Figure 1: Panel (a): three measures of the price level. Panel (b):
two measures of output. All series use 1926 base year. Vertical lines
mark when the U.S. abandoned the gold standard. Sources: Balke and Gordon (1986),
Federal Reserve Board, BEA and BLS from NBER Macrohistory
Database.
The figure has two panels that plot price levels and output from 1928 to 1940. The figure shows the size of the recession and the timing of the subsequent recovery in April 1933.
Panel (a) plots three price level series with a y-axis ranging from 60 to 100 and all series have 1926 as their base year. The GNP deflator series is at 100 near 1930 and then drops to 75 by April of 1933. Thereafter, the series rises towards 85. The CPI series is at 100 in 1928 and drops to 75 by 1933 and then rebounds slightly. The WPI series start at 100 in 1928 and drops to 60 by 1933 and then rises towards 85 by 1938.
Panel (b) plots two output series with a y-axis ranging from 60 to 140 and all series have 1926 as their base year. The Real GNP series is at a peak of 110 by 1930. Thereafter, the series drops to 70 by 1933 where it then rebounds to about 110 throughout the rest of the period shown. The IP series starts at a peak of 125 by 1930. Thereafter, the series drops to 60 by 1933 and then rebounds to 140 by 1940.
Figure 2: Consumer price index since the 1834 Coinage Act set the
price of one ounce of gold at $20.67. Rescaled to make mean from
1834–1933\(=\)100. Sample includes
periods when convertibility was suspended. Source: Officer and Williamson (2018)
and authors’ calculations.
The figure shows the consumer price index throughout the 19^(th) and the early 20^(th) centuries to show how far it had dropped in the Great Depression and how Franklin D. Roosevelt’s pre-Depression target was 60 percent higher than the longer-run average. The x-axis ranges from 1830 to 1933 and the y-axis ranges from 70 to 185. There is horizontal line at 100 and a second series that plots the consumer price index rescaled so that the average from 1834 to 1933 equals 100. This series starts out at 80 in 1830 and remains below there until 1860 when it climbs to 140. Starting around 1860, the series begins a long descent towards 80 which it reaches around 1895. Thereafter, the series climbs above 180 by 1920 where it then drops to roughly 160 which is the average for the years 1921-1929. From 1929 to 1933, the serie5s declines towards a value of 120.
Figure 3: Panel (a): surpluses defined as total receipts less
expenditures, ordinary or total; primary surplus is gross surplus less
net interest payments. Panel (b): fiscal impulses defined as primary
surplus as a percentage of GNP and primary surplus as a percentage of
the market value of gross debt. Vertical line marks when the U. S.
abandoned the gold standard. Sources: Federal Reserve Board (1943) from
NBER Macrohistory Database, and Balke and Gordon (1986). See Appendix B for details on the
data.
The figure shows surpluses to compare and contrast fiscal impulses from Keynesian Hydraulics and unbacked fiscal expansion.
Panel (a) plots three series of surpluses: primary, gross, and ordinary. The x-axis ranges from 1928 to 1940 and the y-axis from -12 to 4. All three series are tightly correlated from 1928 to 1933 where they are slightly above zero until 1930 and then decline towards -6 percent. In the 1933, the primary and gross series fluctuate between -12 and -2 percent and the ordinary series fluctuates near zero. The figure shows that ordinary surpluses were largely positive in the unbacked fiscal expansion period while primary and gross surpluses were largely negative.
Panel (b) plots two series: surplus/GNP and surplus/Debt. The x-axis ranges from 1928 to 1940 and the y-axis ranges from -30 to 5. The surplus/GNP series is slightly below zero until 1930 where it then declines towards a trough of -5 percent by 1933. It fluctuates near this trough for the remainder of the period shown. The surplus/Debt series is just below zero until 1930. From 1930 to 1933, the series declines towards its trough of -30 percent. It rebounds to about -5 percent by 1938. The panel shows that surplus/debt is much larger than surplus/GNP.
Figure 4: Keynesian hydraulics and wealth effects. Impulse responses
of the price level to persistent spending and transfer shocks with an
autoregressive coefficient of 0.5, long-term debt averaging a duration
of 6 years, \(\delta\) is 40 percent of
annual output, and monetary policy responsiveness, \(\phi = 0.5\).
The figure compares and contrasts the impulse responses of the price level to shocks to transfers and purchases with and without wealth effects. The x-axis ranges from 0 to 22 and the y-axis ranges from -0.1 to 1.1. There are four series. The first series is the price level response to transfers with only Keynesian hydraulics and is at zero for the entire period shown. The second series is the response of the price level with Keynesian Hydraulics and wealth effects. The series starts at 0 and then rises towards 0.5 where it remains from x-value 6 for the remainder of the period shown. The third series is the response to purchases with Keynesian hydraulics only. It rises to 0.7 by x-value 3 and then declines towards -0.1 which it reaches by x-value 12. The fourth and final series is the price level response to purchases with Keynesian Hydraulics and wealth effects. The series rises to 1.1 by x-value 4 and declines towards x-value 0.5 by period 9 where it remains thereafter. The figure shows that price level responses are larger with hydraulics and wealth effects than with only hydraulics.
Figure 5: Responses to unanticipated increases in expenditures. VAR is
recursive in the order \((F^o/B,F^e/B,T/B,P,Y)\), where the three
fiscal variables are scaled by the market value of debt. Solid lines are
modes and dashed lines are 68 percentile probability bands based on
500,000 draws from the posterior distribution of all the VAR parameters.
Appendix C reports the full set of impulse
response functions [figure C.2].
