Abstract
The measurement of productivity change (or difference) is usually based on models relying on strong assumptions such as competitive behaviour and constant returns to scale. This chapter discusses the basics of productivity measurement and shows that one can dispense with most if not all of the usual, neo-classical assumptions. Various models are reviewed and their relationships discussed. Throughout the chapter the equivalence of multiplicative and additive models, as well as the equivalence of productivity measurement and growth accounting, is highlighted. By virtue of their structural features, the various measurement models are applicable to individual establishments and aggregates such as industries, sectors, or economies.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The neo-classical model figured already prominently in the 1979 report of the U. S. National Research Council’s Panel to Review Productivity Statistics (Rees 1979). An overview of national and international practice is provided by the regularly updated OECD Compendium of Productivity Indicators (OECD 2019).
- 2.
There is another, minor, difference between the approach defended here and the usual story. The usual story runs in the framework of continuous time in which periods are of infinitesimal short duration. When it then comes to implementation several approximations must be assumed. The approach in this book does not need this kind of assumptions either, because entirely based on accounting periods of finite duration, such as years.
- 3.
A classic treatment in the neo-classical framework, based on the distinction between shortrun and longrun output or cost, is provided by Morrison Paul (1999). For a modern treatment (with keywords: representative firm, dynamic cost minimization, Cobb-Douglas production function with constant returns to scale, Hicks-neutral technological change) the reader is referred to Comin et al. (2020).
- 4.
The utilization rate of the labour input factors is assumed to be 1. Over- or underutilization from the point of view of jobs or persons is reflected in the wage rates. At the economy level, unutilized labour is called ‘unemployment’.
- 5.
- 6.
Sometimes zero profit is imposed by considering profit as the remuneration (price) for entrepreneurial activity (of which the quantity is set equal to one), and adding this to business services S. In terms of National Accounts profit equals gross operating surplus (GOS) minus the imputed income of self-employed persons and capital input cost.
- 7.
The inclusion of natural capital (including subsoil assets) in K is only meaningful if the production units are economies; see Brandt et al. (2017).
- 8.
It is easy to see, for example, that increasing profit can occur simultaneously with decreasing profitability.
- 9.
Note that real change means nominal change deflated by some price index, not necessarily being a (headline) CPI. ‘Stripping’ is of course a vague term, and a more precise definition will be given later.
- 10.
This approach goes back to Hicks (1940).
- 11.
More on this in Diewert (2018).
- 12.
Balk (2008) provides an up-to-date treatment, including its history and all the historical references.
- 13.
Note, however, that this is not unproblematic. For instance, when the Törnqvist price index P T(.) is used, then the implicit quantity index (p 1 ⋅ y 1∕p 0 ⋅ y 0)∕P T(.) does not necessarily satisfy the Identity Test. The Identity Test for a quantity index prescribes that such an index equals unity whenever quantities have not changed.
- 14.
- 15.
See Sheng et al. (2017) for an application to Australian agriculture over the period 1949–2012.
- 16.
Note that it is not necessary to assume that R t = C t (t = 0, 1).
- 17.
Thus, saying that output growth outpaced input growth because TFP increased is “like saying that the sun rose because it was morning”, to paraphrase Friedman (1988, p. 58). Of course, when TFP change is decomposed into factors such as technological change or efficiency change, and one is able to measure such factors independently, more can be said.
- 18.
This approach follows Balk (2003b).
- 19.
For any two strictly positive real numbers a and b their logarithmic mean is defined by \(\mathit {LM}(a,b) \equiv (a - b)/\ln (a/b)\) when a ≠ b, and LM(a, a) ≡ a. It has the following properties: (1) \(\min (a,b) \leq \mathit {LM}(a,b) \leq \max (a,b)\); (2) LM(a, b) is continuous; (3) LM(λa, λb) = λLM(a, b) (λ > 0); (4) LM(a, b) = LM(b, a); (5) (ab)1∕2 ≤LM(a, b) ≤ (a + b)∕2; (6) LM(a, 1) is concave. See Balk (2008, 134–136) for details.
- 20.
This is based on the identity a 1 b 1 − a 0 b 0 = (1∕2)(a 0 + a 1)(b 1 − b 0) + (1∕2)(b 0 + b 1)(a 1 − a 0).
- 21.
Though, taken strictly, ‘multi factor’ differs from ‘total factor’, the two adjectives are generally used as synonymous.
- 22.
The notation used might be slightly confusing, yet has been chosen so as to stay in line with conventional practice. Capital L without super- or subscripts denotes the set of labour types. Capital L with a time supercript denotes the total number of labour units.
- 23.
Capital productivity is closely related to the business accounting concept ‘(total) asset turnover ratio’, defined as net sales divided by (the value of) total assets of a company. In the literature this is considered as a measure of efficiency.
- 24.
An early advocate of the value added output concept was Burns (1930). Specifically, he favoured what later in this chapter will be defined as net value added. Burns was aware of the possibility that for very narrowly defined production units and small time periods value added may sometimes become non-positive. The Covid-19 pandemic caused in economies all over the world dramatic declines of (nominal or real) value added, at various levels of aggregation, but never to the extent of becoming negative.
- 25.
In between the KLEMS-Y model and the KL-VA model figures the KLE”M”S-Margin model, applicable to distributive trade units. Here the set of material inputs M is split into two parts, M′ denoting the goods for resale and M″ the auxiliary materials. Likewise E, the set of energy inputs, is split into E′ and E″. The Margin is then defined as \(R^t - C_{E' \cup M'}^t\). See Inklaar and Timmer (2008).
- 26.
The typical magnitude of α is 0.3, as in Gordon (2016, 546).
- 27.
An example is provided by Niebel et al. (2017).
- 28.
