People do not accept the fiscal theory for an obvious reason: nobody can believe that a substantial fraction of the investors are trying to figure out present value of future deficits. No matter how good is your narrative or data fit. The mechanism is not implausible but ridiculous. If you are right (and I think you are probably right), it is for the wrong reasons.
Now, the big question is, how short term efficient (for short horizons, prices are martingales) but long term indifferent (see Euro countries sovereign spreads in 2007) capital markets are affected by long term fiscal variables?
Until you have that mechanism, Fiscal Theory of Price Level is absolutely implausible, no matter how attractive it is (and it is extremely attractive).
To reiterate a frequently given answer... it's the same way the average person figures out the present value of future dividends, and decides to buy stocks or bonds. Basically, people do or don't have a faith that the government will repay its debts, motivated mostly by a general feeling of confidence in the institutions of repayment. Also, it's not each person, but the market consensus, weighted by money. If you accept price is driven by present value of dividends, it's the same. If you reject that too, well, good luck.
First of all thank you for your nice answer to my slightly grumpy question.
Regarding equity we know that a substantial percentage of equity investors are trying to predict the free cash flow generating capacity of firms (I am more agnostic on dividends).
How many fixed income investors are forecasting primary fiscal surpluses? Are the banks that hold the excess reserves that fund the euro area quantitative easing making optimal mean variance portfolios? I have some degree of contact with fixed income managers, and they follow unemployment levels, confidence indices, gdp data, etc. Debt sustainability analysis? No one really understands or follow it.
Of course, reality correct expectations incompatible with sustainable fiscal trajectories. As a consequence of this ex-post correction the fiscal theory is probably to some extent predictive. But when markets pay attention to fiscal data it is because the situation is already desperate.
During WWII, at year end of 1942, Federal Debt was $72 billion. In 1943 the budget deficit was $55 billion, or 80% of national debt and around 25% of GDP. While serious post war inflation emerged, it was no worse than in the '70s and was quickly damped down. Yet in the '70's, the deficit rose above 10% of debt in only one year. How do you account for this? It's all about money. Please see https://charles72f.substack.com/p/aint-nothin-but-a-party
That's the point. They didn't have to include money. The split of debt between actual debt and money didn't seem to matter. The contrast of QE -- same monetization, no inflation -- with covid proves the point.
Any thoughts as to how this process might work in practice? If deficit spending financed through actual debt issuance causes inflation, it seems to imply that (in aggregate) the money that was used to buy that debt wasn't actually circulating in the economy - there is no crowding out effect. Or maybe that the debt was purchased using newly-created bank credit or some form of indirect money creation, or the debt securities issued were rehypotheticated to create more credit? Some kind of multiplier effect on new government debt issuance.
Robert J. Barro's and Francesco Bianchi's Figure 1, "Change in Headline CPI Inflation Rate versus Composite Government-Spending Variable", shows Norway's ("NOR") data plotted (incorrectly) at (-0.053, -0.014). The correct data point for NOR is (-0.053, +0.014).
The plotting error skews the linear OLS regression line slope and intercept giving the impression of greater statistical agreement than the correctly plotted linear OLS regression line is capable of supporting.
The linear OLS regression model with NOR incorrectly plotted at (-0.056, -0.014) gives the following regression parameter values (m, b), and statistics (r^2, F-stat) for the 21 independent data points (dof = 19) in Table 1:
slope, m = 0.4999
intercept b = 0.00868
r^2 = 0.635
F-statistic = 33.1
The linear OLS regression model with NOR correctly plotted at (-0.056, +0.014) gives the following regression parameter values (m, b), and statistics (r^2, F-stat) for the 21 independent data points (dof = 19) in Table 1:
slope, m = 0.3662
intercept b =0.01344
r^2 = 0.4212
F-statistic = 13.8
When the data is correctly plotted, Barrio and Bianchi's regression model's explanatory power is lower by a significant amount. The correlation coefficient ( r ) is 0.649 for the correctly plotted data, versus 0.797 for the incorrectly plotted data. The difference in the F-statistic for the correctly plotted data and the incorrectly plotted data is notable, 13.8 for the correctly plotted data vs. 33.1 for the incorrectly plotted data.
