2.1. The monetarist-structuralist debate

The orthodox or monetarist view of the money supply process, which is closely associated with the work of ^{Friedman (1959a}, ^1959b) and ^{Friedman and Schwartz (1963}, ¹⁹⁸²⁾, states that the money supply is under the direct (exogenous) control of the Central Bank through its control over the monetary base. A key assumption of monetarism is that the economy is always operating at full capacity in the long run, so that expansionary monetary policy has no effect on output and employment. The only effect of a one-percentage point increase in the money growth rate by the Central Bank would be to increase the inflation rate by the same amount. Or put in another way, the real economy is super-neutral with respect to a permanent change in the money growth rate (^{Bullard, 1999}; ^{De Grauwe and Polan, 2005}). The super-neutrality proposition of monetarism embraces a main tenet of orthodox economics in general, namely that the supply side of the economy is fixed or insensitive to demand in the long run (also see the long-run super-neutrality prediction of ^{Sidrauski’s, 1967} model).

The post-Keynesian/structuralist view of the money supply process presents a fundamental departure from the monetarist approach (^{Wachter, 1976}; ^{Moore, 1988}; ^{Palley, 1994}; ^{Nell, 2000-2001}; ^{Ghatak and Sánchez-Fung, 2007}; ^{Vera, Guijarro, and Pérez, 2022}). According to post-Keynesians and structuralists, the money supply is not under the direct control of the Central Bank but to a large extent endogenous (responsive) to the demand for bank credit. In this framework, the Central Bank sets the supply price of reserves/credit through its interest rate policy and stands ready to supply reserves/credit on demand. Even in a situation where the monetary policy instrument is not the interest rate, but the monetary base, Central Banks will tend to accommodate the demand for bank credit to preserve the liquidity of the financial system and to avoid slow growth and high unemployment rates.

The main policy implications of the two opposing views are the following. If the money supply is under the direct control of the Central Bank, consistent with the monetarist view, observed inflation is most likely caused by demand expansions in excess of fixed supply capacity. In this scenario, demand-side policies have no long-run effects on real variables such as output and the unemployment rate. To avoid the price level from rising, ^{Friedman (1959a)} proposes a fixed monetary growth rule, in which the money supply (through the Central Bank’s control over the monetary base) is constrained to grow at the same rate as real output.

In contrast, when the money supply is endogenous, consistent with the post-Keynesian/Structuralist view, the underlying cause of the inflation rate may not necessarily originate from excess demand financed by excessive monetary expansion, but rather from structural sources. In this case, increases in the money supply accommodate the inflationary process, in which the supply side of the economy is slow to adjust to demand due to several supply rigidities and bottlenecks in the system. Because inflation does not originate from pure excess demand pressure, restrictive monetary policy may slow down the natural growth and development process of the economy.

To illustrate why inflation may be the inevitable outcome of the growth and development process, consider four prime examples of structural inflation in a typical developing economy (^{Campos, 1964}; ^{Johnson, 1967}; ^{Argy, 1970}; ^{Wachter, 1976}; ^{Thirlwall, 1974b}; ^{Johnson, 1984}; ^{Ghatak and Sánchez-Fung, 2007}; ^{Nell, 2004}, ²⁰¹⁸). The first example of structural inflation relates to the dispersion of demand between sub-markets (^{Thirlwall, 1974b}). The process of development is often characterised by resource shifts (capital and labour) from declining to expanding sectors. If capital and labour are immobile, and prices and wages are rigid downwards in the declining sector, the development process is likely to be accompanied by rapid inflation, even when aggregate demand matches aggregate supply. Although the supply side of the economy is slow to adjust, wage and price increases in the expanding sectors are necessary to reallocate labour and capital in the most efficient way (^{Johnson, 1984}).

Secondly, consider a situation in a developing economy where faster demand growth in the domestic economy induces physical capital accumulation to meet demand (^{Nell, 2004}, ²⁰¹⁸). Because imported capital goods from abroad with embodied technical progress and intermediate goods are more productive than domestic resources, a large proportion of investment demand must be financed with foreign exchange earnings. However, foreign exchange shortages due to an undiversified export sector, together with the high demand for imported capital and intermediate goods, cause the nominal exchange rate to depreciate and the price of imported goods to rise. The initial rise in import prices triggers a wage-price spiral and a self-perpetuating inflationary process that is accommodated by endogenous increases in the money supply. For monetary policy purposes, some structural inflation due to foreign exchange shortages must be tolerated and viewed as part of the natural growth and development process.

Thirdly, structuralists often point to the underdeveloped nature of the agricultural sector in low-income economies as a main source of inflationary pressure (^{Cardoso, 1981}; ^{Canavese, 1982}; ^{Johnson, 1984}). According to this view, the process of industrialisation in underdeveloped economies is typically characterised by rapid urbanisation and rising demand for food. Given the inelastic nature of agricultural supply and downward rigidities in non-agricultural prices, the high demand for food causes upward pressure on the absolute price level and, in reaction to rising living costs, wages in the urban sector increase as well. Because prices in industry are set as some mark up over wage costs, the initial increase in the demand for food offsets a wage-price spiral that makes the inflationary process self-perpetuating. The wage-price spiral is accommodated by endogenous increases in the money stock.

Fourthly, structuralist writings emphasise the effect of autonomous increases in the price of food and imported goods on the fiscal deficit and how this leads to an increase in the inflation rate (for an informative review, see ^{Ghatak and Sánchez-Fung, 2007}). The initial rise in the domestic price level from structural sources would tend to widen the public deficit. This occurs because the expenditure side of the budget is typically fixed in real terms, meaning that the money value rises in the same proportion as prices. Taxes, on the other hand, are set in money terms and only adjust slowly to rising prices, causing the real value of taxes to fall. The net effect is that the fiscal deficit tends to increase in real terms, thus forcing the Central Bank to finance the deficit through expansionary monetary policy, which sets in motion a self-perpetuating inflationary process.

Monetarists, on the other hand, strongly contest the structuralist view of the inflationary process (^{Thirlwall, 1974b}; ^{Wachter, 1976}; ^{Johnson, 1984}; ^{Ghatak and Sánchez-Fung, 2007}; ^{Thirlwall and Pacheco-López, 2017}; ^{Nell, 2004}, ²⁰¹⁸). The argument is advanced that the initial increase in the real value of the fiscal deficit is the result of imprudent fiscal management and inflation from exogenous increases in the money stock by the Central Bank. The initial fiscal deficit induces an endogenous increase in the money supply to finance it, which causes a further rise in the inflation rate and a widening of the fiscal deficit in real terms. If the process is allowed to continue unabated it may lead to hyperinflation, such as those experienced in many Latin American countries (^{Aghevli and Khan, 1977}; ^{Phylaktis and Taylor, 1993}). Along the same lines, monetarists typically do not regard supply inelasticities in the agricultural sector and foreign exchange shortages as natural characteristics of underdevelopment, but rather the outcome of irresponsible macroeconomic policies. To moderate high inflation rates from excessive expansions in the money stock, price controls are often imposed on food prices which, in turn, lead to supply inelasticities in the agricultural sector. Foreign exchange bottlenecks, on the other hand, develop because inflation causes an overvalued exchange rate and a deterioration in the international competitiveness of the export sector. From a monetarist perspective, it is inflation that causes structural rigidities in the economic system, rather than the other way around.

