Abstract
The equation of exchange PQ = vM is one equation in four variables, and velocity v is defined by the identity. It becomes a theory of inflation only after closure restrictions are imposed. Constant velocity, output fixed by real factors, and exogenous money form one such closure; other monetary theories of inflation use other closures. Cost of production is absent because the equation reads nominal value from the expenditure side. The income side gives P = (1+k)\,\mathit{ULC}, an identity when k is the residual markup on unit labor cost ( ULC), and a pricing hypothesis only when wages, productivity and k are specified independently. With endogenous money, M = \alpha + \beta Y, and an independent specification of velocity, nominal income is Y^{*} = v\alpha/(1 - v\beta), which depends jointly on \alpha, \beta and v. If money accommodates independently determined income, the exchange identity supplies no additional determination of prices. At \alpha = 0, positive income requires v\beta = 1, and the nominal level is indeterminate. The aggregate identity does not independently determine sectoral relative prices (such as the producer price index components, energy, food, and other goods and services); an input-output price system does, given technical coefficients, wages, import prices and markups, and a stated bridge carries its gross-output prices to the GDP deflator.
1. Identity and Theory
The equation of exchange appears in textbooks as a theory of the price level:
\quad (1)\qquad PQ = vM
P is an aggregate price index, Q an aggregate output index, M the money stock, and v velocity. With the normalization of Section 2, PQ = Y in levels. Dividing by Q gives P = v(M/Q). If we write m = M/Q, the one-period change appears to be \Delta P = v\,\Delta m. The postulate looks simple, but three questions follow. Is the pair (P, Q) circular, given that each index is weighted by the other? Under what conditions is the increment formula exact? And how can an equation without cost of production or services claim to explain inflation?
This post addresses each question with algebra alone. It then introduces an endogenous money supply, M = \alpha + \beta Y, and asks whether equation (1) admits a sectoral form P_j for energy, food, and other components. The purpose is neither to defend nor to reject the quantity theory. The purpose is to separate what the identity contains from what the closure assumptions add.
2. The Index Numbers Are Not Circular
Nominal output Y = \sum p_i q_i is measured directly. The price and quantity indexes do not determine Y; they factor it. Take the Paasche price index and the Laspeyres quantity index between periods 0 and 1:
\begin{aligned}
&(2)\quad
P = \frac{\sum p_1 q_1}{\sum p_0 q_1},
\qquad
Q = \frac{\sum p_0 q_1}{\sum p_0 q_0}
\\[8pt]
&\phantom{(2)}\quad
P \cdot Q = \frac{\sum p_1 q_1}{\sum p_0 q_0} = \frac{Y_1}{Y_0}
\end{aligned}The product equals the value ratio, without approximation (Appendix A). Equation (2) is a ratio, while equation (1) is written in levels. The two are connected by a normalization: set the base-period price index to one, P_0 = 1, and measure output in base-period prices, Q_1 = Y_0 \cdot (\sum p_0 q_1 / \sum p_0 q_0) = \sum p_0 q_1. Then P_1 Q_1 = Y_1, and equation (1) holds in levels.
The Paasche-price, Laspeyres-quantity pair gives an exact factorization, but its two members use different formulas. The factor-reversal test, in its conventional sense, applies one formula to prices and the same formula, with prices and quantities interchanged, to quantities, and requires the product to equal the value ratio. The Fisher indexes, geometric means of the Laspeyres and Paasche formulas, satisfy that test (Fisher 1922; BEA, “Fisher Index”). Chained Fisher indexes preserve the product property link by link, and the national accounts compute real GDP and its price index with that formula (BEA, “Chain-Type Indexes”). The GDP deflator is the implicit index P = Y/Q. A separate property, theoretical exactness, is the result of Diewert (1976): the Fisher index is exact for a homogeneous quadratic aggregator function. The argument here requires only the algebraic factorization.
The weights create an impression of circularity, but no variable is solved in terms of itself. Nominal value is the observed magnitude; P and Q are a decomposition of it, unique up to the choice of index formula and normalization. The left side of equation (1) is therefore well defined. The difficulty lies on the right side.