Cash flow is also called gross or variable profit. The National Accounts concept gross operating surplus (GOS) corresponds to cash flow plus the labour cost of self-employed persons. In some sectors it occasionally occurs that production units exhibit negative cash flows during certain periods. An example of such a sector is agriculture.
- 29.
This also seems to be the model favoured by Hulten and Schreyer (2010). Anyway they consider what is here called NNVA as “entry point for measuring welfare change.”
- 30.
Rymes (1983) would single out the KL-NVA model as the “best” one, but this is clearly not backed by the argument presented here.
- 31.
- 32.
Here are some examples. The U. S. Bureau of Labor Statistics (see Eldridge et al. 2018) and Statistics Canada (see Baldwin et al. 2014) use mainly endogenous rates, except that for outcomes that are deemed unrealistic endogenous rates are replaced by more appropriate exogenous rates. Moreover, Statistics Canada replaces annual rates by 3-year moving averages. The Australian Bureau of Statistics uses, per production unit considered, the maximum of the endogenous rate and a certain exogenous rate (set equal to the annual percentage change of the CPI plus 4%) (see Roberts 2008). Statistics New Zealand (2014) uses throughout an exogenous rate, namely the annual percentage change of the CPI plus 4%. The Swiss Federal Statistical office employs an intricate system: per production unit the simple mean of the endogenous rate and a certain exogenous rate is used as the final exogenous rate (see Rais and Sollberger 2008). Concerning the endogenous rates, however, these sources are not clear as to which concept is used precisely. Statistics Netherlands sets the interest rate equal to the so-called Internal Reference Rate, which is the interest rate that banks charge to each other, plus 1.5% (see Van den Bergen et al. 2008). It seems that Jorgenson (2009) is also proposing endogenous rates of return for the four sectors considered.
- 33.
It is assumed here that the unit user costs of used and unused assets are the same. See Hulten (2010) for a brief discussion of this issue.
- 34.
There is some scattered evidence, such as Lovell and Lovell (2013), who compared ITFPROD VA and ITFPROD Y, both with Πt ≠ 0, on the Australian Coal Mining industry, 1991–2017. They found differences in line with the Domar factor.
- 35.
For an interesting corroboration of this statement the reader is referred to Gandhi et al. (2017). Staying in a strictly neo-classical framework, these authors employed plant-level Manufacturing data from Colombia (1981–1991) and Chili (1979–1996) to estimate gross-output based as well as value-added based production functions in order to obtain productivity figures. They found out that the “dispersion in productivity appears far more stable both across industries and across countries when measured via gross output as opposed to value added”, and thus that “insights derived under value added, compared to gross output, could lead to significantly different policy conclusions.”
- 36.
See Balk (1998, Section 3.7) for a formal proof.
- 37.
On stochastic productivity measurement see Chambers (2008).
- 38.
Recall that for any two strictly positive real numbers a and b their logarithmic mean is defined by \(\mathit {LM}(a,b) \equiv (a - b)/\ln (a/b)\) when a ≠ b, and LM(a, a) ≡ a.
- 39.
An interesting comparison of \(Q_{\mathit {VA}}^{dd1}(.)\) and \(Q_{\mathit {VA}}^{sd'}(.)\) on UK data was carried out by Oulton et al. (2018).
- 40.
Hence the name ‘double deflation’. This should be the standard, rather than an option.
- 41.
The first part of this Appendix is based on Balk (2009).
- 42.
Diewert (1978) showed that any superlative index is a second-order approximation of the Montgomery-Vartia index at the point where price and quantity relatives are equal to 1.
References
Baldwin, J.R., W. Gu, and B. Yan. 2007. User Guide for Statistics Canada’s Annual Multifactor Productivity Program. The Canadian Productivity Review No. 14 (Statistics Canada, www.statcan.gc.ca)
Baldwin, J.R., W. Gu, R. Macdonald, W. Wang, and B. Yan. 2014. Revisions to the Multifactor Productivity Accounts. The Canadian Productivity Review No. 35 (Statistics Canada, www.statcan.gc.ca)
Balk, B.M. 1998. Industrial Price, Quantity, and Productivity Indices: The Micro-Economic Theory and an Application. Boston/Dordrecht/London: Kluwer Academic.
Balk, B.M. 2003b. The Residual: On Monitoring and Benchmarking Firms, Industries, and Economies with Respect to Productivity. Journal of Productivity Analysis 20: 5–47. Reprinted in National Accounting and Economic Growth, ed. John M. Hartwick, The International Library of Critical Writings in Economics No. 313 (Edward Elgar, Cheltenham UK, Northampton MA, 2016).
Balk, B.M. 2008. Price and Quantity Index Numbers: Models for Measuring Aggregate Change and Difference. New York: Cambridge University Press.
Balk, B.M. 2009. On the Relation Between Gross-Output and Value-Added Based Productivity Measures: The Importance of the Domar Factor. Macroeconomic Dynamics 13 (Supplement 2): 241–267.
Balk, B.M. 2018. Empirical Productivity Indices and Indicators. In The Oxford Handbook of Productivity Analysis, ed. E. Grifell-Tatjé, C.A.K. Lovell, and R.C. Sickles. New York: Oxford University Press. Extended version available at SSRN: http://ssrn.com/abstract=2776956.
Ball, V.E., Färe, R., Grosskopf, S., and D. Margaritis. 2015. The Role of Energy Productivity in U. S. Agriculture. Energy Economics 49: 460–471.
Banerjee, A.V., and E. Duflo. 2019. Good Economics for Hard Times (PublicAffairs, New York).
Basu, S., and J.G. Fernald. 2002. Aggregate Productivity and Aggregate Technology. European Economic Review 46: 963–991.
Bennet, T.L. 1920. The Theory of Measurement of Changes in Cost of Living. Journal of the Royal Statistical Society 83: 455–462.