Would this correction necessarily alter Barro's and Bianchi's conclusions? Perhaps.
One further point. Barro and Bianchi use reduced statistics for the Euro-zone data. This single data point replaces the seventeen Euro-zone countries that would otherwise have been plotted individually. This short-cut approach can only be justified if the Euro-zone country data is identical for each of the seventeen Euro-zone member countries. A quick scan of Table 1, Table 2, and Table 3, data suggests that the short-cut is unwarranted, a. s.
They plotted Europe as one point (which I suggested, in response to an earlier graph that plotted them separately) because Europe has one currency. Whatever causes different measured inflation across euro countries is not monetary or fiscal policy. as we do not use US states separately.
Plotting the Euro-zone data separately, rather than as a "weighted average" (weighting formula not disclossed by Barro and Bianchi) makes no statistically significant difference in the corrected linear OLS regression coefficients (i.e., slope and intercept, r^2; the F-statistic is improved by virtue of increase in data points). Note that the data point for Norway ("NOR") is incorrectly plotted in Barro and Bianchi's figure 1 which leads to a steeper regression line (higher slope, lower intercept values).
Francesco (Bianchi) here. Thanks for the interest in our work! The data point for Norway is corrected with the border dummy because Norway borders with Russia. In other words, the effect of the border dummy is removed from the change in inflation. We do the same for all countries that border with Russia. I hope that this helps.
Thank you for your comment. The ordinate in Figure 1 of your paper plots the change in the rate of headline inflation 2020-2022 versus 2010-2019. The ordinate value for the data point NOR lies between minus 0.010 and minus 0.020. NOR = Norway. Table 1, Part I, Column (1) at "Norway" reads plus 0.0140. From this observation, I deduced that the data point for NOR in Figure 1 is incorrectly plotted. Now, there may be an adjustment which you and your co-author have made which yields an ordinate value of minus 0.0140 in Figure 1, in which case the data point for NOR in Table 1, Part I, Col. (1) is erroneous (i.e., Norway experienced disinflation between 2010-2019 and 2020-2022). However, when I look at the regression coefficient values and standard errors of the coefficient values appearing in Table 5, Column "Headline CPI inflation rate" labelled "(1)", the constant = 0.0134 (0.0037) and the slope = 0.369 (0.099) I find that your coefficient values and the coefficient values that I obtain in a linear OLS regression coefficients with NOR = ( - 0.053, + 0.014 ), i.e., constant = 0.0134 (0.00374) and slope = 0.3661 (0.0985). When I set NOR = ( - 0.053, - 0.014), the regression coefficients change to constant = 0.00868 (0.0033) and slope = 0.4999 (0.087).
Based on the agreement between the regression model in Table 5 (column "(1)") and the Excel worksheet regression model constructing using the data in Table 1, Part I, Col. "(1)" for the ordinate, and the data in Table 2 Col. "(1)" and Col. "(3)", and the data in Table 3 Col. "(2)", the only conclusion that I can fairly draw is that the data point labelled "NOR" is incorrectly plotted at ( - 0.053, - 0.014). It likely results from an inadvertent typographical error in the worksheet from which Figure 1, appearing on page 44 of your paper, was generated.
Given that the regression model coefficient values set out in Table 5 Col "(1)" agree with the Table 1 Part I, Table 2, and Table 3 data, the specific observations and conclusions drawn in your paper are unaltered. Only Fig. 1 would need to amended to bring the plot into agreement with the data and the corresponding regression model statistics.