3. From Identity to Theory
Velocity has no independent measurement. It is defined as v \equiv PQ/M, so equation (1) holds for every data set. With m = M/Q, the exact one-period change is (Appendix B):
\quad
(3)\qquad
\Delta P_t = v_t\,\Delta m_t + m_{t-1}\,\Delta v_tThe postulate \Delta P = v\,\Delta m therefore requires \Delta v = 0. For positive variables the log decomposition is exact:
\quad (4)\qquad \Delta \ln P = \Delta \ln v + \Delta \ln M - \Delta \ln Q
In finite growth rates the exact relation is multiplicative (Appendix B):
\quad
(5)\qquad
1 + \pi = \frac{(1 + g_v)(1 + g_M)}{1 + g_Q}The additive form \pi \approx g_v + g_M - g_Q is its first-order approximation.
A theory of inflation emerges only after three restrictions are added to the identity:
- Constant velocity: g_v = 0. This is a behavioral hypothesis about money demand. Velocity that is exogenous but variable would leave g_v in equation (5); the reduction below requires constancy.
- Real output: Q is fixed at a full-employment level, independent of M.
- Exogenous money: the monetary authority sets M, and causation runs from M to P.
Under all three, equation (5) gives \pi = (g_M - g_Q)/(1 + g_Q), and \pi \approx g_M - g_Q to first order (Friedman 1956; Humphrey 1974). These restrictions describe one closure. Monetary theories of inflation that admit variable velocity, or output that responds to money in the short period, rest on other closures, and the algebra of this post does not dispose of them. If M is the monetary base rather than a broad aggregate, v also absorbs the ratio of broad money to the base. The contraction of base velocity after the reserve expansion that followed 2008 is the standard evidence against a constant velocity for the base.
4. Where Cost of Production Enters
The equation of exchange contains no cost term because it reads nominal value from the expenditure and money side. National accounting reads the same nominal value from the income side. Let w be the wage rate and L employment, so that wL is total labor compensation, and let S \equiv Y - wL be the non-labor residual. Then Y = wL + S holds exactly. For GDP the residual S is not operating profit: it comprises taxes on production and imports less subsidies, net operating surplus, and consumption of fixed capital (BEA, “Income Approach”). Measured GDP and GDI also differ by a statistical discrepancy (BEA, “Why Do GDP and GDI Differ?”), so the identity holds exactly for GDI and for GDP up to that discrepancy. Define the residual markup on labor cost:
\quad
(6)\qquad
k \equiv \frac{Y - wL}{wL} \;\Rightarrow\; P = \frac{Y}{Q} = (1+k)\,\frac{w}{Q/L} = (1+k)\cdot \mathit{ULC}With this definition, equation (6) is an identity. It becomes a pricing hypothesis only when w, labor productivity Q/L and k are specified independently of P. The residual k should not be identified with Kalecki’s markup on unit prime costs: it is measured over labor cost alone, and it contains depreciation, indirect taxes and the surplus of every sector, while Kalecki’s markup applies to wages and materials at the level of the firm or industry (Kalecki 1971, 43–61).
Equations (1) and (6) describe the same Y, so they can be set equal:
\quad (7)\qquad vM = (1+k)\,wL
The algebra cannot decide between a money theory and a cost theory of the price level, because both expressions hold at once. The disagreement concerns closure: which variables are exogenous, and which variable absorbs the residual.
| Closure | Exogenous | Residual | Price determination |
|---|---|---|---|
| Quantity theory | M, v, Q | w or k | P = vM/Q |
| Cost closure | w, k, productivity Q/L | M, then v | P = (1+k)\cdot\mathit{ULC} |
In the cost closure, k is a behavioral markup specified in advance, not the residual of equation (6). The direction of causation is an empirical question about which closure is valid (Kalecki 1971; Kaldor 1982). The next section makes the money side of the second closure explicit.
5. Endogenous Money: M = \alpha + \beta Y
Let the money stock respond to nominal income, with \alpha an autonomous component and \beta the response of bank deposits to nominal income. Bank lending creates deposits, which are part of the broad aggregate; it does not create central-bank reserves directly (McLeay, Radia and Thomas 2014). The autonomous component need not be a policy instrument; it collects the part of the money stock that is independent of current income. If M is read as the monetary base, the same equation is an accommodation rule of the central bank, not a description of bank credit.
