Why did the Federal Reserve change the data about USA GDP?

Question

This is the data published on 20 March 2013 FOMC meeting: (http://www.federalreserve.gov/monetarypolicy/fomcprojtabl20130320.htm)

20 March 2013,

Change in real GDP (actual):

• year 2008: -3.3%
• year 2009: -0.1%
• year 2010: +2.4%
• year 2011: +2.0%
• year 2012: +1.6%

This is the data published on 18 September 2013 FOMC meeting: (http://www.federalreserve.gov/monetarypolicy/fomcprojtabl20130918.htm)

18 September 2013,

Change in real GDP (actual):

• year 2008: -2.8%
• year 2009: -0.2%
• year 2010: +2.8%
• year 2011: +2.0%
• year 2012: +2.0%

As you can see, the GDP data for year 2008, 2009, 2010 and 2012 that published on 20 March 2013, and the data that published on 18 September 2013 is different.

Why did the Fed change the data?

Answer 1

@BB King's answer is good, but I want to make one point very clear:

The Fed did not change the data.

The Bureau of Economic Analysis revised the data as a part of its 2013 comprehensive revision process, which it engages in every five years using data from the most recent Economic Census.

It may be worth noting that the BEA, which is the agency that is responsible for estimating US GDP, is housed in the US Department of Commerce, and as such is wholly independent of the Federal Reserve.

The changes made in the 2013 comprehensive revision are discussed here, with a table showing annual changes on page 16. Note that the annual numbers shown in the article do not align with the numbers shown in the Fed's table because the BEA publishes percent change in annual real output, while the numbers used by the Fed in your links are showing the percent change in 4th-quarter real output, which can affect how the timing of changes appears. As a result, the major slowdown that occurred in 2008 Q4, for example, shows as a large drop in 2008 4th-quarter output and a small change in 2009 4th-quarter output using the quarterly comparison shown by the Fed, while it shows up as a small decrease in overall 2008 output and a large decrease in overall 2009 output using the annual numbers shown in the BEA's table.

Answer 2

This happens all the time and is normal. It's due to measurement error. It is hard to measure these things in real time and as time goes by more data comes in, which allows measurement of GDP and other macroeconomic variables to be better so they are revised. Sometimes the revisions can be quite large. Typically the latest data is also the least reliable.

For some advanced information on how such measurements are made, the issue of measurement error and a discussion of how to improve estimates see this paper

For a discussion on the issues these revisions of GDP data and measurement error pose for monetary policy, see the work of Orphanides, for example here and here. For his discussion on the reliability of GDP estimates in real time see this paper.