It is clear from the examples of structural inflation above and the counterarguments of monetarists that the correct diagnosis of inflation is crucial for the conduct and efficiency of monetary policy. If inflation is diagnosed as a pure excess demand phenomenon, whereas in reality it is structural in nature, then overly conservative inflation and monetary growth targets may constrain the natural growth and development process.

Structuralist theories of inflation imply that inflation should be higher in developing countries than in developed countries. Indeed, this is what the recent evidence presented by ^{De Carvalho, Ribeiro, and Marques (2018)} confirms: Inflation, on average, is much higher in lower-income countries than advanced economies. In addition, empirical evidence in favour of structuralist theories of inflation would have to show that the money supply is endogenously determined, and that the underlying cause of inflation is not demand-pull but structural in nature. In this scenario, increases in the money supply accommodate structural sources of inflation due to supply rigidities and bottlenecks in the economy.

Looking at the monetarist-structuralist debate in Latin American countries during the 1950s and 1960s, it would perhaps be a safe bet to take a middle way and attribute high inflation rates in these economies to structural rigidities in the economic system as well as exogenous expansions in the money stock (^{Thirlwall and Pacheco-López, 2017}). But what about other developing economies that have historically not recorded excessive (hyper) inflation? There is a growing body of empirical evidence which shows that the money supply is endogenously determined in non-Latin American developing countries (see the comprehensive survey in ^{Vera, Guijarro, and Pérez, 2022}). Moreover, structural and cost-push sources of inflation appear to play a dominant role relative to monetary factors in low-income African countries (^{Heintz and Ndikumana, 2011}; ^{Durevall, Loening, and Yohannes, 2013}; ^{Nguyen et al., 2017}; ^{Ndikumana et al., 2023}), as well as in emerging market economies such as South Africa (^{Nell, 2004}, ²⁰¹⁸).

The secondary role of monetary factors as an explanation of inflation in low-income countries is often attributed to the inability of an underdeveloped banking system to transmit monetary policy shocks to the real economy (^{Durevall, Loening, and Yohannes, 2013}; ^{Mishra and Montiel, 2013}; ^{Nguyen et al., 2017}; ^{Bhattacharya and Jain, 2020}). This view certainly has merit, but it should also be kept in mind that all inflations usually require some expansions in the money supply, either through exogenous increases by the Central Bank or endogenously driven by structural and cost-push sources. Central banks in low-income and emerging market economies may therefore choose not to intervene, given the dominance of structural and cost-push forces of inflation, otherwise restrictive monetary policy measures may thwart the growth and development process.

A case in point is the different empirical results that ^{Bhattacharya and Jain (2020)} obtain for advanced countries and emerging market economies when they analyse the effect of restrictive monetary policy in response to a rise in food inflation. For advanced countries, their results show that an increase in the monetary policy (interest) rate led to further increases in food inflation as well as overall inflation, because the disinflationary effect of contractionary demand-side measures was more than offset by a rise in production costs. In contrast, for emerging market economies the monetary policy transmission mechanism is statistically insignificant, which ^{Bhattacharya and Jain (2020)} attribute to the underdeveloped nature of their financial systems. However, it is difficult to believe that the emerging market economies included in their sample (Brazil, Russia, India, China, South Korea, Chile, Mexico, Turkey and Hungary) suffer from an underdeveloped financial system in the same way as a typical low-income country does. Rather, the Central Banks in their group of emerging market economies may have chosen to be less aggressive in their policy responses because the underlying cause of inflation is not demand-pull but structural or cost-push in nature. We are back to a structuralist interpretation, in which Central Banks tend to accommodate inflationary pressure that originates from the supply side.

The structuralist interpretation provided here implies that the twin objectives of low inflation and fast growth may not be obtainable through restrictive macroeconomic policy measures. Instead, what is required is a coordinated, economy-wide anti-inflationary programme that directly addresses the rigidities that originate in the agricultural sector, trade sector, labour market and product market (^{Nell, 2018}).

To summarise, the heated debate between monetarists and structuralists that took place in the 1950s, 1960s and some part of the 1970s to analyse the high-inflation experiences of many Latin American countries is just as relevant in today’s developing countries. Although the policy instrument of many Central Banks is no longer the monetary base but the interest rate, the influence of monetarism is still evident in the belief that the underlying cause of inflation is excess demand, and that Central Banks are able to effectively control this source of inflation through their control over interest rates (^{Nelson and Schwartz, 2008}; ^{Jahan and Papageorgiou, 2014}). This point is underscored by recent developments in the world economy to moderate the surge in inflation from supply-side bottlenecks that developed after the COVID-19 pandemic and the sharp rise in food and energy prices (cost inflation) related to the War in Ukraine. Central Banks in advanced economies, such as the United States (US), the United Kingdom (UK) and European Union, have reacted by raising their monetary policy rates.

In a special issue organized by the editors of Review of Keynesian Economics (2023) on the effectiveness of restrictive monetary policy measures in advanced countries, the contributors seem to be unanimous in their view that restrictive demand-side policy measures may not have the desired effect, despite the view by Central Banks that generous COVID-19 relief measures created excess demand (^{DeLong, 2023}; ^{Vernengo and Pérez Caldentey, 2023}; ^{Chowdhury and Sundaram, 2023}; ^{Blanchflower and Bryson, 2023}). In fact, restrictive demand-side policy measures in the presence of structural and cost-push inflation would most likely cause a recession and generate further inflationary pressure by raising production costs, as underlined by the empirical evidence on food inflation in advanced countries discussed above (see ^{Bhattacharya and Jain, 2020}). The only way to maintain growth and employment levels and subdue inflationary pressure at the same time is to design policy measures that directly address supply-side bottlenecks and cost inflation (e.g. by speeding up the switch to renewable energy sources), with monetary policy playing an accommodating role. Allowing inflation to rise from supply-side bottlenecks may also assist in allocating labour and capital in a more efficient way and eventually moderate inflationary pressure.

The overly conservative monetary policy stance in advanced economies may further cause undesirable ripple effects in underdeveloped and emerging market (UEM) economies. Large interest rate differentials between UEM economies and advanced countries could trigger another round of imported inflation in UEM economies through capital outflows and nominal exchange rate depreciations (^{Chowdhury and Sundaram, 2023}). To counter imported inflation triggered by advanced countries, UEM economies may be forced to raise their own monetary policy rates as well. The importance of the monetarist-structuralist debate in present-day economies is self-evident.