Hi, we have a regression with two regressors, and we want to isolate the effect of the rescaled change in spending. So, we need to control for the effect of the border dummy. We are going to adjust the caption to make this clearer, but we already say what we do: See note 30, under the Caption of Figure 1 to see how we compute the variable on the vertical axis.
Thanks again for your explanation. I ought to have read the notes to the tables and footnote #30 more closely. Hungary, Norway, and Poland border either UKR or RUS or both, as you note. Because "bordering UKR or RUS" was only significant from February 2022, the more appropriate censuring measure should have been to drop HUN, NOR and POL from the regression model. Eliminating HUN, NOR and POL from the model does not alter the regression coefficients -- the linear OLS regression model coefficients become slope = 0.4264 (0.0876) and intercept = 0.00792 (0.0031), r^2 = 0.597 and the F-statistic = 23.7.
Excluding the EUR mean and adding only those Euro-zone countries (unweighted) that do not border either UKR or RUS yields a linear OLS regression model with slope = 0.364 (0.076), intercept = 0.0108 (0.00256), r^2 = 0.457 and F-statistic = 22.74. Excluding Euro-zone countries that border UKR or RUS in place of the EUR mean does not alter the positive relation between the change in inflation rate between 2010-19 and 2020-22 and government expenditure scaled by the product of government debt and the duration of government debt.
I withdraw my earlier comments posted in this thread. Thanks again for your comments and patience.
The first equation appearing in the blog post above is derived from the accounting identity of the Department of The Treasury's general account with the Federal Reserve Bank.
The first equation is given by Bₜ₋₁ / Pₜ = 𝔼ₜ{∑ₙ₌₀ βⁿ∙sₜ₊ₙ} where I have replaced the index letter j appearing in the infinite series with the letter n for convenience. Using the lag operator, L , we can represent the first equation by L∙Bₜ / Pₜ = 𝔼ₜ{sₜ}/[1 – β∙L⁻¹] (see, J. H. Cochrane, 2021, "Fiscal Theory of the Price Level" for examples of the use of the lag-operator in deriving infinite series representations of state transition equations in discrete-time economic models).
Now, multiply both sides of the lag operator model equation by the denominator on the RHS of the equals sign, i.e., multiply L∙Bₜ / Pₜ = 𝔼ₜ{sₜ}/[1 – β∙L⁻¹] through by [1 – β∙L⁻¹] to obtain [L – β]∙Bₜ / Pₜ = 𝔼ₜ{sₜ}.
Expand this equation and multiply both sides by Pₜ to find the accounting identity of the Department of The Treasury's general account with the Federal Reserve, i.e., Bₜ₋₁ – β∙Bₜ = Pₜ∙𝔼ₜ{sₜ}. But, note that 𝔼ₜ{sₜ} = sₜ and that Pₜ∙sₜ is the nominal dollar primary surplus, Sₜ . Ergo, by 'reverse-engineering' of the fundamental equation of the fiscal theory of the price level, we have found the Department of The Treasury's cash accounting identity, i.e., Bₜ₋₁ – β∙Bₜ – Sₜ = 0. The discount factor β is shown as being time-invariant in the fundamental equation of the fiscal theory of the price level; but, as is well-known, the discount factor varies with calendar time and primarily in accordance with temporal monetary policy (Fedfunds rate, etc.)
Plotting the data series suggested in the blog post against calendar time is instructive for those whose interests extend to ascertaining patterns in economic time series in order to explain monetary policy variations. I offer the following FRED chart as a starting point: https://fred.stlouisfed.org/graph/?g=1nTvM . The data series plotted are:
Blue trace:
- (Net lending or net borrowing(-), NIPAS: Government: Federal+Federal government current expenditures: Interest payments)/(Federal Debt Held by the Public/1000)*100 (units: percent).
Red trace:
- Unemployment Level/Employed, Usually Work Full Time*100 (units: percent).
Green trace:
- Consumer Price Index for All Urban Consumers: All Items in U.S. City Average (units: percent-change, year-over-year).
