Deflation and Public Finances
Evidence from the Historical Records1
Author: Mr. Nicolas End, Mr. Sampawende J Tapsoba, Mr. G. Terrier, and Renaud Duplay

Contributor Notes

Authors’ E-Mail Addresses: NEnd@imf.org; STapsoba@imf.org; GTerrier@imf.org; RDuplay@imf.org
Publication Date:
28 Jul 2015
eISBN:
9781513528243
Language:
English
Keywords:
WP; inflation rate; fiscal policy; interest rate; baseline scenario inflation; deflation subsidy; impact of deflation; dummy t-3; deflation situation
Download PDF (6.4 MB)

This paper examines the impact of deflation on fiscal aggregates. With deflation relatively rare in modern history, it relies mostly on the historical records, using a dataset panel covering 150 years and 21 advanced economies. Empirical evidence shows that deflation affects public finances mostly through increases in public debt ratios, reflecting a worsening in interest rate–growth differentials. On average, a mild rate of deflation increases public debt ratios by almost 2 percent of GDP a year, this impact being larger during recessionary deflations. Using a simulation model that accounts for composition effects and price expectations, we also find that, for European countries, a 2 percentage point deflationary shock in both 2015 and 2016 would lead to a deterioration in the primary balance of as much as 1 percent of GDP by 2019.

Abstract

This paper examines the impact of deflation on fiscal aggregates. With deflation relatively rare in modern history, it relies mostly on the historical records, using a dataset panel covering 150 years and 21 advanced economies. Empirical evidence shows that deflation affects public finances mostly through increases in public debt ratios, reflecting a worsening in interest rate–growth differentials. On average, a mild rate of deflation increases public debt ratios by almost 2 percent of GDP a year, this impact being larger during recessionary deflations. Using a simulation model that accounts for composition effects and price expectations, we also find that, for European countries, a 2 percentage point deflationary shock in both 2015 and 2016 would lead to a deterioration in the primary balance of as much as 1 percent of GDP by 2019.

I. Introduction

With falling oil and commodity prices, inflation has been declining in advanced economies and is running significantly below targets, raising concerns over the risk of deflation. Fear of deflation is generally premised on the belief that it is associated with recession (Stern, 2003), reflecting developments during the 1930s, when the combination of deflation and economic contraction triggered debt deflation.2 This experience has helped shape the belief that deflation is deeply perilous and should be avoided: falling prices would harm economies through rigid interest rates and price-setting mechanisms and a spiral of expectations and deflation (DeLong, 1999; Bernanke, 2002; Furhrer and Tootell, 2003; Svensson, 2003; and Beckworth, 2008).

Understanding the consequences of deflation on fiscal aggregates is a key question for policymakers. The focus of this paper is to assess how deflation may buffet already-strained public finances and further complicate fiscal policy. Specifically, what are the effects of declining prices on fiscal aggregates? Deflation is clearly associated with mechanical increases in debt-to-GDP ratios: debt mostly consists of preexisting stocks, and its term structure, together with downward rigidities in sovereign interest rates, tends to further compound this increase. At the same time, a number of nominal flow variables—such as revenue and nominal GDP but also, to a smaller extent, expenditure—tend to decrease mechanically. This study investigates the impact of falling prices on public finances using historical records. In particular, it explores whether the impact of deflation on fiscal aggregates is asymmetrical from that of inflation. It uses an original panel dataset covering a long timeframe (over 150 years) and data for inflation, growth, and fiscal aggregates—including debt-to-GDP ratios for 21 advanced economies.3

What does available evidence tell us? In the literature, much of the attention has focused on either the effect of fiscal stances on price dynamic (Catao and Terrones 2005) or the fiscal consequences of very high inflation (Oliveira, 1967 and Tanzi, 1977, Aghevli and Kahn, 1978 and Heller, 1980).4 Existing research on the effects of deflation on fiscal aggregates is scant, compared to the abundant literature on the corresponding effects of high inflation. In addition, the existing literature on deflation has focused mostly on the role of fiscal policy to inflate aggregate demand with a view to exiting from deflation (Auerbach and Obstfeld, 2004; Cochrane, 2011). Another strand of the literature investigates the link between deflation and recession.5 Studies based on records of deflation over the past two centuries argue that, with the exception of the Great Depression, deflation was not systematically associated with persistent and deep economic recession but was often driven by increases in aggregate supply. Borio and Filardo (2004, 2005) proposes three broad categories of deflation: good deflations, which arise from positive supply shocks; bad deflations, which are associated with recessions and include the experience of Japan in the 1990s; and ugly deflations, which represent periods of steep declines in prices associated with severe recessions, such as the early 1930s Great Depression. Both bad and ugly deflations stem from a collapse in aggregate demand.6