2.2. The benefits of mild inflation

A main premise in ^{Thirlwall’s (1974b)} ISGD book is that the growth benefits of mild inflation come from demand inflation instead of structural or cost-push inflation. Structural inflation is the unavoidable outcome of the growth and development process itself. To reduce structural inflation through restrictive demand-side policy measures would be to impair growth. But there is nothing in the process of structural inflation and pure cost inflation (autonomous increases in unit labour costs and rising food and energy prices) that directly stimulates growth in a causal sense (^{Thirlwall, 1974b, p. 233}).

To formalise the growth-effects of demand inflation à la ^{Thirlwall (1974b)}, consider an initial equilibrium position, using ^{Harrod’s (1939)} three different growth rates and some of the modifications suggested by ^{McCombie (2011)}:

Where g_d is the growth rate of demand or actual growth rate; g_w is the warranted growth rate; g_s is the growth rate of supply; and g_n is the natural growth rate. The warranted growth rate is the rate of growth that keeps entrepreneurs satisfied in the sense that their capital is fully utilised, so that their plans to invest (i^P) are equal to their plans to save (s^P). Any deviation of demand growth (g_d) from the warranted growth rate (g_w) would cause entrepreneurs to revise their investment plans. The natural growth rate (g_n) is the maximum potential growth rate that an economy can achieve, given the rate of growth of population, physical and human capital accumulation, and technological progress. Note that the maximum or natural growth rate is a hypothetical growth rate, which means that a typical underdeveloped economy would tend to grow below this rate in the long run (see ^{McCombie, 2011}).

Suppose a monetary or fiscal stimulus initially raises demand growth above the warranted growth rate and the rate of growth of supply:

Plans to invest now exceed plans to save (i^P > s^P). One way to finance the deficit is to borrow from domestic and foreign capital markets, but these sources of finance involve interest payments and future debt obligations. An attractive alternative is to allow inflation from excess demand pressure in the product market to raise the actual saving-to-income ratio (s) by redistributing income from wage earners to profits earners and from the private sector to the government sector. In effect, the inflationary pressure from excess demand pulls up the warranted growth rate and the growth rate of supply:

Consistent with the main hypothesis in ^{León-Ledesma and Thirlwall (2002)}, equation [3] implies that the growth rate of supply is endogenous to the growth rate of demand through static economies of scale associated with large scale production techniques and lower average production costs, as well as dynamic economies of scale from a faster rate of capital accumulation, embodied technical progress and learning-by-doing effects. Formally, ^{Setterfield (2006)} shows that the growth rate of supply adjusts to the growth rate of demand through a rise in the Verdoorn coefficient (for more details, see ^{McCombie, 2011}).

The discussion now turns to a more detailed examination of the Kaldor-Thirlwall model of forced saving and the inflation tax model. Although ISGD highlights these two models as the main channels through which inflation could stimulate development, it also covers other Keynesian views and neoclassical monetary growth models of inflationary finance. The discussion below briefly incorporates some of these additional views on the growth benefits of mild inflation, as well as more recent ones that have gained prominence after the publication of ISGD.

2.2.1. The Kaldor-Thirlwall model of forced saving

Suppose the Central Bank decides to implement more expansionary demand-side policy measures to finance development through inflationary means. In ^{Thirlwall’s (1974a}, ^1974b) Kaldorian-type model (^{Kaldor, 1955-1956}), the initial response of the private sector to the demand-side stimulus is to raise planned investment relative to planned saving to match the increase in demand for goods. Excess demand in the product market generates inflation, which redistributes income from wage earners to profit earners through a fall in real wages. If the marginal propensity to save out of profit income exceeds the marginal propensity to save out of wage income, the aggregate saving ratio rises. Thus, investment finances itself through a rise in the saving ratio. Or, put in another way, inflation finances business fixed investment and faster economic growth.

Formally, ^{Thirlwall (1974a}, ^1974b) shows that the extra increase in the saving ratio for each percentage point increase in the inflation rate can be calculated from the following equation:

Where S_t/Y_t is the saving-to-income ratio at time t; p is the inflation rate; α is the wage-price coefficient; W₀ is the initial wage bill; Y₀ is the initial income level; a is autonomous wage growth; s_w is the marginal propensity to save out of wages; and s_r is the marginal propensity to save out of profits.

It is apparent from equation [4] that the success of inflationary finance will crucially depend on the response of wage growth with respect to the rate of inflation (α) and the difference between the marginal propensity to save out of wages and the marginal propensity to save out of profits. Without money illusion (α = 1) or with equal saving propensities across wage earners and profit earners (s_w = s_r), the extra increase in the saving ratio for each percentage point increase in the inflation rate is zero: ∂(S_t/Y_t)/∂p = 0. On the other hand, with money illusion (α < 1) and a higher saving propensity of profit earners relative to wage earners (s_r > s_w), inflation finances investment through an increase in the saving ratio: ∂(S_t/Y_t)/∂p > 0. The calibration exercises of equation [4] in ^{Thirlwall (1974b)} and ^{Thirlwall and Pacheco-López (2017)} suggest that, for reasonable parameter values, there is some scope to raise the saving ratio by a non-trivial amount without reverting to excessive inflation.

One objection that has been raised against the Kaldor-Thirlwall model is that the money illusion assumption (α < 1) only holds in the short run (^{Woodfield and McDonald, 1978}; ^{Johnson, 1984}). Critics argue that wage earners will resist real wages from falling below a certain threshold, implying that the wage-price coefficient approximates unity in the long run. This interpretation, however, assumes a static economy without labour productivity growth. As ^{Thirlwall (1974b}, ¹⁹⁷⁸⁾ formally demonstrates, the behaviour of wage earners would most likely be different in a growing economy. Although real wages decrease from an increase in inflation (α < 1), the overall effect on real wage growth could be positive if autonomous wage growth (a > 0) from other determinants (e.g. efficiency wages) more than offsets the fall in real wages from inflation. Furthermore, as long as positive growth in real wages falls below labour productivity growth, the share of profit income will rise. Thus, wage earners are likely to be less resistant to real wage cuts from inflation in a growing economy, implying that α < 1 holds in both the short run and long run.

These counterarguments imply that inflation via the Kaldor-Thirlwall model does not necessarily imply an overall fall in real wages. It also means that a faster rate of capital accumulation through inflationary finance may reduce structural unemployment in capital scarce developing economies (^{Thirlwall, 1974a}, ^1974b).