Brandt, N., P. Schreyer, and V. Zipperer. 2017. Productivity Measurement with Natural Capital. The Review of Income and Wealth 63, Supplement 1: S7–S21.
Burns, A.F. 1930. The Measurement of the Physical Volume of Production. Quarterly Journal of Economics 44: 242–262.
Chambers, R.G. 2008. Stochastic Productivity Measurement. Journal of Productivity Analysis 30: 107–120.
Comin, D.A., J. Quintana Gonzalez, T.G. Schmitz, and A. Trigari. 2020. Measuring TFP: The Role of Profits, Adjustment Costs, and Capacity Utilization. Working Paper 28008, National Bureau of Economic Research, Cambridge, MA.
Coremberg, A. 2008. The Measurement of TFP in Argentina, 1990–2004: A Case of the Tyranny of Numbers, Economic Cycles and Methodology. International Productivity Monitor 17: 52–74.
Dean, E.R., and M.J. Harper. 2001. The BLS Productivity Measurement Program. In New Developments in Productivity Analysis, ed. C.R. Hulten, E.R. Dean, and M.J. Harper, Studies in Income and Wealth, vol. 63. Chicago and London: The University of Chicago Press.
De Haan, M., and J. Haynes. 2018. R&D Capitalisation: Where Did We Go Wrong? Eurona — Eurostat Review on National Accounts and Macroeconomic Indicators 1/2018: 7–34.
De Haan, M., E. Veldhuizen, M. Tanriseven, and M. van Rooijen-Horsten. 2014. The Dutch Growth Accounts: Measuring Productivity with Non-Zero Profits. The Review of Income and Wealth 60: S380–S397.
Diewert, W.E. 1978. Superlative Index Numbers and Consistency in Aggregation. Econometrica 46: 883–900.
Diewert, W.E. 1992. The Measurement of Productivity. Bulletin of Economic Research 44: 163–198.
Diewert, W.E. 2008. The Bern OECD Workshop on Productivity Analysis and Measurement: Conclusions and Future Directions. In Productivity Measurement and Analysis: Proceedings from OECD Workshops, OECDiLibrary; Swiss Federal Statistical Office, Neuchâtel.
Diewert, W.E. 2018. Productivity Measurement in the Public Sector. In The Oxford Handbook of Productivity Analysis, ed. E. Grifell-Tatjé, C.A.K. Lovell, and R.C. Sickles. New York: Oxford University Press.
Diewert, W.E., and N. Huang. 2011. Capitalizing R&D Expenditures. Macroeconomic Dynamics 15: 537–564.
Diewert, W.E., and D. Lawrence. 2006. Measuring the Contributions of Productivity and Terms of Trade to Australia’s Economic Welfare. Report by Meyrick and Associates (Productivity Commission, Canberra).
Diewert, W.E., H. Mizobuchi, and K. Nomura. 2005. On Measuring Japan’s Productivity, 1955–2003. Discussion Paper No. 05-22, Department of Economics, University of British Columbia, Vancouver.
Diewert, W.E., and A.O. Nakamura. 2003. Index Number Concepts, Measures and Decompositions of Productivity Growth. Journal of Productivity Analysis 19: 127–159.
Dumagan, J.C., and V.E. Ball. 2009. Decomposing Growth in Revenues and Costs into Price, Quantity, and Total Factor Productivity Contributions. Applied Economics 41: 2943–2953.
Eldridge, L.P., C. Sparks, and J. Stewart. 2018. The US Bureau of Labor Statistics Productivity Program. In The Oxford Handbook of Productivity Analysis, ed. E. Grifell-Tatjé, C.A.K. Lovell, and R.C. Sickles. New York: Oxford University Press.
Fisher, I. 1922. The Making of Index Numbers. Boston: Houghton Mifflin.
Friedman, B.M. 1988. Lessons on Monetary Policy from the 1980s. Journal of Economic Perspectives 2(3): 51–72.
Gandhi, A., S. Navarro, and D. Rivers. 2017. How Heterogeneous is Productivity? A Comparison of Gross Output and Value Added. Working Paper 2017-27, Centre for Human Capital and Productivity, University of Western Ontario, London, Canada.
Geary, R.C. 1944. The Concept of Net Volume of Output, with Special Reference to Irish Data. Journal of the Royal Statistical Society 107: 251–261.
Gollop, F.M. 1979. Accounting for Intermediate Input: The Link Between Sectoral and Aggregate Measures of Productivity Growth. In Rees (1979).
Gordon, R.J. 2016. The Rise and Fall of American Growth: The U. S. Standard of Living Since the Civil War. Princeton and Oxford: Princeton University Press.
Grifell-Tatjé, E., and C.A.K. Lovell. 2015. Productivity Accounting: The Economics of Business Performance. New York: Cambridge University Press.
Hicks, J.R. 1940. The Valuation of Social Income. Economica 7: 105–124.
Hulten, C.R. 2010. Growth Accounting. In Handbook of the Economics of Innovation, ed. B.H. Hall, and N. Rosenberg, vol. 2. North Holland: Elsevier.
Hulten, C.R., and P. Schreyer. 2010. GDP, Technical Change, and the Measurement of Net Income: The Weitzman Model Revisited. Working Paper 16010, National Bureau of Economic Research, Cambridge, MA.
Inklaar, R. 2010. The Sensitivity of Capital Services Measurement: Measure All Assets and the Cost of Capital. The Review of Income and Wealth 56: 389–412.
Inklaar, R., and M.P. Timmer. 2008. Accounting for Growth in Retail Trade: An International Productivity Comparison. Journal of Productivity Analysis 29: 23–31.
Isaksson, A. 2009. The UNIDO World Productivity Database: An Overview. International Productivity Monitor 18: 38–50.
Jorgenson, D.W. 2009. A New Architecture for the U. S. National Accounts. The Review of Income and Wealth 55: 1–42.
