DEFENSE NEWS

The aim of “Defense News” is to reflect changes to planned government defense expenditure previously unanticipated by the public. More precisely, the aim is to chronicle changes in the information set available to an informed member of the public and calculate their implications for the expected present value of future defense expenditure. This variable can be thought of as capturing exogenous shocks to a key component of government spending.

The series for changes in the expected present value of government expenditure on defense for the United Kingdom in the interwar period has been constructed using a similar method to that employed by Ramey who mainly used information from Business Week (Ramey Reference Ramey2009).Footnote ⁷ Our key source is The Economist magazine, which was published weekly through the interwar period. This source gives details of defense estimates, which were usually published in government papers in February and March each year, but there were sometimes also supplementary estimates. The Economist gave a detailed yearly account of actual spending at the time of the annual budget in April. It also published quarterly running totals at the beginning of January, April, July, and October each year, and from time to time provided commentary on the prospects for defense spending in editorials and in featured news items.

In order to construct the series for “Defense News,” we examined every issue of The Economist from 1919 through 1938 and compiled a complete log of all entries relating to defense expenditure. The relevant material includes reports of defense estimates, announcements of policy changes with possible spending implications, special reports such as those on air force and naval expenditure in February 1936, commentaries on the credibility of government estimates and announcements, and predictions of future developments. Most of these items lend themselves to straightforward quantification. Wherever The Economist provides a clear indication as to its expectation of the likely course of future defense expenditure, we have accepted that as part of the public's information set. The volume of material varies considerably over the years; for example, there are many entries for the late 1930s but few for the late 1920s. Statistical information obtained from The Economist has been cross-checked against the detailed descriptions of British budgets provided by Bernard Mallet and Oswald George and by Basil Sabine (Mallet and George Reference Mallet and Oswald George1929, Reference Mallet and Oswald George1933; Sabine Reference Sabine1970).

Using the log, we made an estimate of the present discounted value of defense expenditure for 1920 quarter 1 and then updated it each quarter until 1938 quarter 4. In moving forward through time, for each quarter we compiled a present value figure based on the information set available in the previous quarter and one using the current information. The difference between the two estimates is “Defense News” for the quarter. Expected values were calculated at 1938 prices for a horizon of five years using a discount rate of 5.1 percent.Footnote ⁸ Much of the time there was no news, i.e., the information set was the same as in the previous quarter. Estimates of “Defense News” are reported in Table 2.

We rely on the major historical studies of military policy, such as those of John Ferris and George Peden to interpret the commentary of The Economist (Ferris Reference Ferris1989; Peden Reference Peden1979). The former fills in valuable background on the “Ten-Year Rule” (see below) and the latter reports archival information that bears out some claims relating to rearmament made by The Economist. Indeed, the main difficulty in constructing the Defense News series concerns the period after the policy of rearmament was announced in the 1935 Defense White Paper (Cmd. 4827). The White Paper made clear that there would be substantial additional defense expenditure in future but offered no details as to magnitudes or timing.

A full discussion of how the “Defense News” variable was constructed can be found elsewhere (Crafts Reference Crafts2012). Here we provide some context and briefly discuss the two episodes which require careful interpretation because government policy was in flux.

The interwar period started with fiscal consolidation following the explosion of the public debt during and immediately after World War I; for defense this meant disarmament. In 1919 the government set out its view that “normal expenditure” on the fighting forces would be £93.5 million per year (British Parliamentary Papers, Cmd. 376).Footnote ⁹ The plan was based on the assumption that no expeditionary force would be required for a great war in the next decade: the “10-Year Rule.” The interpretation of this rule was, however, disputed in budget planning discussions, with each service arguing that it had to be ready for a major conflict by 1929 while the Treasury wanted to delay the start of such preparations until 1929 at the earliest. The struggle between the Treasury and the armed forces lasted until 1927, when spending reached a steady-state level of £108 million.Footnote ¹⁰

TABLE 2

DEFENSE NEWS: ESTIMATES OF CHANGES IN NET PRESENT VALUE OF EXPECTED DEFENSE EXPENDITURE (£ million, 1938 prices)

Source:

Authors' own calculations, see the text.

The second phase covers the period from 1927 through 1934. Until 1932, defense policy was quite settled but then the 10-Year Rule was cancelled in March following the invasion of Manchuria and the attack on Shanghai by Japan. The next three years show modest increases in defense expenditure related first to a new naval construction program in 1933 and then to new squadrons for the Air Force in 1934. Already in 1932 Neville Chamberlain (Chancellor of the Exchequer) was contemplating a rise of defense expenditure of a further 10 percent by 1935 (Peden Reference Peden1979, p. 67). In 1934/35 defense expenditure was £124.7 million.