^{FitzGerald (2006)} seems to raise another issue against the practical relevance of the Kaldor-Thirlwall model. He constructs backward- and forward-looking wage-price models, in which the behaviour of real wage growth seems to be ‘explosive’ with respect to increasing inflation rates, and concludes, on the basis of these instability properties, that the Kaldor-Thirlwall model of forced saving is only relevant in the short run (^{FitzGerald, 2006, p. 255}). However, there seems to be a disconnect between ^{FitzGerald’s (2006)} models and the actual assumptions of the Kaldor-Thirlwall in equation [4]. Forced saving in equation [4] does not require increasing inflation rates, but a one-off increase in the inflation rate. ^{FitzGerald (2006, p. 255)} further remarks: “Moreover, in the modern world of independent central banks and inflation targeting, rates of domestic inflation sufficient to force up the savings rate are not feasible, and in any case the uncertainty this causes seem to reduce savings…”

The first point to make is that the debate is not whether a country adopts an inflation-targeting regime or not, but rather whether the official or unofficial inflation target is too low, given the benefits of mild inflation. The second point is that the argument is about the benefits of mild inflation rather than excessive inflation that may discourage voluntary saving. In the US economy, for example, there is a strong view that the unofficial inflation target of 2% is too low and that the Federal Reserve should increase it to at least 4% (^{Palley, 2000}; ^{Blanchard, Dell’Ariccia, and Mauro, 2010}; ^{Ball, 2013}; ^{Krugman, 2014}). Interestingly, one of the arguments in favour of a higher inflation target in the US economy has some affinity with the money illusion assumption of the Kaldor-Thirlwall model. ^{Akerlof, Dickens, and Perry (1996)} show that downward rigidity in nominal wages and price stability (zero inflation) may cause a rise in the natural unemployment rate because real wages in declining sectors are inflexible downwards (also see ^{Akerlof et al., 2000}). A higher inflation rate up until 4% has the advantage of creating real wage flexibility to keep the natural unemployment rate at a lower level (^{Palley, 2000}; ^{Krugman, 2014}). In the case of emerging market economies with inflation targets, the policy debate is whether targets in the range of 3%-5% should be revised upwards to say 6%-8%, and whether slow-growing underdeveloped countries should strive for low inflation rates, such as those observed in more developed economies.

As an alternative to inflationary finance of development, ^{FitzGerald (2006)} develops a model that links a real exchange rate depreciation to a real wage reduction. It is then shown that the saving ratio in equation [4] is a positive function of a real exchange rate depreciation rather than the inflation rate. However, assuming downward rigidity in nominal wages, and the fact that the real exchange rate is a function of real wage changes, it is not clear how real wages are supposed to fall without reverting to some inflation (see e.g.^{Gala, 2008}). Although the open-economy model of ^{Fitzgerald (2006)} is an interesting extension of the Kaldor-Thirlwall model, the saving ratio remains a positive function of the inflation rate if it is assumed that nominal wages are rigid downwards. The open-economy implications of inflationary finance are discussed in a later section.

2.2.2. The inflation tax model

In addition to the Kaldor-Thirlwall model of forced saving, ^{Thirlwall’s (1974b)} ISGD book highlights the potential of an inflation tax on real money holdings as another source of finance in underdeveloped economies. The relevance of the inflation tax model is premised on the observation that tax revenue with respect to income is typically inelastic in underdeveloped countries, while tax collection involves large administrative costs. It is further assumed that there are many productive investment projects of the infrastructure type and research and development (R&D) activities with high social returns, but low private returns because private entrepreneurs cannot capture all the benefits themselves, that need to be undertaken by the government. In this situation, the government’s plans to invest in excess of plans to save cannot be financed with normal tax revenue due to the inelastic nature of the tax system.

One way to finance government expenditure in excess of normal tax revenue is to impose an inflation tax on real money holdings (^{Bailey, 1956}; ^{Cagan, 1956}; ^{Mundell, 1965}; ^{Johnson, 1967}; ^{Friedman, 1971}). Consider a simplified version of the inflation tax model in Figure 1.

Sources: ^{Thirlwall (1974b)}; ^{Ghatak and Sánchez-Fung (2007)}.

Figure 1 The inflation tax model 

The vertical axis measures the opportunity cost of holding real money balances and is calculated as the difference between the real rate of return of physical capital (r) and what is earned from holding real money balances (i - p): r - (i - p), where i is the nominal interest rate and p is the inflation rate. Assuming that the holdings of real money balances pay no nominal interest, the opportunity cost of holding money balances is simply r + p. The horizontal axis in Figure 1 measures the holdings of real money balances, (M/P), where M is nominal money balances and P is the price level.

The initial equilibrium position at point A in Figure 1 assumes zero inflation and a corresponding level of real money holdings of (M/P)₁. Suppose the government borrows from the Central Bank to finance new investment outlays on productive infrastructure projects. Based on the Quantity Theory approach that underlies a monetarist view of inflation, the increase in the rate of growth of the monetary base to finance the public deficit leads to an equi-proportional increase in the inflation rate from p = 0 to p₁ > 0. The real yield from the inflation tax (R) for investment purposes is the shaded area in Figure 1 and equal to the tax rate (p₁) multiplied by the tax base (M/P)₂:

Note that, with inflation, the tax base (M/P) continuously deteriorates unless it is offset by equi-proportional increases in M. To keep the tax base constant at (M/P)₂, the model assumes a real balance effect where the holders of money consume less and use the additional resources (forced saving) to supplement M. Put in another way, the real yield from the inflation tax, R, is equivalent to the real income foregone by the holders of money to keep their real money balances intact in the presence of inflation. Thus, an inflation tax diverts real resources from the holders of money to the government who invests the proceeds on behalf of society.

The textbook model in Figure 1 implies that the optimum inflation rate (tax rate) that maximises the real yield is the rate at which the elasticity of demand for real money balances with respect to the inflation rate is unity (see ^{Friedman, 1971}; ^{Thirlwall, 1974b}; ^{Ghatak and Sánchez-Fung, 2007}). For inflation (tax) rates beyond this optimum rate, the tax base falls by a larger percentage and the real yield declines. ^{Friedman (1971)}, however, shows that the optimum inflation rate that maximises the real yield is lower when the rates of growth of population and real income are positive, rather than zero as assumed in the textbook model. The revenue-maximising inflation rate is lower under these conditions because the issuer of money (government) obtains an additional source of revenue from positive growth in the demand for real money balances as income and population rise.

Despite ^{Friedman’s (1971)} own, positive contribution, and the fact that the inflation tax model originates from monetarist theories, he remains sceptical about the effectiveness of inflationary finance for development purposes: “My own personal view is that inflation is neither desirable nor necessary, that the most effective road to development is through free enterprise and private investment, and that the government can serve best by limiting itself to essential government functions, keeping taxes of all kinds low, refraining from intervention into the economy, and providing a stable monetary framework.” (p. 847).

^{Friedman’s (1971)} aversion to inflation is consistent with his monetary growth rule mentioned earlier (^{Friedman, 1959a}), which implies an optimum inflation rate of zero, because the real economy is insensitive to expansionary demand-side policies in the long run. Economists of a structuralist and Keynesian persuasion, in contrast, take a very different view. The emphasis is on supply rigidities in the tax system, and the fact that inflationary demand-side policies can help capital scarce developing countries to raise their long-run growth rates through a faster rate of physical capital accumulation (^{Thirlwall, 1974b)}.