Orange trace:
- Federal Funds Effective Rate (units: percent per year).
The inflation rate (green trace) appears to follow a 'reversion-to-the mean' process, as does the unemployment rate (red trace). The primary surplus to debt ratio, up until Q2-2013, followed a recovery process after each recession, but after Q2-2013 that process no longer holds (for whatever reason) and the primary surplus to debt ratio has remained negative since then. This suggests that there has been a fundamental change in monetary and fiscal policy in the U.S., starting from the end of Q2-2013. If so, then it is worth further investigation inasmuch as it may have implications for U.S. price stability going forward, according to the fiscal theory of the price level according to the fundamental equation of the FTPL.
Not sure if I'm missing something, but I think your "primary surplus/gdp" graph is wrong. You say "Surplus is net lending or borrowing, AD02RC1Q027SBEA, primary surplus subtracts interest payments A091RC1Q027SBEA" But if I take the first series minus the second, I don't get that. I get this: https://fred.stlouisfed.org/graph/?g=1o5sF It is never positive.
Finally, I ADDED the two series and divided by real GDP (GDPC1) although most measures or real or potential real look the same... and I get a graph that looks like yours. Here's the link: https://fred.stlouisfed.org/series/AD02RC1Q027SBEA#0 .
If your graph has them added, then the comments/stories around it are off too.
I suspect my graph with them subtracted isn't right because we did indeed run primary surpluses in the late 1990s... but what data to look at?
Any thoughts?
I'm very interested in finding good data to play with and think about FTPL. So, I'd be interested in any suggestions on good time series for real primary surpluses or appropriate measures of debt, etc. I plan to look at these globally over the summer, but wanted to start with the best data I know of which is USA FRED data so I feel confident that I'm looking at what I want to look at then I can worry about non-US measures.
People do not accept the fiscal theory for an obvious reason: nobody can believe that a substantial fraction of the investors are trying to figure out present value of future deficits. No matter how good is your narrative or data fit. The mechanism is not implausible but ridiculous. If you are right (and I think you are probably right), it is for the wrong reasons.
Now, the big question is, how short term efficient (for short horizons, prices are martingales) but long term indifferent (see Euro countries sovereign spreads in 2007) capital markets are affected by long term fiscal variables?
Until you have that mechanism, Fiscal Theory of Price Level is absolutely implausible, no matter how attractive it is (and it is extremely attractive).
To reiterate a frequently given answer... it's the same way the average person figures out the present value of future dividends, and decides to buy stocks or bonds. Basically, people do or don't have a faith that the government will repay its debts, motivated mostly by a general feeling of confidence in the institutions of repayment. Also, it's not each person, but the market consensus, weighted by money. If you accept price is driven by present value of dividends, it's the same. If you reject that too, well, good luck.
First of all thank you for your nice answer to my slightly grumpy question.
Regarding equity we know that a substantial percentage of equity investors are trying to predict the free cash flow generating capacity of firms (I am more agnostic on dividends).
How many fixed income investors are forecasting primary fiscal surpluses? Are the banks that hold the excess reserves that fund the euro area quantitative easing making optimal mean variance portfolios? I have some degree of contact with fixed income managers, and they follow unemployment levels, confidence indices, gdp data, etc. Debt sustainability analysis? No one really understands or follow it.
Of course, reality correct expectations incompatible with sustainable fiscal trajectories. As a consequence of this ex-post correction the fiscal theory is probably to some extent predictive. But when markets pay attention to fiscal data it is because the situation is already desperate.
Awesome answer.
During WWII, at year end of 1942, Federal Debt was $72 billion. In 1943 the budget deficit was $55 billion, or 80% of national debt and around 25% of GDP. While serious post war inflation emerged, it was no worse than in the '70s and was quickly damped down. Yet in the '70's, the deficit rose above 10% of debt in only one year. How do you account for this? It's all about money. Please see https://charles72f.substack.com/p/aint-nothin-but-a-party
Barro's graph is quite impressive. However, he does not include monetary phenomenon.