Our findings center around three key messages. First, while deflation negatively affects debt-to-GDP ratios, it also impacts nominal budgetary variables, that is, government revenue and expenditure. Piloting fiscal policy in the midst of deflation might be challenging, similar to the context created by negative output growth. This points to asymmetric economic behaviors and responses, which was noted in Fisher (1928)’s money illusion theory.7 Yet, deflation signals more a strain to achieving fiscal targets than a terrible accident, and what ultimately matters is the authorities’ capacity to steer expenditure skillfully in the context of declining revenues. Second, we find that linkages have evolved over time. Public finance management has changed from a fiscal policy that was relatively marginal, nominally-driven and cash constrained, to modern governments operating on large aggregates and with indexation mechanisms; deflation episodes have been relatively rare in the recent period, with the notable exception of Japan in the 1990s.8 Finally, we find that not all deflations are alike and, in particular, the impact is very different between deflations associated with positive growth and deflations accompanied with recessions.

Our findings agree with the recent literature on the macroeconomic consequences of deflation, which highlights that deflations are not necessarily detrimental to public finances. Our findings are also supported by the deflationnary experience of Japan. Overall, deflation in Japan was associated with deteriorating fiscal aggregates, mostly the debt-to-GDP ratio. Public debt doubled, due primarily to a snowball effect (i.e., an unfavorable differential between interest rate and growth); the denominator effect of deflation explains roughly one fifth of the debt increase. The impact of deflation on the primary deficit was blurred by the combination of demographic changes and policy responses (Appendix A). Expenditure-to-GDP ratios increased largely because of rising age-related spending and explicit downward rigidities.9 The impact on the tax-to-GDP ratio is more difficult to identify, as new tax measures were introduced simultaneously with a shift in the tax base. In addition, as deflation was protracted and anticipated, its effect could be partially offset in the annual budgets.

Yet, some caveats refrain from translating these results directly to the current situation in advanced economies. First, historical data relate only to ex-post outturns and incorporate discretionary measures adopted in the face of deflation, including new tax measures aimed at boosting revenue. As a result, underlying trends are difficult to identify. Second, to a large extent, these data do not account for structural changes introduced in policymaking in recent decades, given that deflation has been relatively rare during that period. A prospective simulation exercise, built to include expectations and modern features of today’s governments under a no-policy change assumption, confirms nevertheless that piloting fiscal policy in the midst of deflation might be a complicated task. In particular, consideration should be given to composition effects within the fiscal balance and the role of expectations.

In addition, the present paper strictly follows a positive approach and does not analyze the optimal response of fiscal policy makers to deflation.10 It is organized as follows. The next section discusses how deflation could theoretically affect fiscal aggregates. It gives the general intuition that deflation should matter for fiscal policymakers. Section III describes our empirical methodology and the historical dataset that we built in order to encompass a sufficient number of deflation episodes. Section IV reports the results of the analysis, which Section V attempts to put in a narrative, historical perspective. Section V closes by presenting a deficit-debt simulation exercise for the Euro Area that would capture modern features of public finances. Section VI draws some tentative policy implications from our main results.

II. Theoretical Background

Public finances are vulnerable to deflation on several accounts. The effect of deflation on debt ratios was described as early as in the 1930s by Fisher (1933). In addition, deflation can affect primary balances through its impact on revenue and expenditure.

A. Primary Balance

During deflation periods, the primary balance is affected according to the magnitude and speed of the respective paces of adjustment of revenue and expenditure.