Jorgenson, D.W., M.S. Ho, and K.J. Stiroh. 2005. Productivity Volume 3: Information Technology and the American Growth Resurgence. Cambridge, MA and London: The MIT Press.
Karmel, P.H. 1954. The Relations Between Chained Indexes of Input, Gross Output and Net Output. Journal of the Royal Statistical Society Series A 117: 441–458.
Lawrence, D., W.E. Diewert, and K.J. Fox. 2006. The Contributions of Productivity, Price Changes and Firm Size to Profitability. Journal of Productivity Analysis 26: 1–13.
Lovell, C.A.K., and J.E. Lovell. 2013. Productivity Decline in Australian Coal Mining. Journal of Productivity Analysis 40: 443–455.
Morrison Paul, C.J. 1999. Cost Structure and the Measurement of Economic Performance Boston/Dordrecht/London: Kluwer Academic Publishers.
Niebel, T., and M. Saam. 2016. ICT and Growth: The Role of Rates of Return and Capital Prices. The Review of Income and Wealth 62: 283–310.
Niebel, T., M. O’Mahony, and M. Saam. 2017. The Contribution of Intangible Assets to Sectoral Productivity Growth in the EU. The Review of Income and Wealth 63: Supplement 1, S49–S67.
OECD. 2019. OECD Compendium of Productivity Indicators (OECDiLibrary).
Oulton, N., A. Rincon-Aznar, L. Samek, and S. Srinivasan. 2018. Double Deflation: Theory and Practice. Discussion Paper 2018-17, Economic Statistics Centre of Excellence, London.
Rais, G., and P. Sollberger. 2008. Multi-Factor Productivity Measurement: From Data Pitfalls to Problem Solving — The Swiss Way. In Productivity Measurement and Analysis: Proceedings from OECD Workshops (OECD/Swiss Federal Statistical Office, Paris/Neuchâtel). https://doi.org/https://doi.org/10.1787/9789264044616-en
Rees, A. (Chairman). 1979. Measurement and Interpretation of Productivity. Report of the Panel to Review Productivity Statistics, Committee on National Statistics, Assembly of Behavioral and Social Sciences, National Research Council (National Academy of Sciences, Washington DC).
Roberts, P. 2008. Estimates of Industry Level Multifactor Productivity in Australia: Measurement Initiatives and Issues. In Productivity Measurement and Analysis: Proceedings from OECD Workshops (OECDiLibrary; Swiss Federal Statistical Office, Neuchâtel).
Rymes, T.K. 1983. More on the Measurement of Total Factor Productivity. The Review of Income and Wealth 29: 297–316.
Schreyer, P. 2001. Measuring Productivity: Measurement of Aggregate and Industry-Level Productivity Growth. OECD Manual (OECDiLibrary).
Schreyer, P. 2010. Measuring Multi-Factor Productivity when Rates of Return are Exogenous. In Price and Productivity Measurement, Volume 6, ed. W.E. Diewert, B.M. Balk, D. Fixler, K.J. Fox, and A.O. Nakamura. Trafford Press, www.vancouvervolumes.com, www.indexmeasures.com.
Sheng, Y., T. Jackson, S. Zhao, and D. Zhang. 2017. Measuring Output, Input and Total Factor Productivity in Australian Agriculture: An Industry-Level Analysis. The Review of Income and Wealth 63: S169–S193.
SNA. 2008. System of National Accounts (European Commission, International Monetary Fund, Organisation for Economic Co-operation and Development, United Nations, World Bank).
Spant, R. 2003. Why Net Domestic Product Should Replace Gross Domestic Product as a Measure of Economic Growth. International Productivity Monitor 7: 39–43.
Statistics New Zealand. 2014. Productivity Statistics: Sources and Methods, 10th ed. Available from www.stats.govt.nz.
Timmer, M.P., M. O’Mahony, and B. van Ark. 2007. EU KLEMS Growth and Productivity Accounts: An Overview. International Productivity Monitor 14: 71–85.
van den Bergen, D., M. van Rooijen-Horsten, M. de Haan, and B.M. Balk. 2008. Productivity Measurement at Statistics Netherlands. In Productivity Measurement and Analysis: Proceedings from OECD Workshops (OECDiLibrary; Swiss Federal Statistical Office, Neuchâtel). Also downloadable as Working Paper from www.cbs.nl (search for Growth Accounts).
Vancauteren, M., E. Veldhuizen, and B.M. Balk. 2012. Measures of Productivity Change: Which Outcome Do You Want? Paper presented at the 32nd General Conference of the IARIW, Boston, MA, August 5–11. Available at www.iariw.org.
Author information
Authors and Affiliations
Appendices
Appendix A: Indices and Indicators
The basic measurement tools used are price and quantity indices and indicators. Indices are ratio-type measures, and indicators are difference-type measures. What, precisely, are the requirements for good tools? The following just serves to introduce and illustrate some concepts used in the main text of this book. For a complete treatment and historical information the reader is referred to Balk (2008).
We consider two time periods and an aggregate consisting of M items (goods and services) with (unit) prices going from \(p_m^0\) to \(p_m^1\) and quantities going from \(y_m^0\) to \(y_m^1\) (m = 1, …, M). The first subsection reviews indices, and the second subsection reviews indicators. It is important to note that, though stated in terms of output, the theory surveyed below is equally applicable to other situations after appropriate modification of the number of items and the interpretation of the variables. When there are more than two time periods involved one can choose between direct (bilateral) indices/indicators or chained (multilateral) indices/indicators. The following survey is restricted to bilateral functions.