Whether an inflation tax supplements total government revenue in a significant way will also depend on lags in the collection of taxes. ^{Tanzi (1977}, ¹⁹⁷⁸, ¹⁹⁸²⁾ emphasises that the real yield obtained from an inflation tax could be offset by a fall in normal tax revenue. Earlier it was mentioned that the real value of normal tax revenue tends to fall with inflation, whereas government expenditure remains fixed in real terms. If there are lags in tax collection, the real yield of an inflation tax may be partly offset by a fall in the real value of normal taxes. If the initial level of normal tax revenue-to-Gross Domestic Product (GDP) ratio is high and there are long lags in tax collection, the net additions to a government’s total revenue from inflationary finance may be greatly diminished. Since the initial level of normal tax revenue-to-GDP ratio is typically low in underdeveloped economies, the effect of inflationary finance on total government revenue would be optimised if strategies could be employed to reduce long delays in tax collection.

Recently, the merits of money-financed government deficits (as depicted in Figure 1) to mitigate the recessionary conditions caused by the 2008-2009 global financial crisis and COVID-19 pandemic have gained popularity (^{Buiter, 2014}; ^{Turner, 2015}; ^{English, Erceg, and Lopez-Salido, 2017}; ^{De Grauwe, 2020}; ^{Agur et al., 2022}). One of the main advantages of a money-financed deficit relative to a debt-financed fiscal stimulus is that the former involves no future debt obligations and is therefore more effective in stimulating aggregate demand (^{De Grauwe, 2020}; ^{Agur et al., 2022}).

Inflation that emanates from a money-financed fiscal deficit may also reduce the cost of borrowing through lower real interest rates and act as stimulus to investment, which, in addition to the Kaldor-Thirlwall model, is another Keynesian view of the benefits of mild inflation (^{Thirlwall, 1974b}; ^{Galí, 2020}). The Keynesian approach further emphasises that the nominal rate of return on investment in physical assets, such as machinery and equipment with embodied technical progress, tends to rise with inflation. Mild inflation, therefore, encourages investment in productive physical assets to maintain profitability relative to money assets, hoarding, inventories and more speculative activities (^{Thirlwall and Barton, 1971}; ^{Thirlwall, 1974b}).

Figure 1 can also be used to illustrate an additional channel through which inflation could raise capital accumulation. In addition to the increase in government revenue from the inflation tax, ^{Tobin’s (1965)} model predicts that the increase in the opportunity cost of holding money from r to r + p₁ would entice wealth-holders to reduce their holdings of real money balances from (M/P)₁ to (M/P)₂ and use the proceeds to invest in physical capital with a higher relative rate of return (^{Thirlwall, 1974b}; ^{Orphanides and Solow, 1990}). Although ^{Tobin’s (1965)} model is an attempt to introduce money into ^{Solow’s (1956)} neoclassical growth model, it is analogous to one of the Keynesian views of inflationary finance discussed above where inflation maintains the profitability of physical assets relative to money assets.

The ^{Tobin (1965)} model rests on the assumption that real money balances and physical capital are substitutes. If, however, they are complements, the additional channel of inflationary finance will not materialise (^{Thirlwall, 1974b}; ^{Orphanides and Solow, 1990}; ^{Ghatak and Sánchez-Fung, 2007}). In ^{Stockman’s (1981)} cash-in-advance model and the complementarity hypothesis of ^{McKinnon (1973)} and ^{Shaw (1973)}, for example, the accumulation of real money balances and physical capital occur in tandem. In the ^{McKinnon (1973)} and ^{Shaw (1973)} hypothesis, the accumulation of real money balances (saving) via financial intermediaries plays an important role to fund large-scale investment projects. Since inflation erodes the real value of money balances and acts as a deterrent to holding money, the rate of physical capital accumulation and output growth fall with inflation.

To summarise, ^{Thirlwall’s (1974b)} ISGD book identifies several models that predict a positive relationship between inflation, investment, and growth. Inflation in these models redistributes income from wage earners to profit earners and from the private sector to the government and, in the process, generates domestic resources to finance plans to invest in excess of plans to save. Mild inflation may also encourage investment in physical assets relative to money assets, but this proposition only holds if physical capital and real money balances are substitutes instead of complements.

2.3. Open-economy implications of inflationary finance

Arguably, the biggest obstacle to a successful policy of inflationary finance is an economy’s balance of payments. ^{Thirlwall’s (1974b)} ISGD book devotes a whole chapter to the open-economy implications of inflationary finance and pre-empts many of the arguments in his seminal paper (^{Thirlwall, 1979}) on balance-of-payments-constrained (BPC) growth. To illustrate how an economy’s balance of payments may thwart a policy of inflationary finance, consider ^{Thirlwall’s (1979)} original growth law:

Where g_b is the BPC growth rate; x is the growth of export; ε is the income elasticity of demand for exports; π is the income elasticity of demand for imports; and z is world income growth.

Augmenting the equilibrium outcome of a monetary or fiscal demand-side stimulus in equation [3] with the BPC growth rate (g_b) in equation [6], gives:

Equation [7] illustrates that, although the growth of supply is endogenous to demand, the rate of growth of demand exceeds the BPC growth rate: gd'>gb' . Because each demand component (investment spending, consumption spending and government spending) comprises an important content, faster demand growth relative to the BPC growth rate implies a growing current account deficit, with imports growing faster than exports. Since an economy cannot borrow indefinitely in foreign markets and because the real exchange rate is an inefficient adjustment mechanism, demand growth must be constrained to its initial level to preserve equilibrium on the current account of the balance of payment. We go back to the initial condition in equation [1] without the demand stimulus, but now with the BPC growth rate imposed:

Equations [7] and [8] capture the main arguments in ^{Thirlwall (2001}, ²⁰¹⁹⁾. The growth of supply is endogenous to demand, but the growth of demand itself is constrained by the balance of payments. To relax the demand constraint imposed by the balance of payments would require an export-led strategy that improves the structural demand features of exports in equation [6] through an increase in the income elasticity of demand for exports (↑ε) and/or import-substitution strategies that reduce the income elasticity of demand for imports (↓π). Foreign exchange earnings generated from faster export growth allow faster demand growth for domestically produced goods and an endogenous response from the growth of supply, by financing the import content of expenditure growth (^{McCombie, 1985}). Foreign exchange earnings from exports can also be used to purchase imported capital and intermediate goods that are more productive than domestic resources.

Is it possible to reconcile a policy of inflationary finance with the BPC growth model? The view advanced here, as well as those in Chapter 8 of ISGD, is in the affirmative, provided that the demand-induced inflationary finance policy is accompanied by foreign capital inflows to finance the current account deficit during the initial stages. Let us first consider in more detail the role of the real exchange rate to support faster demand growth consistent with current account equilibrium.

Generally, the BPC growth literature downplays the role of the real terms of trade (or real exchange rate) as an efficient balance-of-payments adjustment mechanism (^{McCombie, 2011}; ^{Thirlwall, 2011}, ²⁰¹⁹). To illustrate this proposition, consider the growth rate of the real terms of trade, starting from initial values of zero for all the variables:

Where, at time t, p_dt is the growth rate of export prices; p_ft is the growth rate of import prices; and e_t is the growth rate of the nominal exchange rate.