(a) Responses to ordinary expectations
(b)
Responses to emergency expenditures
The figure has 5 panels of impulse response functions. The two columns show impulse responses to ordinary and emergency expenditures, respectively. The first row is expenditures, the second is the price level, and the third is real GNP. All figures have x-axes ranging from 0 to 36. Overall, the figure shows that the responses of the price level and real GNP are larger for emergency expenditures (right column) than ordinary expenditures (left column).
The top-left panel has a y-axis ranging from 0 to 0.6. The ordinary expenditures series starts at 0.6 at x-value 0 and then declines towards y-value 0 which it reaches by x-value 15. There are 68 percent probability bands surrounding the series and these are above from x-value 0 to 10.
The middle-left panel has a y-axis ranging from 0 to 0.025. The price level series starts at 0 and rises to 0.0025 by x-value 5 and remains there for the rest of the period shown. The 68 percent probability bands are only slightly above zero on impact and then drop below zero.
The bottom-left panel has a y-axis ranging from -0.01 to 0.06. The real GNP series starts at 0 and ticks above thereafter. The 68 percent probability bands never rise above zero.
The top-right panel has a y-axis ranging from 0 to 0.06. Like the ordinary expenditures series in the top-left panel, the emergency expenditure series in this panel rises to 0.05 on impact and declines towards 0 which it reaches at x-value 30.
The middle-right panel has a y-axis ranging from 0 to 0.0025. The price level series is at 0.0025 on impact and rises towards 0.01. The 68 percent probability bands are above zero for the entire period shown.
The bottom-right panel has a y-axis ranging from 0 to 0.06. The real GNP series at 0 on impact and rising to 0.02 thereafter. Except for the first few x-values, the 68 percent probability bands are above zero.
Figure 6: Responses to an unanticipated decrease in the primary
surplus in the unbacked fiscal expansion period (April 1933 to June
1940). Solid lines are modes and dashed lines are 68 percentile
probability bands based on 500,000 draws from the posterior distribution
of all the VAR parameters. Appendix C reports
the full set of impulse response functions [figure C.3].
The figure has 6 panels showing the impulse responses of economic variables to a 1 standard deviation decrease in primary surpluses. The x-axes range from 0 to 36 months. The figure shows that prices, GDP, gold, and the monetary base all rise in response to a decrease in primary surpluses.
The top panel has a y-axis ranging from -4 to 1 x 10^-4 and shows that the interest rate response series is at 0 on impact and declines towards -2x10^-4 by x-value 7 where it remains thereafter. The 68 percent probability bands are never below zero.
The second panel has a y-axis ranging from 0 to 0.025 and shows that the monetary base series is at 0 on impact and rises to 0.1 by period 5 where it remains thereafter. The 68 percent probability bands are always above zero.
The third panel has a y-axis ranging from -0.2 to 0 and shows that the primary surplus series is at -0.2 on impact and returns to 0 by period 7. The 68 percent probability bands are always below zero.
The fourth panel has a y-axis ranging from -2 to 10^-3 and shows that the market value of debt rises to 2x10^-3 on impact and settles at 5x10^-3 from x-value 5 onward. The 68 percent probability bands are near zero initially and dip below around x-value 10.
The fifth panel has a y-axis ranging from 0 to 0.05 and shows that the gold stock rises from 0 on impact to 0.03 by x-value 5 where it remains thereafter. The 68 percent probability bands are always above 0.
The sixth panel has a y-axis ranging from 0 to 6 x10^-3 and shows that the price level rises from 0 on impact to 4x10^-3 by x-value 10. The 68 percent probability bands are always above 0.
The final panel has a y-axis ranging from 0 to 0.02 and shows that real GNP rises from 0 on impact to 0.01 by x-value 6. The 68 percent probability bands are always above 0.
Figure 7: Modal response of debt-GNP ratio to an unanticipated
decrease in the primary surplus in the unbacked fiscal expansion period
(April 1933 to June 1940), computed from figure 6.
The figure shows the modal response of the debt-GNP ratio to an unanticipated decrease in the primary surplus in the unbacked fiscal expansion period. The y-axis ranges from -10 to 2 x 10^-3 and the x-axis ranges from 0 to 36. The series starts at 2x10^-3 at x-value 2 and declines towards -8 by x-value 7. The figure shows that most of the responses are eventually negative.
Figure 8: Output multipliers from a $1 increase in the primary
deficit, calculated as \(\Delta Y_{t+k}/\Delta
S_t\) at horizon \(k\). Solid
line is posterior mode from the identified model underlying figure 6, dotted-dashed lines
are 68 percent credible sets for that model, and dashed lines are
posterior modes from alternative recursive orderings. Top panel uses the
mean of real GNP over the full sample, April 1933 to June 1940; bottom
panel uses the mean over the first year of the sample.
The figure shows the output multipliers for a $1 increase in the primary deficit for the full sample (top panel) and the first year (bottom panel). The x-axes range from 0 to 36 on both panels. The figures shows that the multipliers are unambiguously large and positive.
The top panel shows the full sample output multiplier with the y-axis ranging from 0 to 10. The series starts at 0 and ten rises towards 4 by period 5 where it remains thereafter. There are two series representing the posterior modes from alternative recursive orderings and they closely track the main series. The 68 percent credible sets are always above zero.
The bottom panel shows the first-year output multiplier with the y-axis ranging from 0 to 8. The series starts at 0 and then rises towards 4 where it remains thereafter. There are two series representing the posterior modes from alternative recursive orderings and they closely track the main series. The 68 percent credible sets are always above zero.
Figure 9: Responses to an unanticipated increase in the supply of gold
in the unbacked fiscal expansion period (April 1933 to June 1940). Solid
lines are modes and dashed lines are 68 percentile probability bands
based on 500,000 draws from the posterior distribution of all the VAR
parameters.