Most if not all of those countries had both expansive fiscal and monetary policies.
Shouldn't both be included?
That's the point. They didn't have to include money. The split of debt between actual debt and money didn't seem to matter. The contrast of QE -- same monetization, no inflation -- with covid proves the point.
Any thoughts as to how this process might work in practice? If deficit spending financed through actual debt issuance causes inflation, it seems to imply that (in aggregate) the money that was used to buy that debt wasn't actually circulating in the economy - there is no crowding out effect. Or maybe that the debt was purchased using newly-created bank credit or some form of indirect money creation, or the debt securities issued were rehypotheticated to create more credit? Some kind of multiplier effect on new government debt issuance.
Robert J. Barro's and Francesco Bianchi's Figure 1, "Change in Headline CPI Inflation Rate versus Composite Government-Spending Variable", shows Norway's ("NOR") data plotted (incorrectly) at (-0.053, -0.014). The correct data point for NOR is (-0.053, +0.014).
The plotting error skews the linear OLS regression line slope and intercept giving the impression of greater statistical agreement than the correctly plotted linear OLS regression line is capable of supporting.
The linear OLS regression model with NOR incorrectly plotted at (-0.056, -0.014) gives the following regression parameter values (m, b), and statistics (r^2, F-stat) for the 21 independent data points (dof = 19) in Table 1:
slope, m = 0.4999
intercept b = 0.00868
r^2 = 0.635
F-statistic = 33.1
The linear OLS regression model with NOR correctly plotted at (-0.056, +0.014) gives the following regression parameter values (m, b), and statistics (r^2, F-stat) for the 21 independent data points (dof = 19) in Table 1:
slope, m = 0.3662
intercept b =0.01344
r^2 = 0.4212
F-statistic = 13.8
When the data is correctly plotted, Barrio and Bianchi's regression model's explanatory power is lower by a significant amount. The correlation coefficient ( r ) is 0.649 for the correctly plotted data, versus 0.797 for the incorrectly plotted data. The difference in the F-statistic for the correctly plotted data and the incorrectly plotted data is notable, 13.8 for the correctly plotted data vs. 33.1 for the incorrectly plotted data.
Would this correction necessarily alter Barro's and Bianchi's conclusions? Perhaps.
One further point. Barro and Bianchi use reduced statistics for the Euro-zone data. This single data point replaces the seventeen Euro-zone countries that would otherwise have been plotted individually. This short-cut approach can only be justified if the Euro-zone country data is identical for each of the seventeen Euro-zone member countries. A quick scan of Table 1, Table 2, and Table 3, data suggests that the short-cut is unwarranted, a. s.
They plotted Europe as one point (which I suggested, in response to an earlier graph that plotted them separately) because Europe has one currency. Whatever causes different measured inflation across euro countries is not monetary or fiscal policy. as we do not use US states separately.
Plotting the Euro-zone data separately, rather than as a "weighted average" (weighting formula not disclossed by Barro and Bianchi) makes no statistically significant difference in the corrected linear OLS regression coefficients (i.e., slope and intercept, r^2; the F-statistic is improved by virtue of increase in data points). Note that the data point for Norway ("NOR") is incorrectly plotted in Barro and Bianchi's figure 1 which leads to a steeper regression line (higher slope, lower intercept values).
Francesco (Bianchi) here. Thanks for the interest in our work! The data point for Norway is corrected with the border dummy because Norway borders with Russia. In other words, the effect of the border dummy is removed from the change in inflation. We do the same for all countries that border with Russia. I hope that this helps.