Assuming that the growth rate of export prices approximates the domestic inflation rate (p_t), such that p_dt ≈ p_t, inflationary finance would initially cause an appreciating real terms of trade (p_dt > p_ft): p_dt - p_ft - e_t > 0. Under a flexible exchange rate system, the difference between p_dt and p_ft could be offset by a depreciating nominal exchange rate (e_t > 0) to maintain international competitiveness, assuming that the absolute sum of the price elasticities in the export and import demand function is greater than unity (i.e. the Marshall-Lerner condition is satisfied). But a depreciating nominal exchange rate may itself cause a further increase in the domestic inflation rate through a rise in import prices (imported inflation) and an ensuing wage-price spiral, bringing the economy back to its initial position: p_dt - p_ft - e_t > 0. Nevertheless, studies often find that the absolute sum of the price elasticities in the export and import demand function is not significantly different from unity i.e. the Marshall-Lerner condition is just met, which would nullify the adverse balance-of-payments effect of an appreciating real terms of trade (^{Bahmani Harvey, and Hegerty, 2013}; ^{McCombie, 2023}). Thus, although the balance of payments deficit may not necessarily worsen from an appreciating real terms of trade, it is clear that the real terms of trade is an inefficient mechanism to correct the current account deficit from inflationary finance.

Alternatively, a typical developing economy with a high income elasticity of demand for imports would have to rely on foreign capital inflows to support the internal demand expansion during the initial stages (^{Thirlwall, 1974b)}. Foreign borrowing to finance capital imports has the undesirable side-effect that it requires interest payments and future debt obligations in scarce foreign currency. Indeed, the evidence shows that financing investment and growth out of foreign resources (foreign debt) is often too risky and counterproductive (^{Cavallo, Eichengreen, and Panizza, 2018}). Nonetheless, at least in the short run, domestic saving and foreign exchange are not close substitutes, making foreign capital inflows an important source of finance to purchase imported capital and intermediate goods and to support faster expenditure growth with import contents.

It should be acknowledged, however, that a policy of mild demand inflation that keeps resources fully employed and maintains the profitability of investment in physical assets relative to money assets may attract a relatively stable source of external finance, such as foreign direct investment (FDI), and reduce the need to borrow in foreign capital markets to finance the balance-of-payments deficit. The importance of buoyant demand growth in the host country, as a key determinant of FDI, is supported by the theoretical models and evidence presented in ^{Rob and Vettas (2003)} and ^{Conconi, Sapir, and Zanardi (2016)}.

Other sources of external finance in low-income countries are official development assistance (ODA) that can take the form of grants and concessional loans, as well as remittances from abroad. Although the effectiveness of ODA has been hotly debated (see e.g.^{Sachs et al., 2004}; ^{Easterly, 2006}; ^{Doucouliagos and Paldam, 2009}; ^{Lof, Mekasha, and Tarp, 2015}; ^{Herzer et al., 2015}), there is growing evidence that foreign aid has benefited low-income countries, with specific reference to sub-Saharan Africa (SSA) countries, through its positive effect on per capita income and other macro- and micro-related outcomes (^{Juselius, Møller, and Tarp, 2014}; ^{Arndt, Jones, and Tarp, 2015}). Looking at remittances from abroad, the meta-analysis of ^{Cazachevici, Havranek, and Horvath (2020)} across 95 studies suggests a positive, albeit relatively small, effect on economic growth.

A standard argument against foreign aid and remittances is that these sources of external finance undermine the recipient economy’s international competitiveness through a ‘Dutch disease effect’, by raising the price of non-tradables (P_N) relative to the price of tradables (P_T), with a resultant appreciation in the internal real exchange rate: ↓(P_T/P_N) [see the overviews in ^{Addison and Baliamoune-Lutz, 2017}; ^{Cazachevici, Havranek, and Horvath, 2020}; ^{Ahmad, Pentecost, and Stack, 2023}]. An appreciation of the internal real exchange rate, however, should not be seen as an inevitable outcome of foreign aid and remittances. The evidence for African countries in ^{Ahmad, Pentecost, and Stack (2023)} suggests that one-third of the Dutch disease effect of foreign aid is offset by external debt interest repayments. In fact, foreign aid and remittances can support a policy of inflationary finance by providing foreign exchange to purchase essential capital imports and by financing the import content of faster expenditure growth, all of which may offset any downward pressure on the internal real exchange rate (also see ^{Ahmad, Pentecost, and Stack, 2023}).

To summarise, in an open-economy context, a policy of inflationary finance can be effective if it is supported by foreign capital inflows during the initial stages. Although demand creates its own capacity, this is a necessary condition to sustain demand over the medium run. This would ensure that enough time elapses for the capacity-creating effect of demand to favourably affect the structural parameters of the BPC growth model ↑(ε/π) in equation [6] over the long run². Government outlays on productive infrastructure projects, education and R&D activities that are partly financed through an inflation tax would complement private investment (^{Barro, 1990}; ^{Easterly and Rebelo, 1993}; ^{Fedderke, Perkins, and Luiz, 2006}) and, in the process, relax the balance-of-payments constraint by generating export capacity and reducing the propensity to import.

In effect, a successful policy of inflationary finance would alter some of the fundamental determinants of the internal real exchange rate, P_T/P_N. Government investment in productive infrastructure with high social returns, such as ports, electricity supply, water supply and domestic transport, would cause a depreciation in the internal real exchange rate ↑(P_T/P_N) through a fall in the price of non-tradables ↓(P_N). This would raise the relative profitability of tradables and induce another round of investments in physical capital and R&D activities in the tradable sector, all of which may relax the BPC growth rate in equation [6] via an increase in the relative income elasticities (↑ε/π). In line with these arguments, several studies on BPC growth have endogenised the relative income elasticities in equation [6] with respect to a level change in the real exchange rate (^{Missio and Jayme 2012}; ^{Missio, Araujo, and Jayme, 2017}; ^{Setterfield and Ozcelik, 2018}; ^{Cimoli, Pereima, and Porcile, 2019}; ^{Marconi et al., 2021}).

Thus, although inflationary finance could lead to an appreciation in the real exchange rate during the initial stages, as argued in ^{Johnson (1984)}, over the medium to long run it is entirely consistent with a depreciation in the real exchange rate. Indeed, there is a large literature that finds a positive relationship between growth and a one-off depreciation in the real exchange rate (see the extensive surveys in ^{Rapetti, 2020}; ^{Blecker, 2022}; ^{Demir and Razmi, 2022}; ^{Ridhwan, Ismail, and Nijkamp, 2023}). Note, however, that the real exchange rate depreciation of inflationary finance is not achieved by manipulating the nominal exchange rate, but rather through a change in fundamentals. Arguably, this is a more viable development strategy than direct interventions in the foreign exchange market to depreciate the real exchange rate.

2.4. Costs of inflation

The call for inflationary finance of development does not disregard the fact that excessive inflation can be seriously harmful to growth and development (^{Thirlwall, 1974b)}. It is fair to say, however, that standard textbooks and authoritative surveys on the topic tend to emphasise the costs of inflation relative to the benefits of mild inflation and typically ignore the fact that some inflation may be regarded as the inevitable outcome of the development process itself (^{Briault, 1995}; ^{Temple, 2000}; ^{Romer, 2019}). The rest of this section provides a brief overview of the costs of inflation, given its extensive coverage in the literature.