Revenue

The net impact of deflation on public revenue is affected by a variety of factors. An immediate impact is the loss of seigniorage revenue, which represents the real revenues governments acquire by using newly issued money to buy goods and non-money assets. Under a fiat money system and without any monetary policy action, seigniorage revenue is equivalent to an inflation tax, given by the product of the inflation rate and real money balances. In principle, deflation reduces seigniorage for a given level of real money balances—thus generating a “deflation subsidy”. However, if deflation leads to an increase in holdings of real money balances, the effective tax base will augment, leading to a possible increase in seigniorage revenue. Recent evidence suggests that potential gains of seigniorage are limited in today’s advanced economies, as base money is small relatively to GDP.11

Under a fully proportional tax system, deflation would have no impact on the revenue-to-GDP ratio: every component of GDP would be similarly taxed, and both nominal revenue and GDP would react in similar proportions, leaving the ratio unchanged.12 However, in real life, as tax systems always include distortionary features, there are reasons to think that revenue ratios will be affected by deflation.

Some factors tend to decrease revenue-to-GDP ratios during deflation times. First, the progressivity of the tax system matters. In a progressive system, when tax brackets are not perfectly indexed to inflation, deflation will tend to curb revenue ratios by moving some taxpayers to lower tax brackets, leading to lower revenue collections; and vice versa, in a regressive system (Hirao and Aguirre, 1970). Second, revenue-to-GDP ratios will tend to suffer from deflation if tax exemptions are widespread. Such exemptions are often set in nominal terms, and their ensuing costs increase when prices fall.

Some other factors tend to push revenue ratios up during deflation times. First, some revenue components, such as excises and non-tax revenues, are by nature more price-inelastic than income taxes. Their inertia in the face of deflation thus tends to boost revenue as a percent of GDP.13 Second, deflation can have effects on tax bases through behavioral effects. During deflation, consumption tends to shift toward higher-scale goods, in part because prices adjust more quickly than incomes. Since these goods tend to be more heavily taxed, this will help improve revenue.14 Similarly, if consumption prices drop more rapidly than the GDP deflator, then the revenue ratio will tend to rise.15 Finally, central bank actions against protracted deflationary pressures, in the form of quantitative easing policies, can generate seigniorage revenue.

The above effects could be mitigated if tax brackets are indexed. Under a full indexation mechanism, during deflation tax provisions are automatically revised downwards—including minimum income tax thresholds and tax brackets—to maintain the tax pressure constant.16 The net effect of these indexation mechanisms will be similar to that described under a non-distortionary system, with revenue-to-GDP ratios immune from deflations.

Expenditure

In general, public expenditure tends to be more sensitive to deflation than revenue collections because of nominal rigidities in the design of some of its components. It may be politically difficult to reduce wages and social transfers when prices are falling. As a result, during deflation periods, freezing nominal spending may be the only feasible option, leading to increases in expenditure-to-GDP ratios. This is particularly true for social transfers to households (pensions and other benefits) and wages.17

Contractual arrangements, such as multiyear agreements and lagged price indexation provisions, may also delay the transmission of deflation to capital spending (Aghevli and Khan, 1978; Heller, 1980).18 This feature can also apply to some recurrent expenses, if they are specified in multi-annual contracts (maintenance, IT outsourcing, etc.). In those cases, price adjustments will be limited to new and renewed contracts, and will take some time to be fully reflected in fiscal aggregates.

Finally, the design of budgets or fiscal rules could delay the response to an unexpected shock of deflation. Because budgets are usually prepared and executed in nominal terms, it may be difficult to adjust spending lines to unexpected deviations from the budget forecasts within a given fiscal year.

B. Debt

Deflation has a negative impact on debt ratios if not fully anticipated in the level of nominal interest rates. This effect operates through the initial debt stock and the combined effect of the real interest rate and the primary balance. First, for any given debt stock and real growth rates, deflation mechanically increases the debt-to-GDP ratio: It lowers nominal GDP, pushing the ratio up. Second, as discussed in the previous section, the primary balance may deteriorate unexpectedly in a deflationary environment, leading to a further increase in the debt burden. Third, for any given nominal interest and real growth rates, deflation raises the real value of the interest bill. If interest rates are sticky or deflation is not anticipated, nominal rates will not immediately adjust to absorb the shock.19 In general, interest payments are largely based on contractual interest rates, which are mostly fixed and do not adjust to domestic prices in the short run. The impact of this channel depends on the maturity structure and the currency denomination of the sovereign debt, as well as the share of price-indexed bonds in the total debt (Akitoby and others, 2014).