The third phase was rearmament. The new policy was announced, but with no spending commitments, in the Defense White Paper of March 1935 (British Parliamentary Papers, Cmd. 4827). This statement simply said that additional expenditure on the armaments of the defense services could no longer safely be postponed. In late 1935 the Cabinet agreed a total program of £1,075 million over the period 1936–1940 (Peden Reference Peden1979, p. 77). The pace of rearmament was further intensified when the Defense White Paper of February 1937 stated that it would be imprudent to contemplate total expenditure of less than £1,500 million over the next five years, while expenditure over the next two or three years would be greatly increased (British Parliamentary Papers, Cmd. 5374). This announcement of a greatly enhanced military buildup was accompanied by the Defense Loans Act, which gave specific approval for £400 million of this to be financed by borrowing. Finally, in the run up to and immediate aftermath of the Budget of 1938, it became clear that spending would be ahead even of the 1937 program at least during the two years 1938/39 and 1939/40.

The first period that deserves some comment covers 1920 to 1922. This includes the episode of the Geddes Committee, appointed in August 1921 to advise the government on spending cuts, which published its report in February 1922. The prelude to this committee was a big Supplementary Estimate in December 1920 and then a full year Defense Estimate of £138.7 million in April 1921, which showed that the “normal year” of £93.5 million would not be achieved in 1921/22. The Geddes Committee was established with a mandate to propose budget cuts of £100 million at current prices. At the time, The Economist predicted that defense would bear the brunt and that the government would shrink from implementing the full amount of the cuts. It turned out that the Geddes proposal of a defense cut of £58.5 million was scaled back to £38 million.

These developments are reflected in the “Defense News” estimates reported in Table 2. We assume that the extra defense spending announced in early December 1920 is new information for the public in 1920 quarter 4. The Christmas Day issue of The Economist in 1920 forecast very accurately the Defense Estimate for 1921/22 so we assume that, in effect, this became known to the public for the first time at the start of 1921 quarter 1. At the same time, given the weak state of public finances with the large public debt overhang from World War I, it was clear that the Treasury could not afford to allow this extra military spending to become permanent, so we assume that the “normal year” was regarded by the public as postponed but not indefinitely, and we take the subsequent establishment of the Geddes Committee as support for this view. Given that the outcome of the Geddes Committee deliberations was basically predicted by The Economist, we assume that well-informed agents would have received this news at its inception in 1921 quarter 3 rather than 1922 quarter 1.Footnote ¹¹ In making these estimates of “Defense News,” we have, of course, made assumptions but have primarily relied on changes in the information available to the public as reflected in spending announcements and commentary on them by The Economist.

The second period to be reviewed covers the move to rearmament during the years 1935 to 1937. In the last year before the new policy of rearmament, 1934/35, defense expenditure was £124.7 million. Following the publication of the 1935 White Paper, the April 1935 Budget confirmed that defense spending in 1935/36 was expected to be £136 million. However, The Economist of October 5, 1935 reported that spending in quarter 3 had exceeded that of the same period in the previous year by 31 percent and the outturn for 1935/36 was eventually £148.7 million. During February and March 1936, The Economist ran a series of editorials and special reports which suggested that expectations of future defense spending should be greatly increased. The Defense Estimate for 1936/37 was £167.7 million but the 1936 Defense White Paper, published on March 3, indicated that the government had plans of unspecified cost for major re-equipment of all three armed forces which The Economist (March 7 and 14) suggested would imply additional defense expenditure of at least £25 million, and possibly £40 million or even £80 million, during 1936/37. On November 28, 1936 The Economist reported on a brokers' circular which predicted that the Defense Estimate for 1937/38 would be £220 million and described this as “not prima facie unreasonable.” On February 20, 1937 The Economist noted that the Defense White Paper expected defense spending of at least £1500 million over five years and on April 24, 1937 reported that defense spending in 1937/38 was expected to be £278.3 million, of which £80 million would be met by borrowing.