The figure has 7 panels showing the impulse responses of economic variables to a 1 standard deviation increase in the supply of gold. The x-axes range from 0 to 36 months. The figure shows that none of the economic variables other than gold itself have a statistically significant response to the shock.
The top panel has a y-axis ranging from -3 to 3 x 10^-4 and shows that the interest rate response series is slightly above 0 on impact and fluctuates in its vicinity thereafter. The 68 percent probability bands always encompass zero.
The second panel has a y-axis ranging from -5 to 12x10^-3 and shows that the monetary base series is at 0 on impact and rises to 2.5x10^-3 by period 3 where it remains thereafter. The 68 percent probability bands always encompass zero.
The third panel has a y-axis ranging from -0.02 to 0.015 and shows that the primary surpluses are at -0.1 on impact and return to 0 by period 5. The 68 percent probability bands always encompass zero.
The fourth panel has a y-axis ranging from -5 to 5x10^-3 and shows that the market value of debt rises to 3x10^-3 on impact and declines towards 0 by x-value 30. The 68 percent probability bands are always encompassing zero.
The fifth panel has a y-axis ranging from 0 to 0.055 and shows that the gold stock rises to 0.05 on impact and slowly declines to 0.025 by the end of the period shown. The 68 percent probability bands are always above 0.
The sixth panel has a y-axis ranging from -4 to 2 x10^-3 and shows that the price level is 0 on impact and declines towards -2x10^-3 by the end of the period shown. The 68 percent probability bands always encompass zero.
The final panel has a y-axis ranging from -10 to 7x10^-3 and shows that real GNP is close to zero for the period shown. The 68 percent probability bands always encompass 0.
Figure 10: Marginal distributions of responses to the shocks that
satisfy Friedman and Schwartz’s recovery comovements. Median (solid), 68
percentile (dashed), 90 percentile (dashed-dotted), based on 10,000
draws.
The figure has 7 panels showing the marginal distributions of impulse responses of economic variables. The x-axes range from 0 to 36 months. The figure shows that 95 percent of identifications produces declining surpluses.
The top panel has a y-axis ranging from -5 to 3 x 10^-3 and shows that the holding period return response series is near 0 on impact and remains near zero thereafter. The 68 and 90 percent probability bands always encompass zero.
The second panel has a y-axis ranging from 0 to 20x10^-3 and shows that the monetary base series is at 5 on impact and rises to 10 by period 5 where it remains thereafter. The 68 and 90 percent probability bands are always above zero.
The third panel has a y-axis ranging from -0.15 to 0 and shows that the primary surplus series is at -0.15 on impact and returns to 0 by period 7. The 68 and 90 percent probability bands are always below zero.
The fourth panel has a y-axis ranging from -5 to 10^-3 and shows that nominal bonds rise to 2x10^-3 on impact and settles at 3x10^-3 from x-value 3 onward. The 68 and 90 percent probability bands always encompass zero.
The fifth panel has a y-axis ranging from 0 to 0.05 and shows that the gold stock rises from 0.015 on impact to 0.03 by x-value 3 where it remains thereafter. The 68 and 90 percent probability bands are always above 0.
The sixth panel has a y-axis ranging from 0 to 6 x10^-3 and shows that the price level is at 2x10^-3 on impact and rises to 4x10^-3 by x-value 5. The 68 and 90 percent probability bands are always above 0.
The final panel has a y-axis ranging from 0 to 0.02 and shows that real GNP rises to 0.005 on impact and then to 0.01 by x-value 5. The 68 and 90 percent probability bands are always above 0.
Figure 11: Distribution of the present value of primary surpluses
associated with shocks that satisfy Friedman and Schwartz’s recovery
comovements. Based on 10,000 draws of \(\alpha_i\).
The figure shows the distribution of the net present value of primary surpluses associated with shocks that satisfy Friedman and Schwartz’s recovery comovements. The x-axis ranges from -0.5 to 0.2 and the y-axis ranges from 0 to 700. The distribution is roughly bell-shaped with a tight clustering around a mean and median of -0.15. The figures shows that most of the net present value of primary surpluses are negative.
Figure 12: Price level and interest rates. The GNP deflator (\(1926=100\)) is the solid line (right
scale), the commercial paper rate is the dashed line, and the New York
Fed discount rate is the dotted-dashed line. The vertical line marks
when the U.S. abandoned the gold standard. Sources: Federal Reserve Board (1943).
The figure shows interest rates and the GNP deflator with the x-axis ranging from 1928 to 1941. The commercial paper rate and the NY Fed Discount Rate are plotted on the left y-axis from 1 to 8. The two series are near 6 in 1928 and then in 1930 the series begin a decline to 2 that is interrupted by an increase to 3.5 in 1932. Thereafter, they decline below 2 and remain there for the rest of the period shown. The GNP deflator is plotted on the right axis which ranges from 70 to 100. The GNP deflator series is near 100 in 1930 and then drops to 70 by April of 1933. Thereafter, the series rises towards 87.
Figure 13: Market value of gross debt as percentage of nominal GNP
(solid line) and par value of gross debt in billions of dollars (dashed
line); vertical axis measures both percentage and billions of dollars.
Vertical line marks when the U.S. abandoned the gold standard. Sources:
Authors’ calculations, Balke and Gordon (1986).
The figure shows that nominal debt was rising throughout the unbacked fiscal expansion period resulting in stable debt to GDP ratio. The x-axis ranges from 1928 to 1940 and the y-axis ranges from 15 to 50.
The figure plots the market value of gross debt as a percent of GDP. The series is at a trough near 15 in 1930 and then rises to a value of 40 by 1933. Thereafter, it levels off until 1937 and then climbs again to 45 in 1938 where it remains thereafter. The nominal debt series starts near a value of 15 and rises to 45 over the period shown.