2.1.1 Indices
A bilateral price index is a positive, continuously differentiable function \(P(p^1,y^1,p^0,y^0): \Re ^{4M}_{++} \rightarrow \Re _{++}\) that, except in extreme situations, correctly indicates any increase or decrease of the elements of the price vectors p 1 or p 0, conditional on the quantity vectors y 1 and y 0. A bilateral quantity index is a positive, continuously differentiable function of the same variables \(Q(p^1,y^1,p^0,y^0): \Re ^{4M}_{++} \rightarrow \Re _{++}\) that, except in extreme situations, correctly indicates any increase or decrease of the elements of the quantity vectors y 1 or y 0, conditional on the price vectors p 1 and p 0. The number M is called the dimension of the price or quantity index.
The basic requirements (aka axioms) on price and quantity indices comprise (1) that they exhibit the correct monotonicity properties; (2) that they are linearly homogeneous in comparison period prices (quantities, respectively); (3) that they satisfy the Identity Test; (4) that they are homogeneous of degree 0 in prices (quantities, respectively); (5) that they are invariant to changes in the units of measurement of the commodities. Interchanging the variables p t and y t (t = 0, 1) transforms any price index into a quantity index and vice versa.
The Product Test requires that price index times quantity index equals the value ratio; that is, P(p 1, y 1, p 0, y 0) × Q(p 1, y 1, p 0, y 0) = p 1 ⋅ y 1∕p 0 ⋅ y 0. This requirement is especially important in the context of deflation. Deflation of a nominal value means dividing it by a price index in order to obtain a real value; that is, something that behaves as a scalar quantity. In our two-period situation, by deflating the nominal comparison period value p 1 ⋅ y 1 by the price index P(p 1, y 1, p 0, y 0) we get as real comparison period value p 1 ⋅ y 1∕P(p 1, y 1, p 0, y 0). The Product Test then implies that this can be written as p 0 ⋅ y 0 × Q(p 1, y 1, p 0, y 0); that is, base period nominal value inflated by a quantity index. This behaves as a scalar quantity if and only if the quantity index satisfies the basic requirements. However, that the price index satisfies the basic requirements is not at all a sufficient condition for the quantity index to do the same.
As defined, price and quantity indices are functions of prices and quantities. In practice it is usually easier to observe (or estimate) price or quantity relatives and (nominal) values or value shares. It turns out, however, that any function P(p 1, y 1, p 0, y 0) or Q(p 1, y 1, p 0, y 0), that is invariant to changes in the units of measurement, can be written as a function of only 3M variables, namely the price relatives \(p_{m}^1/p_{m}^0\) or the quantity relatives \(y_{m}^1/y_{m}^0\), respectively, and the values \(v_m^t \equiv p_{m}^ty_{m}^t\) (m = 1, …, M;t = 0, 1). Value shares are defined as \(s_m^t \equiv v_m^t/\sum _{m=1}^{M}v_m^t\) (m = 1, …, M;t = 0, 1).
2.1.1.1 The Main Formulas
The indices materializing in this book will now be listed. The reader is invited to check their properties. We start by considering the Dutot price index, defined as
Notice that this index does not depend on quantities. Likewise, the Dutot quantity index is defined as
This index does not depend on prices. Dutot indices are not invariant to changes in the units of measurement of the items. The implication of this is that these indices can only be used when the underlying items (goods and services) can reasonably be considered as homogeneous (or interchangeable). Notice that P D(p 1, y 1, p 0, y 0) × Q D(p 1, y 1, p 0, y 0) ≠ p 1 ⋅ y 1∕p 0 ⋅ y 0, except in special cases. It is also interesting to notice that the implicit Dutot price index, that is, the value index divided by the Dutot quantity index, is known as the unit-value index,
This is a ratio of mean prices (aka unit values). The main disadvantage of the unit-value index is that in general the Identity Test for price indices is violated.
We continue by considering the Laspeyres price index. As function of prices and quantities this index is defined as
which can be written as a function (arithmetic mean) of price relatives and (base period) value shares,
The Laspeyres quantity index is defined as
Notice that P L(p 1, y 1, p 0, y 0) × Q L(p 1, y 1, p 0, y 0) ≠ p 1 ⋅ y 1∕p 0 ⋅ y 0, except in special cases.
The Paasche price index is defined as
which can be written as a function (harmonic mean) of price relatives and (comparison period) value shares,
The Paasche quantity index is defined as
Notice that P P(p 1, y 1, p 0, y 0) × Q P(p 1, y 1, p 0, y 0) ≠ p 1 ⋅ y 1∕p 0 ⋅ y 0, except in special cases. However, P L(p 1, y 1, p 0, y 0)×Q P(p 1, y 1, p 0, y 0) = P P(p 1, y 1, p 0, y 0)×Q L(p 1, y 1, p 0, y 0) = p 1⋅y 1∕p 0⋅y 0.
The Fisher price index is defined as the geometric mean of the Laspeyres and Paasche indices,
The Fisher quantity index is similarly defined as
Notice that P F(p 1, y 1, p 0, y 0) × Q F(p 1, y 1, p 0, y 0) = p 1 ⋅ y 1∕p 0 ⋅ y 0.
The geometric alternative to the Laspeyres price index, called GeoLaspeyres price index, is defined as
The GeoLaspeyres quantity index is defined as
Notice that P GL(p 1, y 1, p 0, y 0) × Q GL(p 1, y 1, p 0, y 0) ≠ p 1 ⋅ y 1∕p 0 ⋅ y 0, except in special cases.
The GeoPaasche price index is defined as
and the GeoPaasche quantity index is defined as
Notice that P GP(p 1, y 1, p 0, y 0) × Q GP(p 1, y 1, p 0, y 0) ≠ p 1 ⋅ y 1∕p 0 ⋅ y 0, except in special cases.