These mechanisms can be summarized within the debt dynamic equation, which links year-on-year changes in the debt-to-GDP ratio to the existing debt stock through the impact of nominal interest rates, inflation, and output growth to the primary balance, and any stock-flow adjustments.20

Δdt=it−πt−gt(1+gt)(1+πt)dt−1−pt+SFt(1)

Equation (1) summarizes the direct channel through which inflation affects the debt-to-GDP ratio: dt is the stock of government debt as a percent of GDP in year t; pt represents the primary balance as a percent of GDP; it, gt, and πt denote the nominal interest rate, the real GDP growth rate, and the inflation rate in year t; and SFt designates stock-flow adjustments (as a percent of GDP); Δ is the first-difference operator. In equation (1), the term (it – Δt – gt) can have a significant impact on debt dynamics. If it is greater than zero, it means that the debt stock increases over time, even in a situation of primary surplus.

III. Empirical Strategy

A. Methodology

To investigate how deflation affects public finances, we assess first the effects on the changes in debt, primary balance, revenue, and expenditure, expressed as ratios to GDP. As argued above, these variables depend upon inflation and growth rates, two key inputs of budget formulation and execution. Accounting for potential persistence effects, we use the following autoregressive model:

Δxi,t=α0+Σj=0Tα1,jxπi,t−j+Σj=0Tα2,jxgi,t−j+Σk=1Kα3,kxTk+α4xΔxi,t−1+ɛi,tx(2)

x stands for either debt, the primary deficit, primary expenditure, or revenues, in percent of GDP. The use of first-differences accounts for possible non-stationarity of the debt series.21 The change in each fiscal ratio is assumed a function of lags and current and past values of inflation and output growth; ɛi,tx captures classic error terms and shocks unrelated to inflation and growth; and α1,0 is the main parameter of interest to be estimated, as it captures the different channels through which inflation impacts the fiscal variables. For instance, with respect to equation (2), current-year higher inflation and growth are expected to reduce the debt stock (i.e., α1,0debt<0,α2,0debt<0). The Tk are dummy variables that capture historical breaks in the estimated parameters. As we are using a long timeframe (1851–2013), we include all estimates dummy variables to account for strucural breaks corresponding to the pre-Great Depression (1851–1928); the Great Depression (1929–34); and the post-WWII (1946–2013) periods.22 In some specifications for debt, we also control for the primary balance and the effective interest rate.

In this paper, we examine whether the regime of inflation—high, low, or negative inflation—matters for fiscal ratios. Hence, we explore whether the impact of deflation is different from that of inflation.23 To that end, we break down the inflation term into positive and negative inflations and estimate the following equation:

Δxi,t=β0+Σj=0Tβ1,jx+(π>0)i,t−j+Σj=0Tβ1,jx−(π≤0)i,t−j+Σj=0Tβ2,jxgi,t−j+Σk=1Kβ3,kxTk+β4xΔxi,t−1+vi,tx(3)

In order to test asymmetry in the effect of deflation, we also scrutinize the effect of a set of dummy variables Ut capturing different regimes of inflation and growth.

Δxi,t=γ0+Σj=0Tγ1,jxUi,t−j+Σj=0Tγ2,jxgi,t−j+Σk=1Kγ3,kxTk+γ4xΔxi,t−1+ηi,tx(4)

Admittedly, estimating equations (2)–(4) poses several challenges. The first, main challenge is the potential reverse causality between fiscal policy and inflation, as fiscal deficits play a role in the formation of the price level. Instrumental variable methods are the typical, technical solution for correcting such a bias. However, good instruments—i.e., variables excluded from the main regressions that would be highly correlated with inflation but uncorrelated with the error terms—are difficult to find, especially over a timeframe as long as the one we use in this paper.24 We explored unsuccessfully a few instrumental variables, including lagged inflation and an index of imported inflation.25 An alternative approach is the generalized method of moments (GMM), which takes advantage of the lagged structure of the data. However, this technique is not well-suited for our focus on historical data, as it produces biased estimates when the time dimension is larger than the cross-sectional one. For this reason, the GMM approach was ruled out. Yet, we mitigate the simultaneity bias by including possible lagged effects of inflation in equations (2)–(4).