The money equation and the identity form two equations in Y and M. Velocity must be specified independently, and here v is a parameter:
\quad (8)\qquad Y = vM, \qquad M = \alpha + \beta Y
Substitution gives the solution for nominal income and the price level (Appendix C):
\quad
(9)\qquad
Y^{*} = \frac{v\alpha}{1 - v\beta},
\qquad
P^{*} = \frac{v\alpha}{(1 - v\beta)\,Q},
\qquad
v\beta \neq 1For the economically intended interior solution we maintain v > 0, \alpha > 0, \beta \ge 0 and v\beta < 1; then Y^{*} > 0 and M^{*} > 0. Define \theta \equiv \beta Y/M, the share of the money stock that responds to income. At the solution, v = Y/M gives v\beta = \theta and 1 - v\beta = \alpha/M, so 0 \le \theta < 1.
Nominal income depends jointly on \alpha, \beta and v. The affine money equation determines nominal income only when supplemented by an independent specification of velocity. If money accommodates independently determined income, the exchange identity supplies no additional determination of prices.
5.1 Limiting cases
- \beta = 0 (exogenous money): Y^{*} = v\alpha. The textbook quantity theory is recovered, with M = \alpha.
- \alpha = 0 (full accommodation): substitution gives (1 - v\beta)\,Y = 0. If v\beta \neq 1, the only algebraic solution is Y = M = 0. For positive income, consistency requires v\beta = 1; velocity is then 1/\beta, the system holds for every Y, and the nominal level is indeterminate. An additional nominal determination must close the system. Cost pricing, equation (6) with w, Q/L and k specified, is one such determination; it is not a logically necessary one.
The approach to the second case is singular. As \alpha \to 0, the limit of Y^{*} = v\alpha/(1 - v\beta) depends on how v\beta approaches one, and it may be zero, finite or unbounded. The second case states the endogenous-money position in algebraic form: when money accommodates in full, the identity carries no causal content for the price level.
5.2 Comparative statics under the money closure
With \alpha, \beta and v positive, \ln Y^{*} = \ln v + \ln \alpha - \ln(1 - v\beta), and the elasticities of nominal income are:
| Parameter | Elasticity of Y* | Expressed in observables |
|---|---|---|
| \alpha (autonomous money) | 1 | 1 |
| v (velocity) | 1/(1-\theta) | M/\alpha |
| \beta (accommodation) | \theta/(1-\theta) | \beta Y/\alpha |
The elasticity in \beta requires \beta > 0; \beta = 0 is a boundary case. In differential form the decomposition is exact; for discrete growth rates it is a local approximation:
\begin{aligned}
&(10)\quad
d\ln P^{*} = d\ln\alpha + \frac{d\ln v}{1-\theta} + \frac{\theta}{1-\theta}\,d\ln\beta - d\ln Q,
\\[8pt]
&\phantom{(10)}\quad
\pi \approx g_{\alpha} + \frac{g_v}{1-\theta} + \frac{\theta}{1-\theta}\,g_{\beta} - g_Q
\end{aligned}At \theta = 0, equation (10) reduces to \pi \approx g_M + g_v - g_Q. As \theta approaches one, the multiplier 1/(1-\theta) increases without bound, and small variations in velocity or in the accommodation coefficient produce large movements in nominal income. The autonomous component retains a unit elasticity, but v and \beta determine Y^{*} as much as \alpha does. The solution does not isolate a quantity-theory channel in \alpha.
5.3 Velocity as a residual under the cost closure
The cost closure is an alternative to the closure of Section 5.2, and the comparative statics of the two should not be combined. There v is a parameter and Y is endogenous; here Y is determined by equation (6) with w, Q/L, k and Q given, and v becomes the residual:
\begin{aligned}
&(11)\quad
v = \frac{Y}{\alpha + \beta Y},
\qquad
\frac{\partial v}{\partial Y} = \frac{\alpha}{(\alpha + \beta Y)^2} > 0,
\\[8pt]
&\phantom{(11)}\quad
\frac{\partial \ln v}{\partial \ln Y} = 1 - \theta = \frac{\alpha}{M}
\end{aligned}Velocity rises with nominal income whenever \alpha > 0, and its elasticity equals the autonomous share of money. A procyclical velocity is thus consistent with the cost closure and does not require a shift in money demand as explanation.
5.4 Counting equations and estimation
The system has three unknowns (Y, M, v) and admits three independent equations: the identity, the money equation, and one closure. The closure is either v = \bar v (quantity theory) or equation (6) with its terms specified (cost theory), not both. Imposing both overdetermines the system, unless k or w adjusts as a residual.