The geometric mean of the GeoLaspeyres and GeoPaasche price indices is called the Törnqvist price index,
The Törnqvist quantity index is defined as
Notice that P T(p 1, y 1, p 0, y 0) × Q T(p 1, y 1, p 0, y 0) ≠ p 1 ⋅ y 1∕p 0 ⋅ y 0, except in special cases.
An alternative to the Törnqvist indices are the Sato-Vartia indices. Here the price or quantity relatives are weighed with logarithmic mean value shares, normalized to ensure that these add up to 1. Thus, the Sato-Vartia price index is
where LM(.) denotes the logarithmic mean.Footnote 38 The Sato-Vartia quantity index is defined as
Notice that P SV(p 1, y 1, p 0, y 0) × Q SV(p 1, y 1, p 0, y 0) = p 1 ⋅ y 1∕p 0 ⋅ y 0.
A second alternative are the Montgomery-Vartia indices. The price and quantity indices are defined as, respectively,
where \(V^t \equiv \sum _{m=1}^{M}v_m^t\) (t = 0, 1). Now the price and quantity relatives are weighed with ratios of means instead of means of ratios. These weights do not add up to 1, though generally the discrepancy appears to be negligible. Notice that P MV(p 1, y 1, p 0, y 0) × Q MV(p 1, y 1, p 0, y 0) = p 1 ⋅ y 1∕p 0 ⋅ y 0.
2.1.1.2 Two-Stage Indices
It is important to consider the concept of two-stage indices. Let the aggregate under consideration be denoted by A, and let A be partitioned arbitrarily into K disjunct subaggregates A k,
Each subaggregate consists of a number of items. Let M k ≥ 1 denote the number of items contained in A k (k = 1, …, K). Obviously \(M = \sum ^{K}_{k=1}M_{k}\). Let \((p^{1}_{k},y^{1}_{k},p^{0}_{k},y^{0}_{k})\) be the subvector of (p 1, y 1, p 0, y 0) corresponding to the subaggregate A k. Recall that \(v_m ^{t} \equiv p_m^ty_m^t\) is the value of item m at period t. Then \(V^{t}_{k} \equiv \sum _{m \in A_k}v^{t}_{m}\) (k = 1, …, K) is the value of subaggregate A k at period t, and \(V^{t} \equiv \sum _{m \in A}v^{t}_{m} = \sum ^{K}_{k=1}V^{t}_{k}\) is the value of aggregate A at period t.
Let P(.), P (1)(.), …, P (K)(.) be price indices of dimension K, M 1, …, M K respectively that satisfy the basic requirements. Recall that any such price index can be written as a function of price relatives and base and comparison period item values. Replace in P(.) the price relatives by subaggregate price indices and the item values by subaggregate values. Then the price index defined by
is of dimension M and also satisfies the basic requirements. The index P ∗(.) is called a two-stage index. The first stage refers to the indices P (k)(.) for the subaggregates A k (k = 1, …, K). The second stage refers to the index P(.) that is applied to the subindices P (k)(.) (k = 1, …, K). A two-stage index such as defined by expression (2.137) closely corresponds to the calculation practice at statistical agencies. All the subindices are then usually of the same functional form, for instance Laspeyres or Paasche indices. The aggregate, second-stage index may or may not be of the same functional form. This could be, for instance, a Fisher index.
If the functional forms of the subindices P (k)(.) (k = 1, …, K) and the aggregate index P(.) are the same, then P ∗(.) is called a two-stage P(.)-index. Continuing the first example, the two-stage Laspeyres price index reads
and one simply checks that the two-stage and the single-stage Laspeyres price indices coincide; that is,
However, this is the exception rather than the rule. For most indices, two-stage and single-stage variants do not coincide. Put otherwise, most indices are not Consistent-in-Aggregation (CIA).
Two-stage quantity indices can be defined analogously.
2.1.2 Indicators
The continuous functions \(\mathcal {P}(p^1,y^1,p^0,y^0): \Re ^{4M}_{++} \rightarrow \Re \) and \(\mathcal {Q}(p^1,y^1,p^0,y^0): \Re ^{4M}_{++} \rightarrow \Re \) will be called price indicator and quantity indicator, respectively, if they satisfy the following basic requirements (aka axioms): (1) the functions exhibit the correct monotonicity properties; (2) they satisfy the Identity Test; (3) they are homogeneous of degree 1 in prices or quantities, respectively; (4) they are invariant to changes in the units of measurement of the commodities. Interchanging the variables p t and y t (t = 0, 1) transforms any price indicator into a quantity indicator and vice versa. The analogue of the Product Test requires that price indicator plus quantity indicator equals the value difference; that is, \(\mathcal {P}(p^1,y^1,p^0,y^0) + \mathcal {Q}(p^1,y^1,p^0,y^0) = p^1 \cdot y^1 - p^0 \cdot y^0\). Notice that these functions may take on negative or zero values.
As values are additive, we have
Given a pair of price and quantity indicators satisfying the Product Test, it must then be the case that
where the dimension of the indicators at the right-hand side of this equation equals 1. A natural requirement is then that \(\mathcal {P}(p^1,y^1,p^0,y^0) = \sum _{m \in A}\mathcal {P}(p_m^1,\) \(y_m^1,p_m^0,y_m^0)\) and \(\mathcal {Q}(p^1,y^1,p^0,y^0) = \sum _{m \in A}\mathcal {Q}(p_m^1,y_m^1,p_m^0,y_m^0)\). Any such indicator is CIA, as
and
Recall that any function \(\mathcal {P}(p^1,y^1,p^0,y^0)\) or \(\mathcal {Q}(p^1,y^1,p^0,y^0)\), that is invariant to changes in the units of measurement, can be written as a function of only 3M variables, namely the price relatives \(p_{m}^1/p_{m}^0\) or the quantity relatives \(y_{m}^1/y_{m}^0\), respectively, and the values \(v_m^t\) (m = 1, …, M;t = 0, 1). This leads to further specifications of the indicators.