Thank you for your comment. The ordinate in Figure 1 of your paper plots the change in the rate of headline inflation 2020-2022 versus 2010-2019. The ordinate value for the data point NOR lies between minus 0.010 and minus 0.020. NOR = Norway. Table 1, Part I, Column (1) at "Norway" reads plus 0.0140. From this observation, I deduced that the data point for NOR in Figure 1 is incorrectly plotted. Now, there may be an adjustment which you and your co-author have made which yields an ordinate value of minus 0.0140 in Figure 1, in which case the data point for NOR in Table 1, Part I, Col. (1) is erroneous (i.e., Norway experienced disinflation between 2010-2019 and 2020-2022). However, when I look at the regression coefficient values and standard errors of the coefficient values appearing in Table 5, Column "Headline CPI inflation rate" labelled "(1)", the constant = 0.0134 (0.0037) and the slope = 0.369 (0.099) I find that your coefficient values and the coefficient values that I obtain in a linear OLS regression coefficients with NOR = ( - 0.053, + 0.014 ), i.e., constant = 0.0134 (0.00374) and slope = 0.3661 (0.0985). When I set NOR = ( - 0.053, - 0.014), the regression coefficients change to constant = 0.00868 (0.0033) and slope = 0.4999 (0.087).
Based on the agreement between the regression model in Table 5 (column "(1)") and the Excel worksheet regression model constructing using the data in Table 1, Part I, Col. "(1)" for the ordinate, and the data in Table 2 Col. "(1)" and Col. "(3)", and the data in Table 3 Col. "(2)", the only conclusion that I can fairly draw is that the data point labelled "NOR" is incorrectly plotted at ( - 0.053, - 0.014). It likely results from an inadvertent typographical error in the worksheet from which Figure 1, appearing on page 44 of your paper, was generated.
Given that the regression model coefficient values set out in Table 5 Col "(1)" agree with the Table 1 Part I, Table 2, and Table 3 data, the specific observations and conclusions drawn in your paper are unaltered. Only Fig. 1 would need to amended to bring the plot into agreement with the data and the corresponding regression model statistics.
Hi, we have a regression with two regressors, and we want to isolate the effect of the rescaled change in spending. So, we need to control for the effect of the border dummy. We are going to adjust the caption to make this clearer, but we already say what we do: See note 30, under the Caption of Figure 1 to see how we compute the variable on the vertical axis.
Thanks again for your explanation. I ought to have read the notes to the tables and footnote #30 more closely. Hungary, Norway, and Poland border either UKR or RUS or both, as you note. Because "bordering UKR or RUS" was only significant from February 2022, the more appropriate censuring measure should have been to drop HUN, NOR and POL from the regression model. Eliminating HUN, NOR and POL from the model does not alter the regression coefficients -- the linear OLS regression model coefficients become slope = 0.4264 (0.0876) and intercept = 0.00792 (0.0031), r^2 = 0.597 and the F-statistic = 23.7.
Excluding the EUR mean and adding only those Euro-zone countries (unweighted) that do not border either UKR or RUS yields a linear OLS regression model with slope = 0.364 (0.076), intercept = 0.0108 (0.00256), r^2 = 0.457 and F-statistic = 22.74. Excluding Euro-zone countries that border UKR or RUS in place of the EUR mean does not alter the positive relation between the change in inflation rate between 2010-19 and 2020-22 and government expenditure scaled by the product of government debt and the duration of government debt.
I withdraw my earlier comments posted in this thread. Thanks again for your comments and patience.
Thanks for the interest in our work and for taking the time to check the results.
The first equation appearing in the blog post above is derived from the accounting identity of the Department of The Treasury's general account with the Federal Reserve Bank.
The first equation is given by Bₜ₋₁ / Pₜ = 𝔼ₜ{∑ₙ₌₀ βⁿ∙sₜ₊ₙ} where I have replaced the index letter j appearing in the infinite series with the letter n for convenience. Using the lag operator, L , we can represent the first equation by L∙Bₜ / Pₜ = 𝔼ₜ{sₜ}/[1 – β∙L⁻¹] (see, J. H. Cochrane, 2021, "Fiscal Theory of the Price Level" for examples of the use of the lag-operator in deriving infinite series representations of state transition equations in discrete-time economic models).