Equation counting does not establish the properties of an estimator. Once the money equation is written with a disturbance, M_t = \alpha + \beta Y_t + u_t, consistent least-squares estimation of \beta requires that Y_t be orthogonal to u_t in the limit, together with the usual regularity conditions. A cost closure does not by itself establish them. Conversely, causation from M to Y does not by itself prove simultaneity bias. Common shocks, feedback between money and income, measurement error, and nonstationarity each require separate treatment, through instruments, cointegration methods or estimation of the full system (Davidson and MacKinnon 2004).
5.5 Dynamics
If money responds with a one-period lag, M_t = \alpha + \beta Y_{t-1}, nominal income follows a first-order difference equation. For v\beta \neq 1 (Appendix D):
\quad
(12)\qquad
Y_t = v\alpha + v\beta\,Y_{t-1},
\qquad
Y_t - Y^{*} = (v\beta)^t\,(Y_0 - Y^{*})Stability requires |v\beta| < 1, and convergence is monotone when 0 < v\beta < 1. In the lagged model the responsive share of money is \theta_t \equiv \beta Y_{t-1}/M_t = v\beta\,Y_{t-1}/Y_t, which equals v\beta at a positive steady state but not during adjustment.
At v\beta = 1 there is no finite fixed point when \alpha > 0, and the deviation formula does not apply. The difference equation becomes:
\quad
(13)\qquad
Y_t = Y_{t-1} + v\alpha \;\Rightarrow\; Y_t = Y_0 + t\,v\alphaThis is deterministic linear growth. With an additive disturbance \varepsilon_t, Y_t = Y_{t-1} + v\alpha + \varepsilon_t is a random walk with drift v\alpha; a driftless random walk additionally requires \alpha = 0.
6. Disaggregation: P_j for Energy, Food, and Other Components
6.1 What the aggregate identity determines
The equation of exchange constrains the scalar Y and, given Q, the scalar P. It does not independently determine sectoral relative prices. That absence follows from the identity having a single equation; it is not by itself the classical dichotomy, which is the hypothesis that real factors determine relative prices and money only their level.
Sectoral spending can be decomposed without commodity-specific money stocks. Let s_j = P_j Q_j / Y be the expenditure share of component j:
\quad (14)\qquad P_j Q_j = s_j\,vM, \qquad \sum_j s_j = 1
Equation (14) is an accounting decomposition. It requires no sectoral money stock, but it adds no independent price theory, since the shares s_j and the split of each P_j Q_j into price and quantity are determined elsewhere. Endogenous money does not alter this: equation (8) adds equations for the scalars Y and M, and none for relative prices.
For a fixed-basket illustration, the Laspeyres price index between periods 0 and 1 is a weighted sum of component price relatives, with base-period expenditure shares as weights:
\quad
(15)\qquad
P^{L} = \sum_j s_{j0}\,\frac{p_{j1}}{p_{j0}}A chained Fisher index, which the national accounts use, is not in general a weighted sum of component indexes of this form, and aggregation of components must then follow the chain formula itself.
Consider an energy price increase with M and v fixed. Nominal spending vM is unchanged, so a higher energy price must coincide with lower prices elsewhere, lower real output, or both; a higher energy price and lower real output can coexist. If nominal spending accommodates the shock, the accommodation runs through v or M.
6.2 The input-output price system
The cost-side price system determines sectoral prices. Let a_{ij} be the domestic input of good i per unit of good j, e_j the imported energy per unit of j at import price p_e, h_j the direct labor hours per unit of j, w the wage rate, and k_j the sectoral markup on current cost:
\quad
(16)\qquad
p_j = (1 + k_j)\Big(\sum_i p_i\,a_{ij} + p_e\,e_j + w\,h_j\Big)In row-vector form, with K = \operatorname{diag}(k_j) and e, h the row vectors of energy and labor coefficients (Appendix E):
\quad
(17)\qquad
p = (p_e\,e + w\,h)\,(I + K)\,\big[I - A(I + K)\big]^{-1}If A(I + K) is non-negative with spectral radius below one, the inverse exists and is non-negative. Strictly positive prices additionally require that each sector have positive direct cost, p_e e_j + w h_j > 0, or draw, directly or through other sectors, on sectors that do. An imported-energy price shock enters through p_e. It reaches sector j only if j uses energy directly or through connected inputs; the shock reaches connected sectors, not necessarily every sector. A change in e_j or a_{ij} represents a change in input requirements, not a price shock. Pasinetti (1977) presents the input-output price system with a uniform rate of profit; the sector-specific markup on current cost, K = \operatorname{diag}(k_j), is the specification adopted here.