These developments inform the large values for “Defense News” for 1935_Q2, 1936_Q1, and 1937_Q1 in Table 2. We assume that by 1935 quarter 2 the public expected a substantial and sustained increase in defense expenditure. Given that The Economist of March 16, 1935 commented that the White Paper implied that in future spending would show a big increase over the estimate for 1935/36, we assume that at this point the actual 1935/36 spending was anticipated and expected to be the normal level in future. We believe that the public's expectations of future defense expenditure were raised again in the light of the next White Paper and Defense Estimates in March 1936. The Defense Estimate for 1936/37 published in March 1936 was £168 million (compared with £136 million a year earlier) but a big supplementary estimate was generally expected, which would push expected spending to well over £200 million according to The Economist. The press was, in effect, largely correctly anticipating what we now know the Cabinet had agreed, namely, a total program of £1,075 million over the period 1936–1940 (Peden Reference Peden1979, p. 77). We assume that at this point the public expected defense expenditure of £212 million per year for the next five years. We also assume that the public's expectations were ratcheted up again in early 1937. Specific commitments were made in the Defense White Paper and the Defense Loans Act underpinned the credibility of this announcement. We assume that in 1937 quarter 1, the public expected future expenditure to be £300 million per year through 1942 to match the Defense White Paper plans which The Economist of March 20 thought would be carried out.

In Table 2, the observations for “Defense News” in 1935 through 1937 are abnormally large. We have constructed them on the basis of a careful and thorough reading of The Economist.Footnote ¹² Nevertheless, there must have been considerable uncertainty at the time both about the detail of the government's plans and the extent to which they would be implemented. This could lead to downward-biased estimates of the normal government expenditure multiplier either because the “Defense News” numbers are outliers or if they are unreliable. Accordingly, in what follows we present results both for a full interwar sample period and also for a period truncated at 1934_Q4 which omits the potentially problematic years at the end.

ECONOMETRIC MODELING

The defense news variable that we use in our econometric modeling, news, is, following Ramey, the series given in Table 2 divided by the one-quarter lagged value of real GDP; news is shown in Figure 1 (Ramey Reference Ramey2011b). We begin by employing an approach similar to that of Barro and Redlick (Reference Barro and Redlick2011).Footnote ¹³ This has the following general specification

(1)

D_t is the level of defense spending while \nabla y_t = \log \left({{{GDP_t} \mathord{\left/ {\vphantom {{GDP_t} {GDP_{t \minus 1}}}} \right. \kern-\nulldelimiterspace} {GDP_{t \minus 1}}}} \right)\] is quarterly real GDP growth. The sample period runs from the first quarter of 1922 to the end of 1938. Starting in 1922 avoids the immediate aftermath of World War 1, which is known to have produced a shift in the process generating GDP.Footnote ¹⁴ The lag length was set at 4 to model any seasonality present in the data (seasonally unadjusted data was used throughout). The contemporaneous term of news was included but all other variables were lagged to avoid problems of endogeneity. The control variables included were lags of export growth, changes in the money multiplier, consol yields and the tax rate, and the unemployment rate. Growth rates and changes were used to ameliorate problems caused by the non-stationarity of many of the variables when expressed as levels. The error term u_t is specified as the ARCH(1) process u_t^2 = \delta _0 + \delta _1 u_{t \minus 1}^2 \, which effectively models the volatility of GDP growth during 1926 and 1927 and precludes the need for lagged values of GDP growth to be included as regressors: including such lags with u_t assumed to be homoskedastic leads to a significant deterioration of fit. Figure 2 clearly shows the volatility of GDP growth in the boom and bust after World War I and during 1926 and 1927.Footnote ¹⁵

FIGURE 1

DEFENSE NEWS DIVIDED BY LAGGED REAL GDP: 1920_Q2–1938_Q$

Sources: Derived using estimates in Table 2 and GDP data from Mitchell, Solomou, and Weale (2012).

FIGURE 2

QUARTERLY GROWTH OF REAL GDP: 1920_Q2–1938_Q$

Sources: Derived from Mitchell, Solomou, and Weale (2012).

In each case, we report estimates based on a final specification, in which insignificant variables have been deleted (see Table 3). For the estimated coefficients of the included control variables, see Appendix Table 1, column 1. Throughout we use standard errors robust to possible residual autocorrelation and heteroskedasticity, all included variables are significant and the reported equation passes a variety of standard tests for misspecification. The news variable is significantly positive at lags of one and two quarters with coefficients estimated to be, with one-standard error bounds, 0.0325 ± 0.0025 and 0.0179 ± 0.0039.

Barro and Redlick's multiplier definition is based on using annual data with single lags of the regressors. Our quarterly counterpart m = 16\sum\nolimits_{i = 1}^4 {\beta _{1i}} + \sum\nolimits_{i = 1}^4 {\beta _{2i}}. Given the estimates in Table 2, the multiplier is m = 0.45 ± 0.13. For the period truncated at 1934Q₄, the estimated multiplier is m = 0.81 ± 0.37.