It is also useful to know that any indicator can be transformed into an index, and vice versa (Balk 2008, 128–129). In the process of transformation, however, certain properties may get lost.
2.1.2.1 The Main Formulas
The indicators materializing in this book will now be reviewed. The Laspeyres price indicator as function of prices and quantities is defined as
This indicator can be written as a function of price relatives and (base period) values,
The Laspeyres quantity indicator is similarly defined as
Notice that \(\mathcal {P}^{L}(p^1,y^1,p^0,y^0) + \mathcal {Q}^{L}(p^1,y^1,p^0,y^0) \neq p^1 \cdot y^1 - p^0 \cdot y^0\), except in special cases.
The Paasche price indicator, defined as
can be written as a function of price relatives and (comparison period) values,
The Paasche quantity indicator is defined as
Notice that \(\mathcal {P}^{P}(p^1,y^1,p^0,y^0) + \mathcal {Q}^{P}(p^1,y^1,p^0,y^0) \neq p^1 \cdot y^1 - p^0 \cdot y^0\), except in special cases. However, \(\mathcal {P}^{L}(p^1,y^1,p^0,y^0) + \mathcal {Q}^{P}(p^1,y^1,p^0,y^0) = \mathcal {P}^{P}(p^1,y^1,p^0,y^0) + \mathcal {Q}^{L}(p^1,y^1,p^0,y^0) = p^1 \cdot y^1 - p^0 \cdot y^0\).
The Bennet price indicator is usually defined by
It is the arithmetic mean of the Laspeyres and Paasche indicators; that is,
which can be written as
The Bennet quantity indicator is defined by
One easily verifies that \(\mathcal {P}^{B}(p^1,y^1,p^0,y^0) + \mathcal {Q}^{B}(p^1,y^1,p^0,y^0) = p^1 \cdot y^1 - p^0 \cdot y^0\).
In the Bennet price indicator the price differences \(p_m^1 - p_m^0\) are multiplied by arithmetic means of base and comparison period quantities \((y_m^0 + y_m^1)/2\). The Montgomery price indicator uses instead logarithmic mean values divided by logarithmic mean prices; that is,
The Montgomery quantity indicator is defined as
Using the definition of the logarithmic mean one easily verifies that \(\mathcal {P}^{M}(p^1,y^1,\) \(p^0,y^0) + \mathcal {Q}^{M}(p^1,y^1,p^0,y^0) = p^1 \cdot y^1 - p^0 \cdot y^0\).
Appendix B: Decompositions of the Value-Added Ratio
Value added is defined as revenue minus the cost of intermediate inputs. Let the revenue ratio R 1∕R 0 as in expression (2.14) be decomposed as
for certain price and quantity indices, and let the intermediate inputs cost ratio \(C_{\mathit {EMS}}^1/C_{\mathit {EMS}}^0\) be decomposed by, not neccessarily the same, price and quantity indices as
How can the (nominal) value added ratio
then be decomposed into meaningful price and quantity components? It is clear that any of these components potentially requires a lot of data, namely \((w_{\mathit {EMS}}^1,x_{\mathit {EMS}}^1,w_{\mathit {EMS}}^0, x_{\mathit {EMS}}^0,p^1,y^1,p^0,y^0)\), data that are not always immediately available. Therefore, traditionally, one had recourse to so-called single deflation, where, for instance, a revenue-based price index is used to deflate nominal value added. That is, one sets \(P_{\mathit {VA}}^{sd}(1,0) \equiv P_R(1,0)\), and let the quantity index be the remainder,
Using the various definitions, this can be written as
but also as
Two important observations must be made. First, if VA t > 0 (t = 0, 1) then \(Q_{\mathit {VA}}^{sd}(1,0) > 0\). Second, it appears that in this setup the quantity component of the value added ratio depends on the price index of intermediate inputs relative to the price index of gross output, P EMS(1, 0)∕P R(1, 0). Since it is not at all certain that this ratio dwells in the neighbourhood of 1, this can create undesirable outcomes.
A variant is mentioned by SNA (2008, par. 15.135), namely \(Q_{\mathit {VA}}^{sd'}(1,0) \equiv Q_R(1,0)\); that is, the growth rate of real value added is set equal to the growth rate of gross output. It appears that
Thus, if the ratio of revenue to value added does not change then the two single-deflation variants coincide. In the literature this ratio is known as the Domar factor.
When the necessary data are available, SNA (2008, par. 15.133) prefers double deflation, which means that the value-added based quantity index is defined asFootnote 39
or alternatively as
Both quantity indices have a two-stage structure. The first stage is given by the quantity indices for gross output and intermediate inputs, Q R(1, 0) and Q EMS(1, 0), respectively.Footnote 40 The second stage is in the case of expression (2.162) a Laspeyres index (with weights from the base period), and in the case of expression (2.163) a Paasche index (with weights from the comparison period). Notice the minus signs for the parts concerning intermediate inputs.
It is then rather natural to propose the geometric mean of the two quantity indices as the desired quantity index. Put otherwise, the quantity component of the value added ratio is defined as the Fisher index of the subindices Q R(1, 0) and Q EMS(1, 0); that is,
Similarly, the price component of the value added ratio is defined as the Fisher index of the subindices P R(1, 0) and P EMS(1, 0); that is,
One easily checks that \(P_{\mathit {VA}}^F(1,0)Q_{\mathit {VA}}^F(1,0) = \mathit {VA}^1/\mathit {VA}^0\). This pair of indices was proposed by Geary (1944), though Karmel (1954) discloses some earlier sources.