6.3 From gross-output prices to the GDP deflator
Equation (17) determines prices of gross output, while GDP concerns final expenditure, equivalently value added. A bridge is required. Let f_t be the column vector of final demand for domestic output in period t, and x_t = (I - A)^{-1} f_t the gross output that supplies it. Value added is gross output less domestic and imported intermediate inputs, so (Appendix E):
\quad
(18)\qquad
Y_t = p_t f_t - p_{e,t}\,e\,x_t,
\qquad
P_t = \frac{p_t f_t - p_{e,t}\,e\,x_t}{p_0 f_t - p_{e,0}\,e\,x_t}The deflator in equation (18) is a double-deflated Paasche index of value added. An imported-energy price shock raises the prices of domestic final demand through equation (17), while the import deduction offsets part of the increase in the GDP deflator. The deflator of domestic final demand and the GDP deflator therefore diverge. Money enters this system only as the financing condition M = \alpha + \beta Y, with Y from equation (18).
7. Conclusion
The equation of exchange is an identity over measured nominal value, and its index numbers are not circular once their normalization is stated. It becomes a theory of inflation through closure restrictions, not through its algebra; constant velocity, fixed full-employment output and exogenous money form one closure among several. Cost of production is absent from it by construction. The income side gives an identity of equal standing, P = (1+k)\cdot\mathit{ULC} with k the residual markup, which becomes a pricing hypothesis only when its terms are specified independently.
With endogenous money, M = \alpha + \beta Y, and an independently specified velocity, nominal income depends jointly on \alpha, \beta and v, and its sensitivity to v and \beta grows with the responsive share \theta. If money accommodates independently determined income, the identity supplies no additional determination of prices. At \alpha = 0, positive income requires v\beta = 1 and leaves the nominal level indeterminate, to be fixed by some other nominal determination, of which cost pricing is one. Finally, the identity does not independently determine sectoral relative prices. The input-output price system (17) determines gross-output prices, and the bridge (18) carries them to the GDP deflator.
Appendix A. Exact Factorization and Factor Reversal
The Paasche price index and the Laspeyres quantity index share the term \sum p_0 q_1, which cancels:
\quad
(\mathrm{A}.1)\qquad
P^{P} Q^{L} = \frac{\sum p_1 q_1}{\sum p_0 q_1} \times \frac{\sum p_0 q_1}{\sum p_0 q_0} = \frac{\sum p_1 q_1}{\sum p_0 q_0} \equiv VThe Laspeyres price index times the Paasche quantity index gives V through the term \sum p_1 q_0. The Fisher indexes are P^{F} = (P^{L} P^{P})^{1/2} and Q^{F} = (Q^{L} Q^{P})^{1/2}, so:
\quad
(\mathrm{A}.2)\qquad
P^{F} Q^{F} = \big(P^{L} Q^{P} \cdot P^{P} Q^{L}\big)^{1/2} = (V \cdot V)^{1/2} = VQ^{F} is the formula of P^{F} with prices and quantities interchanged, so the Fisher pair satisfies factor reversal in the conventional sense. The Paasche-Laspeyres pair satisfies the product property without that symmetry. For levels, with P_0 = 1 and Q_1 = \sum p_0 q_1, the product P_1 Q_1 = \sum p_1 q_1 = Y_1.
