This was discovered while the site was evaluating long term projections for US government finances. It was necessary to decide if the economy grew more slowly than expected whether government spending would slow in the worst case, or if it was reasonable to consider a scenario where it would continue to grow at the same rate. History indicates there is a good chance it may grow faster.
This result may be counter-intuitive to some people who assume the growth rates would match. The faster the private economy grows, the more money government receives in taxes to spend. The more money government gives out in welfare, the more money its recipients have to spend. It turns out that in the short term other factors are usually strong enough to overshadow those connections. For instance government spending displaces private spending. Money taken by government and used for e.g. building "a bridge to nowhere" isn't instead spent to start or grow a private company which creates more wealth and jobs. This page points out the pattern but won't attempt to isolate all the cause and effect relationships at play.
GDP (Gross Domestic Product) is a measure of the size of the economy. GDP figures include a fraction of government spending so it is appropriate to subtract that out to get a measure of the "private GDP". In federal BEA (Bureau of Economic Analysis) tables the government spending included in GDP is labeled "Government consumption and gross investment".
As the BEA points out "Total spending by government is much larger than the spending included in GDP". As this site discusses elsewhere the most complete measure of total spending available seems to be the "Total expenditures" number the BEA provides in its SNA (System of National Accounts) figures which are only available for 1970 onwards. In 2011 the BEA gives total expenditures (by all levels of government) as $6.3 trillion, while government spending included in GDP was only $3.1 trillion (both in 2011 year dollars). Note: all analysis on this page is based on inflation adjusted data (using the GDP deflator measure of inflation).
The graph below shows the % growth in real total government expenditures (federal,state and local) and the % growth in real private GDP
One way to compare sets of data to spot connections between them is to "standardize" them to highlight how they change. For example if you subtract the median value from each data set, it moves the curves down so that about half the data points for each set are above 0 and half are below (aside from median values which are at 0).
The graph qualitatively indicates that most often when government is growing faster than the median value, the private economy grows slower than its median value, and vice versa. The majority of time when growth is slowing for one (the line is headed down), it is speeding up for the other (the line is headed up). There are many factors at play which complicate the pattern so they aren't exactly mirror images. Worldwide economic factors have an influence as do government policies that impact the economy but don't change spending, advances in technology, etc.
In this case 80.49% of the time government spending and private GDP are on opposite sides of 0. The lines are headed in opposite directions 72.5% of the time. If government spending grew at the same rate as private GDP you would expect the lines would always be on the same side of 0 and headed in the same direction. If there were no connection between government spending and private GDP you would expect by random chance about half of the time the lines would be on the opposite side of 0. Since there are causal links that would lead us to expect them to be in synch (e.g. tax collection influencing spending levels), then even if the lines were on opposite sides just 50% of the time it would indicate there are major factors at play working in the opposite direction.
If you look at the change in total federal spending rather than total spending at all government levels, the results are almost the same. The same 80.49% of the time the values are on opposite sides of 0. The lines are headed in opposite directions slightly less often, 70% of the time.
Since most of the economy is private, from a purely economic standpoint it is more important to see the private economy grow fast than the government component grow quickly. From a government funding standpoint, a faster growing private economy will eventually allow government revenue to catch up and collect more in taxes than if the private economy grew slower.
pointed out those numbers don't really include all government spending, they are the closest figures to that available). To analyze world data it is useful to consider another measurement: subtract "total government expenditures" from GDP to get a different "non-government GDP" measure of the part of the economy not under government control.
GDP estimates try to deal with circular flows of funds through the economy in a way that limits "double dipping". It is their choice of method to do this which leads some government spending to not be included directly since it is picked up elsewhere in the flow (e.g. they give welfare to an individual who then buys a product which is counted in GDP).
The "non-government GDP" approach is is another way to approximately break up circular flows and looks at the fraction of GDP that hasn't passed through government control compared to the fraction that has. The approach isn't conceptually perfect since e.g. a government transfer payment in theory might have been stuffed into a mattress and never spent so it wasn't actually counted in GDP. However most people who receive government transfer payments like welfare spend them so it isn't likely a major factor (or they save it in a bank which loans it out so the money continues flowing). Unfortunately it isn't always possible to find economic data perfectly suited to the needs of a particular analysis, there are often imperfections like that. Comparing "non-government GDP" gives a useful different perspective on the fraction of the economy that isn't under control of government.
If you compare US government spending to "non-government GDP" the same way it was compared to "private GDP" above the results are similar. 82.93% of the time the values are on opposite sides of the median, and 80% of the time the lines are headed in opposite directions.