For studying the behaviour of the quantity index \(Q_{\mathit {VA}}^F(1,0)\) the following relation is useful:
When VA t > 0 then \(1 - C_{\mathit {EMS}}^t/R^t > 0\) (t = 0, 1). However, the function \(Q_{VA}^F(1,0)\) is undefined when \(Q_R(1,0) \leq (C_{\mathit {EMS}}^0/R^0)Q_{\mathit {EMS}}(1,0)\) or \(Q_R(1,0) \geq (R^1/C_{\mathit {EMS}}^1)Q_{\mathit {EMS}}(1,0)\). Moreover, Eq. (2.166) implies the following relations:
Karmel (1954), however, showed that in the case of chained indices these relations might be violated. Thus there may occur situations where Fisher indices are undefined.
An alternative decomposition, which does not exhibit this defect, can be developed as follows. For the logarithm of the value added ratio we get by repeated application of the logarithmic mean LM(a, b),
Using the decompositions of the revenue ratio and the intermediate inputs cost ratio, the logarithm of the value added ratio can be expressed as
This can be rearranged as
where ϕ(1, 0) ≡LM(R 1, R 0)∕LM(V A 1, V A 0), that is, mean revenue over mean value added, and \(\psi (1,0) \equiv \mathit {LM}(C_{\mathit {EMS}}^1,C_{\mathit {EMS}}^0)/\mathit {LM}(VA^1,VA^0)\), that is, mean intermediate inputs cost over mean value added. Thus, value-added based price and quantity indices can rather naturally be defined by
These indices generalize the conventional Montgomery-Vartia indices (Appendix A). Their disadvantage is that P R(1, 0) = P EMS(1, 0) (or Q R(1, 0) = Q EMS(1, 0)) not necessarily implies that \(P_{\mathit {VA}}^{MV}(1,0) = P_R(1,0)\) (or \(Q_{\mathit {VA}}^{MV}(1,0) = Q_R(1,0)\)). Formally stated, these indices fail the Equality Test. The reason behind this failure is that generally
is close but not equal to 1 because LM(a, 1) is a concave function. Fortunately, in ‘normal’ situations the discrepancy appears to be unimportant.
Summarizing the foregoing, we see that
-
single deflation is simple, leads to a quantity index that is always positive, but this index may be biased;
-
double deflation by a Fisher index leads to a quantity index that is unbiased but may not always be well-defined;
-
double deflation by a Montgomery-Vartia index leads to a quantity index that is always well-defined but fails the Equality Test.
Thus, there is no theoretically entirely satisfactory solution to the problem of decomposing a value-added ratio into a price index and a quantity index.
Appendix C: The Domar Factor
2.1.1 And the TFP Index
According to the definition of the gross-output based TFP indexFootnote 41
We know that \(C^t = C^t_{\mathit {KL}} + C^t_{\mathit {EMS}}\) (t = 0, 1). It is assumed that the primary inputs cost ratio is decomposed into a price and a quantity index, \(C^1_{\mathit {KL}}/C^0_{\mathit {KL}} = P_{\mathit {KL}}(1,0)Q_{\mathit {KL}}(1,0)\). It is also assumed that the intermediate inputs cost ratio is decomposed into a price and a quantity index, \(C^1_{\mathit {EMS}}/C^0_{\mathit {EMS}} = P_{\mathit {EMS}}(1,0) \times Q_{\mathit {EMS}}(1,0)\). Then Q C(1, 0) can be defined as the Montgomery-Vartia index of Q KL(1, 0) and Q EMS(1, 0);Footnote 42 that is,
where LM(a, b) is the logarithmic mean of a and b. Then, substituting this expression into expression (2.174), we obtain
Next, the definition of value added is rewritten as \(R^t = \mathit {VA}^t + C^t_{\mathit {EMS}}\) (t = 0, 1). Let Q R(1, 0) be defined as the Montgomery-Vartia index of Q VA(1, 0) and Q EMS(1, 0). Then the logarithm of the value-added based quantity index can be backed out as
Substituting this into the definition of the value-added based TFP index delivers
Combining the two expressions (2.176) and (2.178) by eliminating \(\ln Q_{R}(1,0)\) delivers
The factor in front of the square brackets,
is known as the Domar factor: the ratio of (mean) revenue over (mean) value added. One easily checks that if revenue R t is equal to cost C t (or value added V A t is equal to primary inputs cost \(C_{\mathit {KL}}^{t}\), or profit Πt is equal to 0) (t = 0, 1), then expression (2.179) reduces to
This corresponds to Proposition 1 of Gollop (1979). See Jorgenson et al. (2005, 298) for a modern derivation under the usual neo-classical assumptions.
2.1.2 And the Labour Productivity Index
According to the definition of the value-added based labour productivity index, expression (2.81),
Substituting expressions (2.177) and (2.180) delivers
According to the definition of the gross-output based labour productivity index, expression (2.61),
Substituting this into expression (2.183) delivers
The assumption that Q R(1, 0) is the Montgomery-Vartia index of Q VA(1, 0) and Q EMS(1, 0) implies that the sum of its weights is approximately equal to 1; that is,
Employing the definition of the Domar factor, this implies that
Substituting this into expression (2.185) delivers
which is our final result. Notice that the second term at the right-hand side of this approximate equality does not vanish if revenue R t equals cost C t. The term vanishes if labour quantity change Q L(1, 0) equals the aggregate quantity change of intermediate inputs Q EMS(1, 0).
Rights and permissions
Copyright information
© 2021 The Author(s), under exclusive license to Springer Nature Switzerland AG
About this chapter
Cite this chapter
Balk, B.M. (2021). A Framework Without Assumptions. In: Productivity. Contributions to Economics. Springer, Cham. https://doi.org/10.1007/978-3-030-75448-8_2
Download citation
DOI: https://doi.org/10.1007/978-3-030-75448-8_2
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-75447-1
Online ISBN: 978-3-030-75448-8
eBook Packages: Economics and FinanceEconomics and Finance (R0)