Figure 14: Panel (a): surprises in real returns on bond portfolio as
percentage of market value of outstanding debt. Panel (b): decomposition
of surprises in real returns on bond portfolio into components due to
unanticipated inflation (solid line) and unanticipated bond prices
(dashed line). See appendix F for details.
Vertical line marks when the United States abandoned the gold standard.
Source: Hall et al. (2021), CRSP, and
authors’ calculations.
The figure has two panels the left shows the surprise returns as a fraction of debt and the right shows the decomposition of the surprise returns into components attributable to the price level or bond prices. The x-axes in both panels range from 1926 to 1941 and the panels show that surprises were mostly negative and coming from the price level in the post-1933 unbacked fiscal expansion period.
The left panel has a y-axis ranging from -1.6 to 1.6. The surprise component of returns as a fraction of debt series starts near 0.5 in 1926 and is mostly positive until 1933 with fluctuations ranging from -0.5 to 1.6. From 1933 onward, the series is mostly negative fluctuating near -0.5 with a range of -1.6 to 1.5.
The right panel has a y-axis ranging from -1.5 to 1.5 and there are two series. The surprise return due to inflation series fluctuations around 0 from 1926 to 1930, rises to 1.5 and then fluctuates around -0.25 from 1933 onward. The surprise return due to bond prices fluctuates around 0 for the period shown with a drop to -1.5 in 1932.
Figure 15: Relative price of the bond portfolio is the ratio of the
real market value to the nominal par value of debt. Vertical line marks
when the U.S. abandoned the gold standard. Source: Authors’
calculations.
The figure plots the relative price of the bond portfolio and shows that it peaked in 1933 and declined thereafter. The x-axis ranges from 1928 to 1940 and the y-axis ranges from 1 to 1.4. The series has y-value 1 in 1928 and rises to y-value 1.4 where it peaks in 1933. Thereafter, the series declines towards y-value 1.25 by 1934, and it remains there for the rest of the period shown.
Figure B.1: Annual Reports
The figure has two panels each showing a table from the Annual Reports showing the federal interest receipts series. The left panel shows the series from the 1929 Annual Report and highlights that the column is called “Interest, exchange, and dividends on capital stock.” The right panel is from the 1928 Annual Report and highlights a series called, “Interest, premium, and discount”. It also highlights a footnote saying that this table entry includes receipts credited direct to appropriations.
Figure B.2: 1927 Annual Report, page 431
The figure is a table from the 1927 Annual Report and highlights the line items contained in the federal interest receipts series title, “Interest, premium, and discount”.
Figure B.3: 1922 Annual Report, page 107
The figure is a table from the 1922 Annual Report and highlights the line items that total federal interest rate receipts series. These items include “Discount on bonds, notes, and certificates purchased”, “Gain by exchange”, “Interest on farm loan bonds”, “Interest on foreign obligations, sale of surplus property, War Department”, “Interest on public deposits”, “Interest on loans to railroad companies”, and “Interest on advance payments to contractors”.
Figure B.4: Federal receipts and total expenditures aggregated by fiscal
year, billions of dollars. Source: Department of the Treasury (various). See
table B.4 for details.
The figure shows federal receipts and expenditures aggregated by fiscal year in billions of dollars compiled by the authors. The x-axis ranges from 1920 to 1941 and the y-axis ranges from 2 to 10 billion. From 1920 to 1930, receipts and expenditures track each other closely. They start at $7 billion in 1920, drop to $4 billion by 1922 where they remain until 1930. Thereafter, the expenditure series climbs toward $9 billion by 1939. The receipts series drops to $2 billion by 1932 and the rises to $6 billion by 1939 where it remains.
Figure B.5: Federal receipts aggregated by fiscal year, billions of
dollars.
(a) Comparing receipts to Firestone (1960)
(b) Comparing receipts to NBER
Source: Department of the Treasury (various). See table B.4 for details; Firestone (1960); NBER Macrohistory
database (m15004a, m15004b).
The figure shows that the authors’ aggregated federal receipts series closely matches those of Firestone (1960) and the NBER. The x-axis is from 1919to 1941 on both panels and the y-axis from $2 to $7 billion.
The left panel compares the authors’ receipts series to Firestone. The two series follow the description in Figure B.4 with the Firestone series being below the authors’ from 1930 to 1934 and again from 1939 to 1940.
The right panel compares the authors’ receipts series to those of the NBER. The three series follow the description in Figure B.4 with the two NBER series being below the authors’ series from 1930 to 1934.
Figure B.6: Federal receipts aggregated by fiscal year, billions of
dollars.
(a) Receipts, 1936 vintage
(b) Receipts,
1940 and 1941 vintages
Source: Department of the Treasury (various). See
table B.4 for details.
The figure shows that the authors’ aggregated federal receipts series closely matches those found elsewhere in the Annual Reports. The x-axis is from 1919 to 1941 on both panels and the y-axis from $2 to $7 billion.
The left panel compares the authors’ receipts series to the 1936 Annual Report vintage. The two series follow the description in Figure B.4 with the 1936 vintage series above the authors’ from 1933 to 1935.
The right panel compares the authors’ receipts series to the vintages from the 1940 and 1941 Annual Reports. The three series follow the description in Figure B.4 with the 1940 vintage below the authors’ series from 1930 to 1932 and the 1941 below from 1930 to 1932 as well as 1936 to 1941.
Figure B.7: Federal expenditures aggregated by fiscal year, billions of
dollars.