Another key challenge includes the possibility that some omitted variable is correlated with fiscal performance and inflation. Possible candidates are war, colonization, and political instability periods. To mitigate this problem, we control for idiosyncratic and common factors by allowing for both country and year fixed effects. We also have dummy variables to account for strucural breaks corresponding to the pre-Great Depression (1851–1928), the Great Depression (1929–34), and the post-WWII (1946–2013) periods. An additional challenge is a potential multi-colinearity between inflation and other control variables. So far, we have kept the description of the framework simple and have abstracted from second order effects, such as the cross-linkages between inflation, interest rates and growth. In reality, the phenomena behind equations (2)–(4) are complex and their parameters are intertwined (Barro, 1995). For instance, a fiscal adjustment leading to a larger primary balance could affect growth and the nominal interest rate (Clinton and others, 2011).

We test the optimal lag length. For each country, we use the well-known Akaike, Hannan-Quinn, and Schwarz information criteria. In Appendix C, we find that the optimal lag length is between 1 and 3. To account for the entire autocorrelation structure, all estimates are conducted with 3 lags whenever possible. Finally, to produce robust results, autocorrelation and heteroskecasiticty are corrected following the usual practice.

B. Dataset

Episodes of deflation were much more commonplace in the 19th and early 20th centuries than in modern times (since the end of World War II). In moderne times, deflation has been rare (5 percent of observations). Therefore, we rely mostly on the historical records.

The first, main variable is inflation. We use the CPI-based inflation taken form Bordo and Filardo (2005) through 1997, complemented by data from the IMF’s World Economic Outlook (WEO). When observations are missing in Bordo and Filardo’s dataset, we use Warren Weber’s dataset, which contains information on prices and output from 1810–1995.26 Second, economic growth is used as control variable. In order to charaterize the role of growth regimes, we similarly compile the information on real GDP growth rates from Bordo and Filardo (2005), complemented by recent WEO figures. Third, debt ratios are taken from Abbas and others (2010), supplemented with recent WEO data. The dataset compiles government gross debt-to-GDP ratios covering a large set of countries, spanning a long time period. Finally, the other fiscal variables are culled from Mauro and others (2013). Their dataset contains fiscal flows, such as primary balances and their main components; it also includes interest payments, which allows us to compute effective interest rates.

Data limitations are a serious issue for any historical approach, especially when they cover the pre-WWI period. In order to reduce biases related to the presence of outliers, we exclude improbable events or extreme outliers in price dynamics, such as hyperinflation episodes (inflation above 100 percent). The dataset is an unbalanced panel of data for 21 developed economies dating from as early as 1851.27 We define deflation as negative inflation, low inflation as an inflation rate between 0 and 2 percent, and recession as a negative annual growth rate in real GDP.

Descriptive statistics show that inflation is more dispersed than growth: the average inflation rate is 4 percent, with a standard deviation of 8.1 percent; and the average growth rate is 3 percent, with a standard deviation of 4.1 percent (Table 1). The average of the public debt-to-GDP ratios is 53 percent and the primary balance is estimated to show a surplus of 0.5 percent of GDP on average, with both revenue and primary expenditure estimated at around 23 percent of GDP. About one fifth of the total historical sample corresponds to deflation, but this ratio shrinks significantly over the most recent period. Public debt-to-GDP ratios seem the most vulnerable to deflation. They grow on average by 1.7 percentage points of GDP a year during deflation periods, thrice faster than the sample average (Figure 1). The deterioration is worse when deflation is combined with negative growth (depression). On average, the primary balance tends to worsen, but at a moderate pace during deflation (mostly when combined with recession) on the back of stalling revenue-to-GDP and increasing expenditure-to-GDP ratios.

Table 1.

Descriptive Statistics

(Percent)

article image
Figure 1.

Change in Fiscal Aggregates and Inflation Regimes

Citation: IMF Working Papers 2015, 176; 10.5089/9781513528243.001.A001

Source: Authors

IV. Evidence from the Historical Records

In this section, we discuss key findings from the historical database. To that effect, we present the main findings; test for asymmetries between positive and negative inflations; and assess possible differences across growth regimes.