Consider some of the costs of excessive inflation. Deliberate inflationary policies that cause one country to inflate at a much faster rate than others could cause severe balance-of-payments difficulties, requiring import substitution strategies and exchange controls that may lead to inefficiencies in resource allocation (^{Thirlwall, 1974b)}. The uncertainty associated with high and variable inflation could encourage investment in real estate, inventories, and speculative activities instead of productive investment in physical plant and machinery (^{Thirlwall, 1974b}; ^{Briault, 1995}). Moreover, uncertainty about future prices could force investors to delay their decisions because investment in physical capital is a sunk cost and largely irreversible (^{Pindyck, 1991}; ^{Carruth, Dickerson, and Henley, 2000}). Excessive and volatile (unanticipated) inflation may further cause confusion between relative and aggregate price changes, which leads to a misallocation of scarce resources and slower growth (^{Briault, 1995}). Another cost of high and variable inflation is the uncertainty it creates between savers and investors. If savers and investors form different expectations about the ex-ante real interest rate, capital will be misallocated (^{Briault, 1995}).

The theoretical models covered up until this point suggest that the relationship between economic growth, investment and inflation may be non-linear. Up until a threshold inflation rate, mild inflation stimulates investment and growth, but thereafter the distortionary effect of inflation dominates. The discussion now turns to the empirical evidence on the inflation-growth nexus.

3.1. More recent evidence since the early 1990s

Surveys of the literature often interpret ^{Fischer’s (1993)} study as the first one to identify the possibility of a non-linear relationship between inflation and growth, and that his evidence in fact shows a positive relationship at low inflation rates and a negative relationship at high inflation rates (^{Khan and Senhadji, 2001}; ^{López-Villavicencio and Mignon, 2011}; ^{Baharumshah, Slesman, and Wohar, 2016}). To quote ^{Khan and Senhadji (2001)} in full: “The possibility of such a nonlinear relationship was first identified by ^{Fischer (1993)}, who noted the existence of a positive relationship at low rates of inflation and a negative one as inflation rose (which weakened as inflation increased).” (p. 2).

From the overview of earlier studies in the previous section, it is clear that these statements are factually incorrect. ^{Thirlwall’s (1974b)} ISGD book and some of the earlier studies reviewed therein, such as ^{Dorrance (1966)}, were the first ones to examine the empirical possibility of a non-linear relationship inflation and growth. Moreover, ^{Fischer (1993)} does not in any way find a positive relationship at low levels of inflation, as claimed by ^{Khan and Senhadji (2001, p. 2)} and ^{López-Villavicencio and Mignon (2011, p. 455)}. For the three inflation groups identified in ^{Fischer (1993)}, the effect of inflation on growth and other outcome variables is consistently negative and generally statistically insignificant across the inflation ranges of 15% or less, between 15%-40% and above 40% (see Table 8 in ^{Fischer, 1993}). Only for the full sample of countries does ^{Fischer (1993)} find a negative and statistically significant effect of inflation on per capita income growth, capital accumulation and productivity growth. Most studies since the early 1990s reviewed below use ^{Fischer’s (1993)} study as a benchmark. In doing so, however, they neglect some of the major insights of earlier work, such as Thirlwall’s (1974b) ISGD book.

Using more refined spline techniques than the ^{Fischer (1993)} study to identify whether there is a non-linear relationship between per capita income growth and inflation, the panel data studies of ^{Sarel (1996)} and ^{Ghosh and Phillips (1998)} identify threshold inflation rates of 8% and 2.5%, respectively. ^{Sarel (1996)} finds that up until 8% the effect of inflation on per capita income growth is positive but statistically insignificant. Beyond the 8% threshold the inflation effect is negative, statistically significant, and quantitatively large. ^{Ghosh and Phillips (1998)} show that the effect of inflation on per capita income growth is positive and statistically significant up until a low threshold of 2.5%, but thereafter turns negative and statistically significant. Both studies emphasise the costs of inflation since the inflation effect is either statistically insignificant up until the threshold or the threshold is very low. However, both studies do not test whether the threshold inflation rate differs across developed and developing countries, implying that the optimum inflation rates of 8% and 2.5%, together with their respective significance levels, may be biased. We know from earlier studies that the threshold inflation rates may differ across rich and poor countries (^{Thirlwall, 1974b}; ^{Johnson, 1984}).

Table 1 provides a selective summary overview of studies that have tested for a non-linear relationship between inflation and growth in developed and developing countries, using either GDP growth or GDP per capita growth as the dependent variable. All the studies use panel data, except for the ^{Yemba, Kitenge, and Woodburne (2020)} article, that specifically looks at the Democratic Republic of the Congo (DRC) over the period 1975-2017.

Consider the evidence for developed countries in Table 1. Studies tend to identify relatively low threshold values that range between 1% and 5%. The majority of these studies find that the effect of inflation on growth up until 1%-5% is positive and statistically significant, but thereafter turns negative and statistically significant (^{Khan and Senhadji, 2001}; ^{López-Villavicencio and Mignon, 2011}; ^{Kremer, Bick, and Nautz, 2013}; ^{Azam and Khan, 2022}). ^{Ibarra and Trupkin (2016)}, on the other hand, find that the effect of inflation on growth up until 4.5% is positive, but statistically insignificant. Beyond the 4.5% threshold value the inflation effect becomes negative and statistically significant. The ^{Espinoza, Leon, and Prasad (2012)} study is the only one that reports an insignificant non-linear effect, but a negative and statistically significant effect in their linear model. ^{Sepehri and Moshiri (2004)}, in contrast, obtain insignificant non-linear and linear effects for their sample of developed (Organisation for Economic Co-operation and Development, OECD) economies. Overall, and consistent with earlier studies by ^{Thirlwall and Barton (1971)} and ^{Thirlwall (1974b)}, the evidence for developed countries points to the growth benefits of mild inflation up until relatively low thresholds of 1%-5%.

The analysis now turns to the evidence for developing countries in Table 1. Studies tend to find a positive and statistically significant effect of inflation on growth up until relatively high thresholds that range between 6% and 19% (^{Khan and Senhadji, 2001}; ^{Sepehri and Moshiri, 2004}; ^{Pollin and Zhu, 2006}; ^{Bick, 2010}; ^{Baharumshah, Slesman, and Wohar, 2016}; ^{Ndoricimpa, 2017}; ^{Yemba, Kitenge, and Woodburne, 2020}; ^{Azam and Khan, 2022}). It is noteworthy how these results contrast with those in ^{Thirlwall and Barton (1971)} and ^{Thirlwall (1974b)}, whose empirical analyses were constrained by a lack of data over a long time period. Indeed, ^{Thirlwall’s (1974b)} observation that the insignificant inflation effect in developing countries over his short sample period (1958-1968) does not necessarily mean that some inflation is not beneficial for growth, seems prophetic in the context of more recent evidence that draws on much longer time spans (see the time periods in column 2 of Table 1).

