Measuring Productivity Change (Without Neoclassical Assumptions) Bert M. Balk Statistics Netherlands and Rotterdam School of Management Erasmus University Washington DC, 14 May 2008

Opening quote “… productivity measurement is all about comparing outputs with inputs, …” Ross Gittins, “Productivity should be a spin- free zone”, The Sydney Morning Herald, June 23-24, 2007.

Usual neo-classical assumptions – Technology exhibits constant returns to scale (CRS). – Competitive input and output markets (exogenously given prices). – Production unit maximizes profit. – Production unit has perfect foresight (ex post prices used as if ex ante known). Can we dispense with such assumptions?

Generic model (market) Input Unit Output Input value: Output value: Cost Revenue

What is productivity (change)? Usual financial performance measures are: – profit = revenue minus cost (positive, negative, or zero); – profitability = revenue divided by cost (greater than 1, less than 1, or equal to 1). Profit and profitability change can be decomposed in price and quantity components. Natural decomposition for profit is additive (indicators) and for profitability multiplicative (indices). The quantity component of profit (-ability) change, or real profit (-ability) change, is called (total factor) productivity change.

Formal definitions TFP index: IPROD ≡ Q(output) / Q(input) Good choice: Fisher TFP indicator: DPROD ≡ Q(output) - Q(input) Good choice: Bennet Interpretation of difference-type measures requires that all prices be deflated by some general inflation index (for example: headline CPI).

This sounds simple, but … … what precisely is to be considered as input and output?

KLEMS-Y model Capital K Labour L Output Y Energy E Unit (Goods & Materials M Services) Services S Input: (Gross) Output: Cost+ Profit = Revenue

KL-VA model Capital K Revenue Unit minus Labour L E,M,S cost Input: Output: Primary inputs cost + Profit = Value added

K-CF model Value added Capital K Unit minus L cost Input: Output: Capital input cost + Profit = Cash flow

TFP indices (1) IPROD-Y ≡ Q(gross output) / Q(KLEMS) IPROD-VA ≡ Q(value added) / Q(KL) IPROD-CF ≡ Q(cash flow) / Q(K) If Profit = 0 then ln(IPROD-CF) ≥ ln(IPROD-VA) ≥ ln(IPROD-Y).

TFP indicators (1) But for indicators we get DPROD-CF = DPROD-VA = DPROD-Y. Thus the three models are not really different!

Capital input cost Total user cost of all asset types i and ages j is C(capital) t = ∑ i ∑ j u t ij K t ij + ∑ i ∑ j v t ij I t ij, where K and I are quantities of asset-type by age (available at the beginning of year t and invested at midyear respectively).

Unit user cost (ex post) (1) For asset of age j (at midyear): u t j = r t P t- j (P t- j-0.5 – E t- P t+ j+0.5 ) + (E t- P t+ j P t+ j+0.5 ) (j=1,…,J) where r t denotes a certain nominal interest rate, the P’s are prices (valuations), and E is the expectation operator.

Unit user cost (ex post) (2) There are three components –“waiting cost” (interest rate times asset price); –anticipated time-series depreciation (effect of time and ageing); –unanticipated revaluation. This leads to four additional input - output models.

KL-NVA model Capital (waiting KValue added cost) Unit - ex post Labour L t-s depreciation Input: Output: Partial primary + Profit = Net value added inputs cost (ex post)

KL-NNVA model Capital (waiting K Value added cost) Unit - anticipated Labour L t-s depreciation Input: Output: Partial primary + Profit* = Net value added inputs cost (normal)

KL-NNVA model Profit* ≡ Profit + Unanticipated asset revaluation

K-NCF model Capital Net value added (waiting K Unit (ex post) cost) - L cost Input: Output: Waiting cost of + Profit = Net cash flow (owned) capital (ex post)

K-NNCF model Capital Net value added (waiting K Unit (normal) cost) - L cost Input: Output: Waiting cost of + Profit* = Net cash flow (owned) capital (normal)

TFP indices (2) IPROD-NVA ≡ Q(net value added) / Q(KwL) IPROD-NCF ≡ Q(net cash flow) / Q(Kw) If Profit = 0 then ln(IPROD-NCF) ≥ ln(IPROD-NVA) ≥ ln(IPROD-VA) ln(IPROD-NCF) ≥ ln(IPROD-CF) ≥ ln(IPROD-VA)

TFP indicators (2) DPROD-NCF = DPROD-NVA = DPROD-CF = DPROD-VA = DPROD-Y but DPROD-NNVA = DPROD-NNCF are different.

The rate of return (1) Net cash flow (ex post) can be seen as the return to the capital stock. The accounting identity is r t (∑ i ∑ j P t- i,j-0.5 K t- i,j (1/2)∑ i ∑ j P t ij I t ij ) + Profit = Net cash flow. Setting Profit = 0 and solving for r t delivers the so-called “endogenous (or internal, or balancing) rate of return”. Specific for unit. Alternative: use some reasonable, exogenous rate.

The rate of return (2) Alternatively, normal net cash flow can be seen as the return on the capital stock. The accounting identity is r t (∑ i ∑ j P t- i,j-0.5 K t- i,j (1/2)∑ i ∑ j P t ij I t ij ) + Profit* = Normal net cash flow. Setting Profit* = 0 and solving for r t delivers the “normal endogenous rate of return”. This rate absorbs also unanticipated asset revaluations. Alternative: use some reasonable, exogenous rate.

Implementation by Statistics Netherlands Details in Van den Bergen et al. (2007) and Balk & Van den Bergen (2006).

TFP change (%): Commercial sector

Explanation (1): From macro to micro Profitability change of an aggregate unit is the outcome of: – Lower level profitability change, which can be decomposed in –Productivity change –Differential price change – Expansion and contraction of units – The entry of units – The exit of units

Explanation (2): What is productivity change? Technological change (incl. management technics). Efficiency change (technical). Scale effects. Input- and/or output-mix change (for example due to relaxation of capacity restrictions, movement from / toward perfect competition).

And here come the neoclassical assumptions … When efficiency change, scale effects, and input-/output-mix effects are assumed away, and the unit is assumed to have perfect foresight (so that ex post user cost = ex ante user cost), then … … productivity change = technological change. But all these assumptions are highly unlikely!