A. Effects of Deflation on Fiscal Ratios

Overall, evidence from the historical records suggests that deflation increases the debt-to-GDP ratio primarily through a worsening of the interest rate-growth differentials. Primary balances are found to remain broadly unaffected.

Debt

We examine first the impact of inflation on debt accumulation. To that effect, we estimate equation (2) and report the results in Table 2, column (1). We focus mostly on the contemporaneous effect of inflation on the debt-to-GDP ratio.28 The empirical examination identifies a strong and immediate impact of price dynamics on debt-to-GDP ratios. The coefficients are significant at the 5 percent level and are in line with expectations in the debt accumulation equation. All other things equal, higher inflation helps curb the debt-to-GDP ratios. A 1 percentage point increase in the rate of inflation reduces the debt-to-GDP ratio by 0.15 percentage point. There is a delayed positive effect of inflation on the debt ratios after two years. The change in debt-to-GDP ratio is persistent. The dependence coefficient is positive, below unity in absolute terms.

Table 2.

Inflation, Deflation, and Debt 1851–2013

article image
Note: Dependent variable: First difference of fiscal aggregate in percent of GDP. Robust standard errors are in parentheses. Regressions include intercepts and time fixed effects.*** p<0.01, ** p<0.05, * p<0.1

Other controls variables have the expected signs. Higher economic growth also reduces the debt build-up, and at a faster pace (a 1 percentage point increase leads to a decline of 0.40 percentage point in the debt-to-GDP ratio). The impact in the following year is positive, estimated at 0.14 percentage point. A higher nominal interest rate is also associated with a faster debt build-up: a 100 basis points increase in the nominal rate raises the debt-to-GDP ratio by 0.65 percentage point. The delayed impact is negative, though with a lower magnitude.

To assess the particular effect of deflation, we split inflation into deflation and positive inflation. We estimate equation (3). In Table 2, column (2), the estimated coefficients are statistically significant. When inflation is positive, a 1 percentage point increase in its rate translates into a reduction of the debt ratio of 0.15 percent of GDP in a year. When inflation is negative, it leads to an increase of a similar magnitude in the public debt ratio (0.19 percentage point).

Given that our dataset spans a long timeframe, we verified the stability of our results on deflation over three separate periods: before the Great Depression (1851–1928); during the Depression (1929–34); and after World War II. Results are presented in Table 2, columns (4–6). They suggest that deflation was associated with a worsening in debt-to-GDP ratios mostly during the periods prior to World War II, when deflation episodes were more frequent. Before the Great Depression, the debt-to-GDP ratios rose by almost 0.3 percentage point a year; during the depression, they rose by 0.5 percentage point.29

The relevant literature underscores that not all deflations are alike. The origin and impact of deflation are different in recessions and expansions. Atkeson and Kehoe (2004) documented that, although the 1929 deflation was correlated with the Great Depression, the empirical link is not robust from historical perspective. Accordingly, Atkeson and Kehoe (2004) and Borio and Filardo (2004) argued that the effects of deflation depend on the growth regime. In a similar vein, Bordo, Lane, and Redish (2004) use a panel VAR to investigate the United States, the United Kingdom, and Germany during the late 19th century, and find that, overall, deflation over that period was “primarily good” for the real economy. Bordo, Lane, and Redish (2004) also conclude that deflation has received a “bad rap” and stress the importance of distinguishing between good and bad deflations.

In order to test for asymmetry in the effect of deflation on the debt-to-GDP ratio, we estimate equation (4). We consider deflation, recessionary deflation (deflation combined with negative growth), and expansionary deflation (deflation combined with positive growth). The results confirm that a state of deflation permanently increases the debt ratios (Table 3). The debt-to-GDP ratio increases by almost 1.7 percent a year, and this impact varies across growth regimes. As previously assessed, deflation has adverse effects on debt when combined with recession (“bad” or “ugly” deflations). During such episodes, the debt-to-GDP ratio is assessed to increase by 3.2 percent a year. Conversely, expansionary deflations have no impact.

Table 3.