Since the fit of this equation is not particularly good, with an R ² of just 0.19, we tried moving to a more general model specification. To this end we relate changes in output growth, ∇ y _t, again to lags of news _t but now to lags of government spending growth disaggregated into defense and nondefense spending, ∇ d_t and ∇non − d_t, and lags of nongovernment spending (private expenditure on consumption, investment and net exports) growth, ∇ n_t, all growth variables again being defined as one-quarter changes in the logarithms

(2)

TABLE 3

BARRO AND REDLICK SPECIFICATION ESTIMATES

Note:

Figures in parentheses are robust t-ratios. SE is the regression standard error. The estimates for the ARCH(1) error specification for the equation in column 1 are \hat \delta _0 = 9.31 \times 10^{ \minus 9} [297.5] and \hat \delta_1 = 1.7016\;[4.2] and are \hat \delta _0 = 5.98 \times 10^{ \minus 9} [241.0 and \hat \delta_1 = 2.0891\;[3.7] for the equation in column 2.

Tables 4A and 4B report our estimates based on a final specification, in which insignificant variables have been deleted. For the estimated coefficients of the included control variables, see Appendix Table 1, column 2.Footnote ¹⁶ All included variables are highly significant and the reported equation passes a variety of standard tests for misspecification. The news variable is again significantly positive at lags of one and two quarters with coefficients estimated to be 0.0154 ± 0.0044 and 0.0380 ± 0.0031. The sum annualized of these coefficients (0.213 ± 0.025), may be regarded as an estimate of the “direct” multiplier of defense news on GDP.Footnote ¹⁷ The second model offers a superior fit to the Barro and Redlick model, with a much improved R^2 of 0.46 and a regression standard error some 20 percent lower.

However, there may also be “indirect” effects because news may have influence on other various categories of spending that also matter to GDP growth. Regressions of the form \nabla z_t = \gamma _0 + \sum\nolimits_{i = 1}^4 {\gamma _i \nabla z_{t \minus i} + \sum\nolimits_{i = 1}^4 {\theta _i news_{t \minus i} + v_t}} were therefore estimated for z respectively defined as ∇ d, ∇ non − d, and ∇ n, these being reported in columns 2–4 of Tables 4A and 4B. These were then inserted into equation 2 to obtain a set of indirect multipliers and an overall multiplier, defined as the sum of the direct and indirect multipliers: the time path of the direct and the total overall multipliers are shown in Figure 3. The total overall multiplier is then 0.34 on an annualized basis. It is effectively reached after three years.Footnote ¹⁸ For the truncated period ending at 1934Q₄, we obtain an annualized total multiplier of 0.76.

Two additional points are worth noting. First, the equation in column 2 shows that defense news does indeed predict subsequent defense expenditure. Second, it should be remembered that ending the sample period at 1934Q₄ excludes the defense news data which are subject to relatively large margins of error. This implies that finding a multiplier well below 1 is robust to excluding the largest and least certain values for defense news.

The clear implication is that there was some crowding out of private expenditure even in a “depressed economy.” This issue can be explored using an approach recently proposed by Ramey (Reference Ramey2012). In the current context, Ramey's equation may be specified as

(3)

TABLE 4A

REGRESSION ESTIMATES: PREFERRED SPECIFICATION (sample period: 1922_Q1 to 1938_Q4)

Notes:

Figures in parentheses are robust t-ratios. SE is the regression standard error. The estimates for the ARCH(1) error specification for the equation in column 1 are \hat \delta _0 = 2.37 \times 10^{ \minus 7} [322.0] and \hat \delta_1 = 1.5905\,[5.6].

where N_t and G_t are the levels of nongovernment (private) and government expenditure respectively. This equation was estimated for both sample periods and the results are reported in Table 5. Government expenditure has an overall negative effect on private expenditure, with the total impact estimated to be −0.23 for the full sample and −0.28 for the shorter sample.Footnote ¹⁹ This confirms that the multiplier, ΔY/ΔG, is < 1 with implied values of 0.77 for the full sample and 0.72 for the shorter sample.

TABLE 4B

REGRESSION ESTIMATES: PREFERRED SPECIFICATION (sample period: 1922_Q1 to 1934_Q4)

Note:

Figures in parentheses are robust t-ratios. SE is the regression standard error. The estimates for the ARCH(1) error specification for the equation in column 1 are \hat \delta _0 = 1.61 \times 10^{ \minus 8} [244.9] and \hat \delta_1 = 1.9243\,[2.7].