Now, multiply both sides of the lag operator model equation by the denominator on the RHS of the equals sign, i.e., multiply L∙Bₜ / Pₜ = 𝔼ₜ{sₜ}/[1 – β∙L⁻¹] through by [1 – β∙L⁻¹] to obtain [L – β]∙Bₜ / Pₜ = 𝔼ₜ{sₜ}.
Expand this equation and multiply both sides by Pₜ to find the accounting identity of the Department of The Treasury's general account with the Federal Reserve, i.e., Bₜ₋₁ – β∙Bₜ = Pₜ∙𝔼ₜ{sₜ}. But, note that 𝔼ₜ{sₜ} = sₜ and that Pₜ∙sₜ is the nominal dollar primary surplus, Sₜ . Ergo, by 'reverse-engineering' of the fundamental equation of the fiscal theory of the price level, we have found the Department of The Treasury's cash accounting identity, i.e., Bₜ₋₁ – β∙Bₜ – Sₜ = 0. The discount factor β is shown as being time-invariant in the fundamental equation of the fiscal theory of the price level; but, as is well-known, the discount factor varies with calendar time and primarily in accordance with temporal monetary policy (Fedfunds rate, etc.)
Plotting the data series suggested in the blog post against calendar time is instructive for those whose interests extend to ascertaining patterns in economic time series in order to explain monetary policy variations. I offer the following FRED chart as a starting point: https://fred.stlouisfed.org/graph/?g=1nTvM . The data series plotted are:
Blue trace:
- (Net lending or net borrowing(-), NIPAS: Government: Federal+Federal government current expenditures: Interest payments)/(Federal Debt Held by the Public/1000)*100 (units: percent).
Red trace:
- Unemployment Level/Employed, Usually Work Full Time*100 (units: percent).
Green trace:
- Consumer Price Index for All Urban Consumers: All Items in U.S. City Average (units: percent-change, year-over-year).
Orange trace:
- Federal Funds Effective Rate (units: percent per year).
The inflation rate (green trace) appears to follow a 'reversion-to-the mean' process, as does the unemployment rate (red trace). The primary surplus to debt ratio, up until Q2-2013, followed a recovery process after each recession, but after Q2-2013 that process no longer holds (for whatever reason) and the primary surplus to debt ratio has remained negative since then. This suggests that there has been a fundamental change in monetary and fiscal policy in the U.S., starting from the end of Q2-2013. If so, then it is worth further investigation inasmuch as it may have implications for U.S. price stability going forward, according to the fiscal theory of the price level according to the fundamental equation of the FTPL.
Not sure if I'm missing something, but I think your "primary surplus/gdp" graph is wrong. You say "Surplus is net lending or borrowing, AD02RC1Q027SBEA, primary surplus subtracts interest payments A091RC1Q027SBEA" But if I take the first series minus the second, I don't get that. I get this: https://fred.stlouisfed.org/graph/?g=1o5sF It is never positive.
Finally, I ADDED the two series and divided by real GDP (GDPC1) although most measures or real or potential real look the same... and I get a graph that looks like yours. Here's the link: https://fred.stlouisfed.org/series/AD02RC1Q027SBEA#0 .
If your graph has them added, then the comments/stories around it are off too.
I suspect my graph with them subtracted isn't right because we did indeed run primary surpluses in the late 1990s... but what data to look at?
Any thoughts?
I'm very interested in finding good data to play with and think about FTPL. So, I'd be interested in any suggestions on good time series for real primary surpluses or appropriate measures of debt, etc. I plan to look at these globally over the summer, but wanted to start with the best data I know of which is USA FRED data so I feel confident that I'm looking at what I want to look at then I can worry about non-US measures.
Thanks.