Appendix B. The Exact Increment and Finite Growth
With P_t = v_t m_t, add and subtract v_t m_{t-1}:
\begin{aligned}
&(\mathrm{B}.1)\quad
P_t - P_{t-1} = v_t m_t - v_t m_{t-1} + v_t m_{t-1} - v_{t-1} m_{t-1}
\\[8pt]
&\phantom{(\mathrm{B}.1)\quad P_t - P_{t-1}}
= v_t\,\Delta m_t + m_{t-1}\,\Delta v_t
\end{aligned}The alternative grouping gives v_{t-1}\Delta m_t + m_t \Delta v_t. Both are exact, and both reduce to v\,\Delta m only when \Delta v = 0. For finite growth, P_t/P_{t-1} = (v_t/v_{t-1})(M_t/M_{t-1})/(Q_t/Q_{t-1}), which is equation (5). With g_v = 0:
\quad
(\mathrm{B}.2)\qquad
\pi = \frac{1 + g_M}{1 + g_Q} - 1 = \frac{g_M - g_Q}{1 + g_Q}Appendix C. Solution and Elasticities under Endogenous Money
Substitute M = \alpha + \beta Y into Y = vM:
\begin{aligned}
&(\mathrm{C}.1)\quad
Y = v\alpha + v\beta Y \;\Rightarrow\; (1 - v\beta)\,Y = v\alpha
\\[8pt]
&\phantom{(\mathrm{C}.1)}\quad
\;\Rightarrow\; Y^{*} = \frac{v\alpha}{1 - v\beta}, \quad v\beta \neq 1
\end{aligned}At \alpha = 0 the middle expression is (1 - v\beta)Y = 0: either v\beta = 1 and Y is free, or Y = 0 and then M = \beta Y = 0. With v = Y/M and \theta = \beta Y/M, we have v\beta = \theta, and from M = \alpha + \beta Y we have \alpha/M = 1 - \theta. For positive \alpha, \beta and v, take \ln Y^{*} = \ln v + \ln \alpha - \ln(1 - v\beta) and differentiate:
\begin{aligned}
&(\mathrm{C}.2)\quad
\frac{\partial \ln Y^{*}}{\partial \ln v} = 1 + \frac{v\beta}{1 - v\beta} = \frac{1}{1-\theta},
\\[8pt]
&\phantom{(\mathrm{C}.2)}\quad
\frac{\partial \ln Y^{*}}{\partial \ln \alpha} = 1,
\\[8pt]
&\phantom{(\mathrm{C}.2)}\quad
\frac{\partial \ln Y^{*}}{\partial \ln \beta} = \frac{v\beta}{1 - v\beta} = \frac{\theta}{1-\theta}
\end{aligned}Since \ln P^{*} = \ln Y^{*} - \ln Q, the differential form of equation (10) follows by total differentiation; replacing differentials by discrete growth rates gives the local approximation. For residual velocity, v = Y/(\alpha + \beta Y) gives \ln v = \ln Y - \ln(\alpha + \beta Y), so:
\quad
(\mathrm{C}.3)\qquad
\frac{\partial \ln v}{\partial \ln Y} = 1 - \frac{\beta Y}{\alpha + \beta Y} = 1 - \theta = \frac{\alpha}{M}Appendix D. The Lagged System
For v\beta \neq 1 the fixed point satisfies Y^{*} = v\alpha + v\beta Y^{*}. Subtract it from Y_t = v\alpha + v\beta Y_{t-1}:
\quad
(\mathrm{D}.1)\qquad
Y_t - Y^{*} = v\beta\,(Y_{t-1} - Y^{*}) \;\Rightarrow\; Y_t - Y^{*} = (v\beta)^t (Y_0 - Y^{*})The deviation decays if |v\beta| < 1, monotone when v\beta > 0. Since Y_t = v M_t, the responsive share is \theta_t = \beta Y_{t-1}/M_t = v\beta\,Y_{t-1}/Y_t, equal to v\beta only when Y_t = Y_{t-1}. At v\beta = 1, recursion from Y_0 gives:
\quad
(\mathrm{D}.2)\qquad
Y_t = Y_0 + t\,v\alpha + \sum_{s=1}^{t} \varepsilon_swith \varepsilon_s = 0 in the deterministic case. The sum of disturbances is the random-walk component, and t\,v\alpha is its drift.
Appendix E. The Input-Output System and the GDP Bridge
The sectoral equations (16) in row-vector form are p = (pA + p_e e + wh)(I + K). Collect the terms in p:
\begin{aligned}
&(\mathrm{E}.1)\quad
p - pA(I + K) = (p_e e + w h)(I + K)
\\[8pt]
&\phantom{(\mathrm{E}.1)}\quad
\;\Rightarrow\; p = (p_e e + w h)(I + K)\big[I - A(I + K)\big]^{-1}
\end{aligned}By the Perron–Frobenius theorem, the inverse exists and is non-negative when A(I + K) is non-negative with spectral radius less than one. The power series [I - A(I + K)]^{-1} = I + A(I + K) + [A(I + K)]^2 + \cdots traces the rounds of cost pass-through, which travel only along non-zero entries of A.