THE IMPACT OF REARMAMENT ON GDP

We can use our estimates of the multiplier to examine the impact of rearmament on economic recovery in Britain in the 1930s and compare our findings with those of the best-known paper on this topic by Thomas (Reference Thomas1983). The simplest approach is to take the estimates of the government-expenditure multiplier obtained for the period 1922–1934 and apply them to the increase in defense expenditure of £204 million in 1938. Our multiplier estimate of 0.8 gives a rearmament contribution of about £160 million or 3.3 percent of GDP in 1938. This is half the estimate based on Thomas's method with a multiplier calculated as 1.6 (Ibid.).

FIGURE 3

DEFENSE NEWS MULTIPLIERS

Sources: Authors' own calculations, see the text.

The idea of “defense news” is that it contains information relevant to private sector decisions. A credible commitment to future defense spending might well stimulate investment by businesses to meet the anticipated demand for military equipment, munitions, etc. Such investment would amount to an element of “crowding in,” and is an example of what the economics literature on the fiscal multiplier would call an anticipation effect.Footnote ²⁰

Some years ago, A. J. Robertson argued that the main initial effect of rearmament was to encourage growth in the construction industry as armaments manufacturers scrambled to add capacity (Robertson Reference Robertson1983). Robertson drew attention to the impact of works and buildings expenditures in, stimulating economic activity and which sustained a vigorous expansion in construction output even as private house-building stalled. Robertson based this argument on the contribution of William Hornby to the official U.K. History of the Second World War (Hornby Reference Hornby1958). Hornby details the buildup of activity to add capacity especially from 1936 onwards. The push included a notable surge in investment for aircraft production and the formation of three important new armaments manufacturing companies. David Chambers's recent compilation of data on all interwar IPOs shows that 20 of the 27 aircraft manufacturing ventures were launched between 1934Q₃ and 1936Q₂, when the increase in actual defense spending was still relatively small.Footnote ²¹

TABLE 5

ESTIMATING THE EFFECT OF GOVERNMENT EXPENDITURE ON PRIVATE-SECTOR SPENDING

Note:

Figures in parentheses are robust t-ratios. SE is the regression standard error.

Robertson's view suggests we can use our defense news methodology, which was really designed to facilitate an estimate of the multiplier, to simulate the impact of rearmament on GDP. In principle, the overall multiplier time path from our preferred specification could be used to compute the estimated monetary impact on GDP of defense-news shocks as in Figure 4, which uses a five-year horizon on the basis of a multiplier of 0.34.Footnote ²²Figure 4 clearly shows the rapid increase in this impact after 1934, driven by the very large defense-news shocks starting in 1935. Table 6 sets out in more detail the simulated impact of defense news on GDP during the recovery period and also reports actual defense expenditure. Based on the total multiplier (including the direct and indirect components) for the full period, this impact averaged 5.1 percent of GDP in 1938 and amounted to £246.2 million (at 1938 prices) for the four quarters.

FIGURE 4

SIMULATED EFFECT OF DEFENSE NEWS SHOCKS ON GDP USING A FIVE-YEAR HORIZON

Sources: Authors' own calculations, see the text.

One must be careful not to put too much weight on these estimates because the defense-news shocks in these years were very large. For example, the defense-news shock in 1937Q₁ is estimated as £393 million on the basis of the Defense White Paper stating that total defense expenditure over the next five years would be at least £1,500 million. The 1937 announcement was huge compared with the defense news from 1922 to 1934 and compares with actual defense spending in 1936/37 of about half that amount (£194.6 million). Clearly, an impact of 5.1 percent of GDP would entail a very substantial boost to investment from the anticipation effect and is perhaps best regarded as an upper-bound estimate since the increase in investment in manufacturing sectors relevant to the war effort was £57.9 million (at 1938 prices), only £16.5 million more than in 1934.Footnote ²³

TABLE 6

SIMULATED IMPACT OF DEFENSE NEWS SHOCKS ON REAL GDP (£ million, 1938 prices)

Sources:

Authors' own calculations. See the text and data appendix for GDP and defense expenditure.

Overall rearmament provided a significant stimulus to recovery in the late 1930s. That said, the evidence points to a smaller impact than is found (or implied) by the earlier literature on fiscal multipliers in 1930s' Britain. All this suggests considerable caution before accepting without qualification Thomas's often cited conclusion that “the success of rearmament in creating employment … leads us to view the eschewment of fiscal policy in the thirties as a missed opportunity” (Thomas Reference Thomas1983, p. 571). Our results suggest that, if it is believed that rearmament had a substantial impact, this would be on the basis of the implications of massive future spending plans rather than because there was a large fiscal multiplier. The implication is that a conventional temporary program of expenditure on public works of similar magnitude to the actual increase in defense spending between 1934 and 1938 would have had a much smaller impact on GDP than Thomas's model seems to imply.