(a) Total and ordinary expenditures (1920-1933)
(b) Total expenditures, 1934-1940)
Source: Department of the Treasury (various). See table B.4 for details; Firestone (1960); NBER Macrohistory
database (m15004b, m15004c, m15004d); Romer (1992).
The figure shows the authors’ series for total and ordinary expenditures and compares them to those from Firestone, Romer, and the NBER.
The left panel is from 1920 to 1933 and has a y-axis ranging from $2.5 billion to $7 billion. There are five series and the three for ordinary expenditures from the authors’, Firestone, and NBERb are almost indistinguishable. These series are at $6.5 billion in 1920, decline to about $3 billion in 1923, remain there until 1926 when they steadily climb towards $4.5 billion by 1933. The ordinary expenditure series from Romer is from 1926 to 1936 and is mostly below the other three series. The final series, total expenditures, is at $6.5 billion in 1920, declines to $3.75 billion by 1923 and remains there until 1926 where it steadily climbs towards $5 billion by 1933.
The right panel is from 1934 to 1940 and has a y-axis ranging from $6 billion to $10 billion. There are five series and the three for Firestone, NBERc, and NBERd are largely similar. These series start at $6.75 billion in 1934, rise to $8.5 billion by 1936, decline to $7.75 billion by 1938, and then climb to $9 billion by 1940. The series from Romer is always about $0.25 to $0.5 billon below, but follows the same pattern. Similarly our series is eh same or about $0.25 to $0.5 billion above.
Figure B.8: Federal expenditures aggregated by fiscal year, billions of
dollars.
(a) Ordinary expenditures, 1920-1933
(b)
Total expenditures, 1920-1933
Source: Department of the Treasury (various). See
table B.4 for details.
The figure shows a comparison of the authors’ ordinary and total expenditure series to the vintages from the 1934 Annual Report. Both panels have x-axis ranging from 1920 to 1935 and y-axes ranging from $3 billion to $7 billion.
In the left panel, the ordinary expenditures series follows the pattern described in figure B.7. The 1934 vintage is slightly higher in 1920, the same thereafter, and higher again in 1934.
In the right panel, the total expenditure series follows the pattern described in figure B.7. The 1934 vintage is slightly higher in 1934.
Figure B.9: Federal expenditures aggregated by fiscal year, billions of
dollars.
(a) Ordinary expenditures, 1920-1936
(b)
Total expenditures, 1920-1936
(c) Ordinary expenditures,
1920-1940
(d) Total expenditures, 1920-1940
Source: Department of the Treasury (various). See table B.4 for details.
The figure shows a comparison of the authors’ expenditure series to those from the 1935, 1936, 1937, 1939, and 1940 vintages from the Annual Reports.
Panel B.9(a) in the top left has an x-axis from 1920 to 1936 and a y-axis from $2 to $9 billion. There are four series: the authors’ two series for ordinary and general expenditures and ordinary expenditures for the 1935 and 1936 vintages of the Annual Reports. All series but the authors’ series for general expenditures track each other closely. These three series are at $6 billion in 1920, drop to $3 billion by 1922 and remain there until 1930 when the climb to their peak of $8 billion by 1926. The general expenditure series is considerably lower than the other three at $3 billion in 1934 and $6 billion by 1936.
Panel B.9(b) in the top right has an x-axis from 1920 to 1936 and a y-axis from $3 billion to $10 billion. There are three series: the authors’ total expenditures and total expenditures from the 1935 and 1936 vintages of the Annual Reports. All three series track each other closely, starting at $6.5 billion in 1920, dropping to $3.5 billion in 1922 and remaining there till about 1930 where they rise to a peak of $9 billion by 1936.
Panel B.9(c) in the bottom left has an x-axis from 1920 to 1940 and a y-axis from $2 billion to $10 billion. There are five series : two series for the authors’ calculation of ordinary and general expenditures and three from the 1937, 1939, and 1940 vintages of the Annual Reports. All of the series but the general expenditures series follow each other closely and have the same pattern as panel B.9(a) until 1936. After 1936, these four series peak and fluctuate around $8 billion. The general expenditure series is also like that in Panel B.9(a) until 1936. Thereafter, it fluctuates around $5 billion.
Panel B.9(d) in bottom right panel has an x-axis from 1920 to 1940 and a y-axis from $3 billion to $10 billion. There are four series: the authors’ series of total expenditure and those from the 1937, 1939, and 1940 vintages of the Annual Reports. All four series track each other closely and are the same as Panel B.9(b) until 1936. Thereafter, the series fluctuate around their peak of $8 billion.
Figure C.1: Actual and unconditional forecasts of variables in VAR using
the hyperparameters \(\lambda_0=0.6,\lambda_1=0.3,\lambda_3=1.0,\lambda_4=1.75,\mu_5=\mu_6=2.0\),
in the notation of Sims and Zha (1998).
The figure shows the actual and unconditional forecasts of economic variables in 7 different panels under the prior that was chosen from the model’s marginal data density. The x-axis is from 1933 to 1941 for all panels.
The top left panel shows the actual interest rates and the forecast. The y-axis is from 0.7 to 1.4. The forecast series starts at 1.3 in 1933 and gradually declines towards 0.8 by 1937 where it remains for the rest of the period shown. This tracks the actual series which starts at 1.4 in early 1934, drops to 0.7 by mid-1934 where it remains until 1937. Starting in 1937, the series rises to 1 for a few years and drops back to about 0.7 by the end of 1938.
The second left panel shows the actual primary surplus and the accompanying forecast. The y-axis is from -1.7 to 0. The forecast series is at -0.5 for the period shown and the actual series fluctuates around this value with notable dips in early 1934 and mid-1936.
The third left panel shows the actual government spending and its forecast. The y-axis is from 4 to 20. The forecast series starts at 4 and rises to 18 by the end of the period shown. The actual series fluctuates around this trend.
