statistical uncertainty, permanent statistical uncertainty, and conceptual uncertainty, respectively.
Data revisions belong to the first category. This leaves many other problems unresolved. Regarding permanent statistical uncertainty, Manski emphasized the non-response of the survey and imputation. For conceptual uncertainty, he studied the impact of seasonal adjustments.The latter is a thorny issue and provides (as a reader rsm Wanting) all-not just temporary-statistical uncertainty will be very difficult.
Going back to the temporary uncertainty (especially the revised version), I redrawn these charts (from this postal):
figure 1: The actual GDP as of January 31, 2001 (blue), as of April 27, 2001 (red), as of July 27, 2001 (green), and as of June 28, 2018 (black) are standardized as 1999 The first quarter of the year. The decline date defined by NBER is shaded in gray. Source: ALFRED, BEA, NBER and the author’s calculations.
figure 2: As of January 31, 2001 (blue), as of April 27, 2001 (red), as of July 27, 2001 (green), and as of June 28/2018 (black). The decline date defined by NBER is shaded in gray. Source: ALFRED, BEA, NBER and the author’s calculations.
Revision discussion on Econbrowser (short form): [1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12].