For the bridge, gross output satisfies x = Ax + f. The value of gross output less domestic intermediate inputs is px - pAx = p(I - A)x = pf. Subtracting imported intermediate inputs p_e e x gives value added:
\quad
(\mathrm{E}.2)\qquad
Y = px - pAx - p_e e x = pf - p_e e xReal value added in base-period prices is p_0 f_t - p_{e,0} e x_t, and the ratio of the two is the implicit deflator of equation (18).
References
Bureau of Economic Analysis. n.d. “Chain-Type Indexes”; “Fisher Index”; “Income Approach.” BEA Glossary. bea.gov/help/glossary. Defines the chained Fisher formula used for real GDP and its price index, and the income-side components of GDP (compensation, taxes on production and imports less subsidies, net operating surplus, consumption of fixed capital) that make up wL + S in Section 4.
Bureau of Economic Analysis. n.d. “Why Do Gross Domestic Product (GDP) and Gross Domestic Income (GDI) Differ, and What Does That Imply?” BEA FAQ 484. bea.gov/help/faq/484. Explains the statistical discrepancy between the expenditure and income measures, which qualifies the exactness of the income identity for measured GDP.
Davidson, R., and J. G. MacKinnon (2004). Econometric Theory and Methods. New York: Oxford University Press. Conditions for consistency of least squares, simultaneity, instrumental variables and nonstationary regressors; the reference for Section 5.4.
Diewert, W. E. (1976). Exact and superlative index numbers. Journal of Econometrics, 4(2), 115–145. Shows that the Fisher index is exact for a homogeneous quadratic aggregator function and defines the class of superlative indexes. Its theoretical exactness is distinct from the algebraic factor reversal used in Section 2.
Fisher, I. (1911). The Purchasing Power of Money. New York: Macmillan. States the equation of exchange, with the volume of transactions in place of output, and attributes velocity to institutional and payment habits. It is the source of the identity examined here.
Fisher, I. (1922). The Making of Index Numbers. Boston: Houghton Mifflin. Formulates the tests an index number should satisfy, among them factor reversal (one formula applied to prices and, with prices and quantities interchanged, to quantities), and identifies the ideal index as the geometric mean of the Laspeyres and Paasche formulas. It is the basis of Appendix A.
Friedman, M. (1956). The quantity theory of money: A restatement. In M. Friedman (ed.), Studies in the Quantity Theory of Money, 3–21. Chicago: University of Chicago Press. Recasts the quantity theory as a theory of the demand for money, a stable function of a few variables. It is the modern source of restriction 1; the restatement leaves the supply of money and the determination of output to separate hypotheses, which is why restrictions 2 and 3 must be added.
Humphrey, T. M. (1974). The quantity theory of money: Its historical evolution and role in policy debates. Federal Reserve Bank of Richmond Economic Review, 60(3), 2–19. A secondary survey that traces the quantity theory from its seventeenth- and eighteenth-century origins to the monetarist debates, and sets out five propositions (proportionality, the causal role of money, neutrality, the monetary theory of the price level, and exogeneity of the nominal money stock) that convert the identity into a theory. It serves as the reference for the closure restrictions of Section 3.
Kaldor, N. (1982). The Scourge of Monetarism. Oxford: Oxford University Press. Argues that in a credit economy the money stock adjusts to the demand for loans, so that causation runs from nominal income to money. The British monetarist experiment of 1979–1981 is its empirical setting; the endogenous-money equation of Section 5 formalizes its position.
Kalecki, M. (1971). Selected Essays on the Dynamics of the Capitalist Economy, 1933–1970. Cambridge: Cambridge University Press. Chapter 5, “Costs and Prices” (43–61), sets the price of manufactures as unit prime cost, wages and materials, multiplied by a markup that reflects the degree of monopoly. It is the source of the cost closure, not of the aggregate income identity (6), whose residual markup is a different magnitude.
McLeay, M., A. Radia, and R. Thomas (2014). Money creation in the modern economy. Bank of England Quarterly Bulletin, 54(1), 14–27. Explains that commercial-bank lending creates deposits and that reserves do not constrain lending through a fixed multiplier; the basis for reading \beta as the response of broad money in Section 5.
Pasinetti, L. L. (1977). Lectures on the Theory of Production. New York: Columbia University Press. Presents the Leontief and Sraffa price systems in matrix form, with a uniform rate of profit, the conditions for a positive solution, and the construction of vertically integrated sectors. It covers the structure of equations (16) and (17); the sector-specific markups are the specification adopted here.