Inflation and Growth Regimes and Debt, 1851–2013

article image
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

Primary Balance

As developed in the theoretical section, deflation can affect the debt-to-GDP ratios through the primary balance. We investigate whether the impact of deflation on debt documented above operates only through the interest rate-growth differential channel, or if it also operates through an impact on the primary balance. We first estimate equation (2). Results are reported in Table 4, columns (1)–(3). Overall, higher inflation reduces both the revenue and expenditure ratios by similar magnitudes (0.04 percentage point of GDP), leaving the primary balance broadly unchanged.30 The estimated coefficient is positive but not statistically significant.31 We then break down the inflation term into positive and negative inflations, to assess the specific effect of deflation on the primary balance. Results are reported in Table 4, columns (4)–(6). We find that deflation has no significant impact on the primary balance and its components as ratios of GDP. The effect of overall inflation previously identified is attributed to positive inflation—revenue and primary expenditure ratios contract with positive inflation but appear insensitive to deflation. We further check the effect of deflation by substituting a dummy variable to the inflation term (equation (4)) and confirm that deflation has no effect on the primary balance.

Table 4.

Deflation and Primary Balance, 1851–2013

article image
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

As a robustness check, we explore whether low inflation measured as inflation between 0 and 2 percent has an impact on fiscal aggregates different from that of deflation. The 2 percent threshold is set arbitrarily, as in recent decades several advanced economies experienced moderate inflation often below that threshold. From the historical records, we break inflation down into deflation, low inflation (between 0 and 2 percent), and inflation above 2 percent. Results are reported in Table 5. We find that low inflation has no specific effect on debt. Conversely and as documented above, deflation increases the debt-to-GDP ratio while inflation above 2 percent curbs it. We also conduct the same estimates for the primary balance and its components as ratios of GDP and find low inflation (between 0 and 2 percent) has no significant effect on the primary balance.

Table 5.

Low Inflation and Debt, 1851–2013

article image
Robust standard errors in parentheses*** p<0.01, ** p<0.05, * p<0.1

B. Beyond Ratios: Nominal Changes in Fiscal Aggregates

So far, we focused on fiscal ratios. The analysis of fiscal aggregates expressed in percent of GDP suggests that the denominator effect is preponderant on the results, and some of the impact of deflation on the numerator (the nominal fiscal variable) may thus be muted. In an attempt to quantify the impact of deflation on nominal variables, we need to find a way to circumvent the limitation of the data that, for historical records, we only have ratios, not the nominal values. We thus need to mute the denominator effect and, to that effect, we build “pseudo-nominal” changes in the fiscal variable x = X / Y by using the following formula:

Δaxt≔ΔXtYt−1=Δxt+((1+gt)(1+πt)−1)xt(5)

The change adjusted from the denominator effect, Δaxt, proxies the nominal increase, as measured in terms of previous year GDP. We use the specifications in equations (2)–(4) to scrutinize the effects of deflation on these “pseudo-nominal” variables, focusing on primary expenditure and revenue. We obtain the following three additional results, which are robust when we consider a broader definition of deflation (i.e., any inflation rate below 1 percent). This finding is not a surprise, given the upward bias in modern chained price indices, which is exacerbated by the use of more rudimentary indices at times.32,33

First, both nominal expenditure and nominal revenue are positively correlated with inflation. We find that revenue and expenditure are historically more responsive to inflation than to deflation (Tables 6–7, columns (3)). This means that, whereas positive inflation boosts nominal revenue and expenditure, deflation does not deflate them as much. This result is a sign of downward rigidities in expenditures and, probably also, in tax provisions. In addition, the denominator effect would exacerbate these asymmetries: in deflation situations, revenue and expenditure ratios tend to increase more than they decrease in positive inflation situations.

Table 6.

Inflation, Deflation, and Nominal Expenditure, 1851–2013

article image
Note: Dependent variable: First difference of fiscal aggregate in percent of GDP. Robust standard errors are in parentheses. Regressions include intercepts and time fixed effects.*** p<0.01, ** p<0.05, * p<0.1
Table 7.

Inflation, Deflation, and Nominal Revenue, 1851–2013

article image
Note: Dependent variable: First difference of fiscal aggregate in percent of GDP. Robust standard errors are in parentheses. Regressions include intercepts and time fixed effects.*** p<0.01, ** p<0.05, * p<0.1