The fourth left panel shows the actual output series and its forecast. The y-axis is from 70 to 115. The forecast series rises from 70 to a value of 110 by the end of the period shown. The actual series fluctuates around this trend.
The top right panel shows the actual monetary base series and its forecast. The y-axis is from 7 to 22. The forecast is at 7 in 1933 and rises towards 20 by 1941. The actual series fluctuates around this trend.
The second right panel shown the actual market value of debt and its forecast. The y-axis is from 22 to 45. The forecast rises from 22 in 1933 to 45 by 1941 and the actual series fluctuates around this trend.
The third right panel shows the actual price level and its forecast. The y-axis is from 78 to 100. The forecast series starts at 78 in 1933 and rises to 86 by 1940. The price level series also rises following the same pattern, but with a peak closer to 100 in 1937.
Figure C.2: Full moving average representation of the five-variable
identified VAR estimated over the period April 1933 to June 1940.
Identification is recursive with variables ordered \(F^o/B,F^e/B,T/B,P,Y\). Solid lines are
maximum likelihood estimates; dashed lines are 68 percentile probability
bands based on 500,000 draws from the posterior distribution of all the
VAR parameters.
The figure shows the full moving average representation of the five-variable identified VAR over April 1933 to June 1940. The figure has a total of 25 panels with each row representing responses of ordinary expenditures, emergency expenditures, transfers, the price level, and output respectively. The columns are the respective shocks to ordinary expenditures, emergency expenditures, receipts, the price level, and real GNP. The x-axis ranges from 0 to 36 months on all panels.
The first column of panels shows the response of the row variables to a one standard deviation shock to ordinary expenditures which are in matrix position 1,1 in the figure and rise to 0.5 on impact and declines to 0 by period 11. Ordinary expenditures—matrix position 2,1—are on the margin of significance on impact, drop to -0.04 and returning to zero by x-value 20. The price level (4,1) is also on the margin of significance in periods 2 to 5. The series is at 0 on impact and rises to 2 several periods later and remains there in the rest of the period shown. Transfers and output, matrix positions 2,1 and 5,1, respectively, have responses close to zero that are not statistically significant.
The second column of panels shows the response of the row variables to a one standard deviation shock to emergency expenditures which are in matrix position 2,2 in the figure and rise to 0.3 on impact and declines to 0 by period 25. Emergency expenditures and transfers—matrix positions 1,2 and 3,2, respectively—are close to zero and not statistically significant. Both the price level and output—matrix positions 2,2 and 5,2, respectively—are statistically significant starting at 0 on impact and rising to 0.004 by period 20 for the price level and to 0.01 by period 20 for output.
The third column of panels shows the response of the row variables to a one standard deviation shock to receipts (not shown). Only transfers—matrix position 3,3—have a statistically significant response. The series rises to 0.1 on impact and then declines towards zero which it reaches by period 36. The response is statistically significant until period 25.
The fourth column of panels shows the response of the row variables to a one standard deviation shock to the price level which is in matrix position 4,4 in the figure and rises to 0.6x10^-3 and remains there for the period shown. Neither ordinary nor emergency expenditures—matrix positions 1,4 and 2,4, respectively--are statistically significant from zero. The response of transfers in matrix position 3,4 rises from 0 on impact to 0.015 by period 5 and declines towards 0 thereafter. The series is on the margin of statistical significance from periods 2 to 10. The output series is near 5 for the period shown and is statistically significant from 0 to 5.
The fifth column of panels shows the response of the row variables to a one standard deviation shock to real GNP which is in matrix position 5,5 and rises to 0.02 on impact and remains there throughout the period shown. The response of transfers and the price level—matrix positions 3,5 and 4,5, respectively—are near zero and not statistically significant. The response of ordinary expenditures in matrix position 1,5 is 0 on impact and then rises to 0.04 a few periods later and declines to zero by period 20. It is on the margin of significance in period 5. The response of emergency expenditures in matrix positions 2,5 is zero on impact and drops to -0.05 in period 5 with statistical significance from 0 to 10.
Figure C.3: Full moving average representation of the identified VAR
estimated over the period April 1933 to June 1940. Identification from
table C.2. Solid lines are maximum
likelihood estimates; dashed lines are 68 percentile probability bands
based on 500,000 draws from the posterior distribution of all the VAR
parameters.
The figure shows the full moving-average representation of the identified VAR estimated from April 1933 to June 1940. The columns are the responses for the system of equations given in section 5.4.1 of the main text to shocks to monetary policy, money demand, fiscal policy, government debt, the gold stock, the price level, and output. The rows are shocks to interest rates, money. The rows are the variable VARs which are interest rates, the monetary base, surpluses, debt, gold, the price level, and output. The shock to primary surpluses—column 3—has the most statistically significant responses. Most of the responses to the other shocks are statistically insignificant.
The first column shows the responses to a monetary policy shock and only the interest rate—matrix position 1,1--is statistically significant from zero rising on impact and falling towards zero by the end of the period shown.
The second column shows the responses to a money demand shock and only the monetary base—matrix position 2,2—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
The third column shows the response to a positive shock to primary surpluses—the main text shows a shock to primary deficits. The shock rises on impact and declines to zero after 5 periods as shown in matrix position 3,3. Interest rates in 1,3 rise on impact and decline towards zero by the end of the period shown, but are not statistically significant. The monetary base in 2,3 drops on impact and remains unambiguously negative. Government debt in 4,3 drops on impact, remains below zero, and is on the margin of statistical significance until period 5. The gold stock, price level, and output in 5,1, 6,1, and 7,1, respectively, drop on impact and remain unambiguously below zero thereafter.