OECD provides data for several countries. Comparing "non-government GDP" to total government expenditures for each country, combining the results shows that 72% of the time government spending is on the opposite side of 0 (based on 744 data points), and 75% the lines are headed in opposite directions. The chart below shows that for every country they provide data for, at least 50% of the time government spending growth is on the opposite side of the median from non-government GDP:
The chart below adds in the % of time the lines are headed in opposite directions. Only 3 of the countries have lines headed in opposite directions less than 50% of the time:
The IMF has data for more countries (including a far higher percentage of less developed countries) but only back to 1980, and most don't go that far back. For 69.7% the lines are headed in opposite directions, and for 65.2% the values are on opposite sides of the median (based on 3095 data points).
That IMF source doesn't provide the data needed to calculate '"private GDP" (unless its categories are misnamed), only "non-government GDP". The OECD data does allow private GDP to be calculated, and 62% of the time the lines have opposite slopes, and 59% of the time the values are on opposite sides of the median.
One factor that can complicate the results and lower the numbers is if the average growth rate is increasing or decreasing noticeably. e.g. if the typical growth rate for a country (government and non-government) in 1980 were around say 1% but 10% in 2011, with a median of 5% over the whole time, then growth rates would for everything would be more likely to fall below the median early on and above the median later. That means they are on opposite sides of the median less often, which leads to lower numbers. Although the IMF and OECD data is only available for a few decades, even over that period of time some emerging countries can see their growth speed up, and more developed countries see their growth rate slow. Future work might include designing a different technique for evaluating the pattern to account for this.
As this site noted before, the rules for calculating "total expenditure" figures leave off some government spending. The issue may be more pronounced for some countries than the US which may impact the results.
Although "total expenditure" data isn't available for earlier years for the United States, "current expenditure" data is available from 1929 onwards. That includes a large fraction of government spending, even if not all of it. The growth rate for US GDP has slowed over that period of time. Despite those issues, the pattern still shows up if you compare current expenditures to private GDP for that period. 69.51% of the time the line is headed in opposite directions, and 62.65% of the time the growth rate is on opposite sides of the median.
Occam's razor would suggest that at a minimum unless/until there is stronger evidence to the contrary than exists today, the simple possibility that faster government spending growth might have a good chance of slowing the private economy shouldn't be ruled out.
Anyone considering the potential causes for these results should note the pattern is less apparent when using non-inflation adjusted figures. Although people estimate inflation beforehand, it often differs from inflation measured after the fact. That may indicate less of a potential role for intentional planning to be part of the explanation. Projections of near term future GDP growth used for planning budgets are often inaccurate (even calculations of GDP after a quarter ends are often approximations that are revised later), so any examination of the role of planned spending would need to take into account the data they planned with and not just the measured data after the fact.
Despite those caveats, planned responses to growth rates might play some role in the explanation. When an economy is expected to grow faster than usual, a country may be more likely to yield to pressure for tax cuts which could then keep spending from rising as fast as the economy is growing. Governments often use "baseline" budgeting where they project spending to keep increasing in the future. They pretend any reduction in their plan is a "cut", even if spending will still rise. If the economy seems likely to grow more slowly than usual, they may fear they won't have enough revenue to cover planned spending growth. They may push for tax increases to cover their planned spending and will likely try to raise taxes more than required "just in case" to make sure it covers their future budgets. They are likely to spend all the increased revenue that comes in, so spending may grow faster than usual, while the economy grows even slower.
It is possible the largest impact of tax increases may be when they initially go into effect. Businesses and individual investors will change strategies to try to minimize the impact of taxation.The tax may lead to a revenue spike initially while that is in the works, while there may be "friction" and waste incurred from those changes which slows private GDP growth more noticeably at first.
theory doesn't seem plausible to begin with and evidence is lacking or questionable). For many years economists were skeptical of the approach even if politicians like it, so it is also not clear how often it has been intentionally done. Wishful thinking seems to have made the idea more popular lately.
It is useful to note for the moment that aside from other possible flaws, studies claiming additional government spending during a recession will increase economic growth may not always do a good job of considering alternative explanations which might complicate claimed causes and effects. Specific studies won't be cited at the moment, merely examples of the types of concerns that need to be addressed in any such study. These illustrate the difficulty of isolating cause and effect relations between government spending and the economy.