It is further noteworthy that several studies report statistically insignificant inflation effects for developing countries, or different income groups within developing countries, up until relatively high inflation thresholds of 6% to 21% (^{Sepehri and Moshiri, 2004}; ^{López-Villavicencio and Mignon, 2011}; ^{Baglan and Yoldas, 2014}; ^{Espinoza, Leon, and Prasad, 2012}; ^{Kremer, Bick, and Nautz, 2013}; ^{Seleteng, Bittencourt, and van Eyden, 2013}; ^{Ibarra and Trupkin, 2016}; ^{Ndoricimpa, 2017}). At first, it would be tempting to conclude from these results that disinflationary policy is costless in terms of output and employment losses, and that restrictive demand-side measures could be justified to reduce the welfare costs of inflation (see ^{Thirlwall, 1974b} and ^{Briault, 1995} for more on the welfare costs of inflation).

Such an analysis, however, disregards the underlying reason why the threshold inflation rates of 6%-21% in developing countries are much higher than the 1%-5% identified in developed countries. As emphasised throughout this paper, with specific reference to ^{Thirlwall’s (1974b)} ISGD book, inflation thresholds in developing countries would tend to be higher than developed countries because structural sources of inflation (supply bottlenecks between expanding and contracting sectors, foreign exchange bottlenecks, an underdeveloped agriculture sector and a widening budget deficit from autonomous increases in food prices) are more dominant in underdeveloped economies. The overview of the empirical evidence for African countries in section 2 suggests that structural and cost-push sources of inflation tend to dominate monetary factors (^{Nell, 2004}, ²⁰¹⁸; ^{Heintz and Ndikumana, 2011}; ^{Durevall, Loening, and Yohannes, 2013}; ^{Nguyen et al., 2017}; ^{Ndikumana et al., 2023}). In addition, there is growing evidence that the money supply in developing economies is endogenously determined (^{Vera, Guijarro, and Pérez, 2022})³. It is important to reiterate that money supply endogeneity does not necessarily mean that monetary policy is ineffective in controlling inflation, but rather that Central Banks may choose to accommodate structural and cost inflation, otherwise slow growth and high unemployment may well be the inevitable outcomes⁴.

^{Khan and Senhadji (2001)} recognize the relevance of a traditional ‘structural’ source of inflation -i.e. inelasticities in the tax system and the need to revert to an inflation tax- as an explanation for the higher threshold values observed in underdeveloped economies⁵. Generally, however, studies tend to emphasise conventional explanations such as differences in institutional quality and institutional arrangements (e.g. the degree of central bank independence) in developed relative to developing economies, and the absence of well-developed financial intermediaries in the latter to transmit monetary policy shocks in an effective way. (^{Durevall, Loening, and Yohannes, 2013}; ^{Mishra and Montiel, 2013}; ^{Seleteng, Bittencourt, and van Eyden, 2013}; ^{Ibarra and Trupkin, 2016}; ^{Nguyen et al., 2017}; ^{Bhattacharya and Jain, 2020}).

But the more fundamental policy issue is to recognize that structural inflation and, to a lesser extent, cost inflation, are the natural and unavoidable outcomes of the growth and development process in developing economies. To reduce structural inflation with restrictive demand-side policy measures is to impair growth, even in developing countries where the inflation effect is insignificant below the threshold. The only way to squeeze structural inflation out of the system without harming growth is to design measures that directly address the rigidities in the real economy.

From the above, a useful and policy relevant way of interpreting the relatively high inflation thresholds of 6%-21% in developing countries is the following. For illustrative purposes only, assume that inflation of up until 7% originates from structural and cost-push sources. Superimposing demand inflation of 7% gives a threshold inflation of 14%, which is roughly the midpoint of the actual threshold range identified by the studies in Table 1. In some of the studies reviewed above, the demand component of inflation dominates, giving a positive inflation effect up until the threshold. In other studies, with different samples of developing countries and time periods, the positive growth effect of demand inflation is obscured by structural and cost-push forces, hence the insignificant growth effect up until a relatively high threshold.

Finally, the results in Table 1 show that the effect of inflation above all the threshold values identified for developed and developing countries is unambiguously negative and statistically significant. Using the insights in ^{Thirlwall’s (1974b)} ISGD book, the unambiguous nature of the results above the thresholds in developing economies is not surprising. We know that the growth benefits come from demand inflation. However, below the threshold values in developing economies the results may be more mixed because the growth stimulus of demand inflation is masked by structural and cost inflation. In contrast, above the thresholds all types of inflation are harmful to growth, including demand inflation.

3.2. Investment in physical capital as the operative channel

The models in ^{Thirlwall’s (1974b)} ISGD book generally emphasise the importance of investment in physical capital and how the costs and benefits of inflation on growth operate via the investment channel. Nevertheless, the vast majority of studies in Table 1 include either total gross investment or gross fixed investment as a control variable. Column 4 in Table 1 shows that in most cases the investment variable is positive and statistically significant, implying that inflation affects the productivity or efficiency of investment⁶. Or put in another way, inflation affects growth through a productivity channel rather than an investment channel. On the surface, these findings would seem to contradict the theoretical models in ISGD with their emphasis on a direct investment channel.

Nevertheless, it is easy to reconcile the findings in Table 1 with theoretical models that predict a direct effect of inflation on investment. It is plausible to assume that a rise in the nominal rate of return on physical capital from mild demand inflation, as well as the redistribution of income from wage earners to profit earners in the Kaldor-Thirlwall model and from the private sector to the government sector in the inflation tax model involve a change in the composition of investment. In the case of total gross investment, mild inflation in these models implies a shift out of residential investment and inventories into productive structures (infrastructure), machinery and plant. Thus, inflation raises the efficiency of investment by changing its composition rather than the total quantity of investment. In this way, the efficiency channel also captures the investment channel, as envisaged by the theoretical models in ISGD. The opposite occurs when inflation becomes ‘excessive’. In this case, there would be shifts out of productive machinery and plant into inventories, unproductive structures and residential investment (see ^{Feldstein, 1983}; ^{Temple, 2000}; ^{Madsen, 2003}).

Some empirical support for the proposition that inflation changes the composition, rather than the quantity of investment, can be found in ^{Ghosh and Phillips (1998)} and ^{Wallich (1969)}. They find that the effect of inflation on growth remains virtually unchanged, irrespective of whether investment is included in the regression or not. Nevertheless, in line with ^{Thirlwall’s (1974b)} emphasis on the investment channel, it would be informative to explicitly examine the linear and non-linear effects of inflation on total investment and its different sub-components for future research purposes. Although a more in-depth empirical analysis of the investment channel has already been identified by ^{Temple (2000)} as an important research area, studies still tend to focus on the direct effect of inflation on GDP per capita growth, notwithstanding some notable exceptions (see e.g.^{Fischer, 1993}; ^{Madsen, 2003}; ^{Barro, 2013}).