The fourth column shows the responses to a government debt shock and the market value of government debt—matrix position 4,4—is the most statistically significant form zero on impact and remains above zero for the rest of the period shown. The monetary base and government spending—matrix positions 2,4 and 5,4, respectively--rise on impact and are on the margin of statistically significance. Interest rates in matrix position 1,4 drop in impact and are on the margin of statistical significance. All other response variables are near zero and not statistically significant.
The fifth column shows the responses to a gold stock shock and only the gold stock—matrix position 5,5—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
The sixth column shows the responses to a price level shock and only the price level—matrix position 6,6—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
The seventh column shows the responses to an output shock and only the gold stock—matrix position 7,7—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
Figure C.4: Full moving average representation of 8-variable VAR ordered
recursively—\(P,Y,F,T,R,M,G,B\)—estimated over the period
(April 1933 to June 1940). \(F\) is
government expenditures net of interest payments and \(T\) is tax receipts. Solid lines are
maximum likelihood estimates; dashed lines are 68 percentile probability
bands based on 500,000 draws from the posterior distribution of all the
VAR parameters.
The figure is similar to C.3 and shows the full moving-average representation of the recursively identified VAR estimated from April 1933 to June 1940. The columns are the responses for the system of equations given in section 5.4.1 of the main text to shocks to monetary policy, money demand, fiscal policy expenditures, fiscal policy transfers, government debt, the gold stock, the price level, and output. The rows are shocks to interest rates, money. The rows are the variable VARs which are interest rates, the monetary base, surpluses, transfers, debt, gold, the price level, and output. Transfers are included as fiscal policy variable in addition to surpluses. The shocks to primary surpluses and transfers—columns 3 and 4, respectively—have the most statistically significant responses. Most of the responses to the other shocks are statistically insignificant.
The first column shows the responses to a monetary policy shock and only the interest rate—matrix position 1,1--is statistically significant from zero rising on impact and falling towards zero by the end of the period shown.
The second column shows the responses to a money demand shock and only the monetary base—matrix position 2,2—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
The third column shows the response to a positive shock to primary deficits. The shock rises on impact and declines to zero after 5 periods as shown in matrix position 3,3. Interest rates in 1,3 drop on impact and rise towards zero by the end of the period shown with statistical significance for the first 10 periods. The monetary base in 2,3 rises on impact and remains unambiguously positive. Transfers in 4,3 are not statistically significant from 0. Government debt in 5,3 rises on impact, remains above zero, and is statistically significant until the end of the period shown. The gold stock, price level, and output in 6,1, 7,1, and 8,1, respectively, rise on impact and remain unambiguously below zero thereafter.
The fourth column shows the response to a one standard deviation increase in transfers. The shock rises on impact and declines to zero after 5 periods as shown in matrix position 4,4. Interest rates, the monetary base, primary deficits and the gold stock in matrix positions 1,4, 2,4, 3,4, and 6,4 respectively are close to zero and not statistically significant. Government debt, the price level, and output in positions 5,4, 7,4, and 8,4, respectively, drop on impact and are on the margin of statistical significance.
The fifth column shows the responses to a government debt shock and only government—matrix position 5,5—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
The sixth column shows the responses to a gold stock shock and only the gold stock—matrix position 6,6—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
The seventh column shows the responses to a price level shock and only the price level—matrix position 7,7—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
The eighth column shows the responses to an output shock and only the gold stock—matrix position 8,8—is statistically significant form zero on impact and remaining above zero for the rest of the period shown.
Figure C.5: Blanchard and Perotti (2002) output multipliers for
government expenditures net of interest payments. Solid lines are
posterior modes; dashed lines are 68 percent credible sets. Top panel
takes mean of expenditures to output from full sample; bottom panel
takes mean from first year of sample.
The figure shows the Blanchard-Perotti (2002) output multipliers for government expenditures net of interest payments for the full sample and the mean of the first year. The a-axis is from 0 to 36. The figure shows that the output multipliers are above zero and around 4.
The top panel shows the output multiplier for the full sample. The y-axis is from 0 to 15 and the series starts near zero and rises to 4 by x-value 5 where it remains for the rest of the period shown. The 68 percent credible sets surrounding the series are always above zero.
The top panel shows the output multiplier for the full sample. The y-axis is from 0 to 11 and the series starts near zero and rises to 4 by x-value 5 where it remains for the rest of the period shown. The 68 percent credible sets surrounding the series are always above zero.
Figure C.6: Blanchard and Perotti (2002) output multipliers for
government tax receipts. Solid lines are posterior modes; dashed lines
are 68 percent credible sets. Top panel takes mean of receipts to output
from full sample; bottom panel takes mean from first year of
sample.
The figure shows the Blanchard-Perotti (2002) output multipliers for net tax-receipts for the full sample and the mean of the first year. The a-axis is from 0 to 36. The figure shows that the output multipliers rise above 10 over the period shown.
The top panel shows the output multiplier for the full sample. The y-axis is from -10 to 40 and the series starts near zero and rises to 11 by the end of the period shown. The 68 percent credible sets surrounding the series encompass zero.
The top panel shows the output multiplier for the full sample. The y-axis is from -20 to 40 and the series starts near zero and rises to 18 by the end of the period shown. The 68 percent credible sets surrounding the series encompass zero.
Figure E.1: Measures of price levels and real economic activity. All
series use 1926 base year. Vertical line marks when the United States
abandoned the gold standard.
(a) Price levels
(b)
Output
Sources: Balke and Gordon (1986), Federal Reserve Board, BEA and BLS
from NBER Macrohistory Database.
The figure has two panels that plot price levels and output from 1920 to 1940. The figure shows the size of the recession and the timing of the subsequent recovery in April 1933.