Some claim they find signs of a delay between a spike in government spending during a recession and the positive impact on the economy. If a spike in government spending occurs during a recession, and then after it stops the economy grows, obviously it may be instead the economy is growing faster because the spike was over with rather than that there was a delayed benefit. A spike in government spending that overlaps a recession may not have intentionally been planned to counter the recession. It may have occurred because they didn't anticipate a recession and thought the economy was strong and would be growing faster. The spike may have helped contribute to the recession and when spending slowed, the underlying strength led the economy to recover and grow quickly.
There is a cynical explanation for any data that superficially might seem to show "stimulus" planned to counter a recession is beneficial. Politicians are rationally self interested so this possibility can't easily be ruled out without evidence that may be hard to find. If politicians think data shows their economy will eventually recover on its own, the faster the recovery they expect the more money they may risk spending on pork mislabeled "stimulus". Rather than faster growth being caused by more spending, more spending may be caused by expected later faster growth. Of course in some cases their expectations turn out to be wrong and things get worse (possibly due to the "stimulus").
Some economists suggest there can sometimes be a rebound effect after recessions where the economy initially grows faster after a bad recession ends. For example people spend money they held off spending while the economy was slow and start businesses based on a backlog of ideas that accumulated during the downturn. If government policies in reality kept an economy in recession longer then any rebound effect may be larger afterwards. A larger rebound after intentional "stimulus" could therefore be due to a negative impact of government policies creating a longer recession rather than "stimulus" spending having been a positive factor in an eventual larger boom.
Some studies on counter-cyclical spending only use the fraction of government spending included in GDP rather than total government expenditure figures even if they might be more appropriate to their approach. ( in the case of the US noted above the government share of GDP was about half as much as total spending).
The values for the % that are on opposite sides of 0 are a tiny bit lower than they might be in the long run due to an approximation for convenience: If the growth value for a year exactly matches the median it winds up at 0. This always happens if there are an odd number of years in the data, and can occur coincidentally regardless if more than on year has the same growth rate. For convenience the median value is treated as if it were below 0. Since that means there are more values below zero than above, that means there is slightly less chance values will be on opposite sides. This makes only a small difference if say 1 value out of 41 years is exactly at the median. However if there are only e.g. 5 years of data and at least 1 of those is at the median, there are 3 (or more) below 0 and 2 (or less) above. This may make a difference in the numbers for some individual countries in the international data sets since some countries only have a few years worth of data. It shouldn't make much difference for the data set as a whole. In the OECD data set 51% of the values are below the media, and in IMF 52% are. In both cases the odds by chance that values will be on opposite sides of the median aren't quite 50% but are still over 49.9%.
These results should be straightforward for those who are comfortable with spreadsheets and economic data to replicate. Links to data are provided within the main page above. (note: the GDP and government share of GDP data is from the BEA's standard NIPA tables which is linked to first, it was only the "total expenditures" figures that were taken from their SNA tables as more complete figures for government spending). The initial plan was to upload the spreadsheets from day 1, but they won't provide any benefit to the casual reader beyond what was described above. The spreadsheets won't be provided at first to encourage anyone curious to confirm the results independently. The method is simple enough it is likely faster for someone to cut and paste formulas in themselves to many cells to recreate the results than to check each cell in an existing spreadsheet they didn't write to confirm the formula make sense. It would be useful if any academics or policy analysts choose to do so. The spreadsheets used by this site might be added in the future for those that don't want to bother recreating them but are still curious to see them for some reason.
Technique: Simply grab the real GDP and government spending data in spreadsheets provided by the sources (or modify nominal data using deflators, or by using data given as a % of GDP and then use real GDP values). All data used each country's default currency and not US dollar figures to avoid any distortion from conversions into US dollars. Create rows for the yearly percentage change for GDP and non-government GDP and (where available) private-GDP values. Calculate the median of each. Create rows to indicate if each value is above or below the median. e.g. if a growth rates is above the median set the corresponding result cell to 1, otherwise it set it to -1, or a "" empty string cell if there is no data for that year (many countries in the international data sets don't have data for the full time period). Create a summary where the median-comparisons for non-government and government are multiplied. That results in a 1 if they are on the same side of the median, or -1 if they are on opposite sides. The results are then counted to determine the percentage of times they are above or below the median using the functions "countif" to count how many cells contain "-1" and "count" to see how many cells contain data (rather than being blank). Use a similar approach to compare the growth rate for each year to the year prior to see if growth is speeding up (placing a 1 in the result cell) otherwise a -1 result, or an "" empty string if there isn't data.
OpenOffice is used by default for spreadsheets on this site. Those that use spreadsheets to analyze data should be aware that Excel is fine for simple calculations like those on this page. However it it has a history of numerical bugs in things like its statistics routines and trend analysis routines (e.g. see here, here, here and here just for a start).