Panel (a) plots three price level series with a y-axis ranging from 60 to 100 and all series have 1926 as their base year. The GNP deflator series is near 130 in 1920 and then falls to 100 where it remains till 1930 and then drops to 75 by April of 1933. Thereafter, the series rises towards 85. The CPI series is near 115 in 1920 and also drops to 100 where it remains until 1929. The series drops to 75 by 1933 and then rebounds slightly. The WPI series start at 170 in 1920, drops to about 100 and remains there until 1928 where it drops towards a trough of 60 by 1933. The series then rises towards 85 by 1938.
Panel (b) plots two output series with a y-axis ranging from 60 to 140 and all series have 1926 as their base year. The x-axis is from 1920 to 1940. The Real GNP series rises from about 70 in 1920 to 105 by 1930. Thereafter, the series drops to 70 by 1933 where it then rebounds to about 110 throughout the rest of the period shown. The IP series climbs from about 80 in 1920 to its peak of 125 by 1930. Thereafter, the series drops to 60 by 1933 and then rebounds to 140 by 1940.
Figure E.2: Exchange rates, inflation, and interest rates. Exchange
rates in dollars per foreign currency; inflation is annual (quarter over
four quarters prior). First vertical line marks when the United Kingdom
abandoned the gold standard; second line marks when the United States
abandoned the gold standard.
(a) Dollar valuation
(b) Interest rates& price level
Sources: Federal Reserve Board (1943).
The figure shows how departing the gold standard affected exchange rates and interest rates from 1920 to 1941.
Panel (a) shows the dollar/sterling on left y-axis ranging from 3.2 to 5.2 and the dollar/franc exchange rate on the right y-axis ranging from 0.02 to 0.09. The dollar/sterling series is at 3.7 by 1920 and then rises towards 4.8 by 1925 where it remains until 1930. Thereafter, the series drops to 3.4 when the UK departed the gold standard in 1931 and then rebounds to 5 when the U.S. departed the gold standard in 1933. The series remains near 5 until 1940 when it drops back to 3.6. The dollar/franc series is at 0.07 in 1920, rises to 0.09 a year later, and then drops to 0.04 by 1927 and remains there until 1933. When the U.S. abandons the gold standard in 1933, the series jumps to 0.065, stays there for a few years and then declines towards 0.02 by 1940.
Panel (b) shows interest rates and the GNP deflator with the x-axis ranging from 1920 to 1941. The commercial paper rate and the NY Fed Discount Rate are plotted on the left y-axis from 1 to 8. The two series are near 8 in 1920, drop to about 4 by 1922 and fluctuate around there until 1928 when they rise to about 6. In 1930 the series begin a decline to 2 that is interrupted by an increase to 3.5 in 1932. Thereafter, they decline below 2 and remain there for the rest of the period shown. The GNP deflator is plotted on the right axis which ranges from 70 to 100. The GNP deflator series is near 140 in 1920, drops below 100 in 1922 and then fluctuates around 100 until 1930. The series drops to 70 by April of 1933 and then rises towards 87.
Figure E.3: Monetary base and gold held by Federal Reserve banks.
Monetary base is currency in circulation plus non-borrowed reserves.
Vertical line marks when the United States abandoned the gold standard.
Source: Federal Reserve Board (1943) from NBER Macrohistory
Database.
The figure shows how the departure from the gold standard changed the monetary base and gold stock from 1920 to 1941.
The top panel has a y-axis ranging from 0 to 25 and series for the monetary base and the gold stock. The monetary base series is around by 7 million from 1920 to 1933 and then rises towards 20 million by 1941. The gold series is around 4 from 1920 to 1934 and it then jumps in 1934 to 7 when Roosevelt changes the dollar value. Thereafter, it climbs towards 20 by 1940.
The bottom panel has a y-axis ranging from 0.2 to 1 and plots the ratio of gold to the monetary base. The ratio is around 0.4 in 1920 and then jumps to 0.6 in 1922 where it remains until 1932. Thereafter, it drops to 0.5 where it remains until 1934. When the dollar price of gold is revalued, the ratio jumps to 0.8 in 1934 and it remains there for the rest of the period shown.
Figure F.1: Returns and innovations. Because there is a 0.99 correlation
coefficient between returns with clean and dirty prices, the figure only
shows series for dirty prices.
(a) Real and nominal price
returns
(b) Innovations decomposition from eq.
F.41)
The figure shows the real and nominal bond returns along with their decompositions attributable to movements in the price level vs. bond prices. The x-axis runs from 1926 to 1942 and shows that returns fell after the departure from the gold standard in 1933 and was due to both prices and the price level.
Panel (a) shows real and nominal bond returns with a y-axis ranging from 0.97 to 1.05. Both series are quite volatile and are closely correlated. In the pre-gold standard period from 1926 to 1933 both series are mostly above zero, fluctuating between 0.98 and 1.04 with the real return generally higher. In the post-gold standard period from 1933 to 1942 both series are mostly close to zero fluctuating between 0.97 and 1.02. The real return series is mostly lower than the nominal return series.
Panel (b) shows the surprise component of real returns decomposed into components attributable to the price level or bond prices. The y-axis ranges from -0.03 to 0.02. The innovation due to the price level series fluctuates close to, but above, zero in the pre-gold standard period from 1926 to 1933 with values ranging from -0.01 to 0.02. In the post gold-standard period from 1933 to 1942, the series fluctuates below zero with values ranging from -0.03 to 0.01. The innovation due to bond prices fluctuates near zero in the gold standard period with values ranging from -0.02 to 0.02. Thereafter, it fluctuates mostly below zero with values ranging from -0.025 to 0.02.