Economic Modelling – shining light on a dark art

Economic modelling has been likened to driving forwards while looking only in the rear vision mirror, as it can only use historical data and this can be a poor guide to what will happen in the future. Economic forecasting (modelling with more courageous assumptions) has been described as an activity invented to make fortune telling look respectable. These light-hearted definitions are based on the poor record that economic analysis has in predicting the future, especially in the many cases where the future does not continue a trend visible in the recent past. Economic models are based on long run relationships and ratios, and on some strong underlying assumptions that constrain human behaviour and decision–making. Were these constraints removed, it is often perceived that modelling would be too inconclusive to be useful.

Input-output approaches
The regional input output approach is the most common way of building a predictive model of how a regional economy will react to external shocks. Unfortunately, while input-output models have been used effectively at the national scale, at the regional and local scale they cannot reflect enough of the local specialisations, strengths and weaknesses to be reliable. A regional-scale input-output approach known as Generating Regional Input-output Tables (GRIT) is in common use, and a study for the Cotton Catchment Communities CRC in 2008 noted a series of limitations (my comments in bold):

1. A linearity assumption implies that any change has proportionate effects throughout the economy so that there are no substitutions among inputs and products. Implication – fixed supply chains from paddock to export mean that the model cannot respond to adaptation and seasonal variation, for example by reducing need for labour through capital equipment.

2. A set of homogeneity assumptions mean that all of the entities (eg farms) in the specified sectors are the same in terms of production technology, products produced, goods consumed, etc. Implication – model eliminates adaptation amongst growers and between growers.

3. There is no consideration of market effects in the input-output model and all results are based on real changes in production of goods and services. Implication – models allows for no feedback between price signals (and more importantly for growers price expectations and forecasts) and cropping behaviour.

The exclusion of price-reactions in input-output models and the inherent assumption that the main ‘factors of production’ (land, labour and capital) are not constrained are the major criticisms of them levelled by economists.

“The rigidity of input output tables and their inability to conform to the most basic of economic assumptions about the role of price in the economy means that input output modelling should only be used to evaluate relative small changes in the economy.” (Denniss, 2012)

“Multipliers assume that extra output can be produced without constraints on the supply of labour, capital, land, good or service. The factors of production are assumed to be limitless in supply and therefore can be sourced without any price increase.” (WA Treasury, 2002)

Regional input-output models are often used to quantify the flow-on effects through a regional economy of an external ‘shock’ such as a change in commodity prices, labour costs or energy prices. The flow-on effects are estimated by considering the ‘multipliers’ through the regional economy, multipliers which are part of the regional input-output model and which link the inputs into an industry from other industries, and the output from an industry into others. While still widely used, these multipliers are highly misleading and are no longer published by the ABS as they have been used inappropriately:

“While Input-Output multipliers may be useful as summary statistics to assist in understanding the degree to which an industry is integrated into the economy, their inherent shortcomings make them inappropriate for economic impact analysis. These shortcomings mean that Input-Output multipliers are likely to significantly over-state the impacts of projects or events.” (ABS, 2009 )

Economists criticise the results of input-output modelling when used to demonstrate the significance of an industry and the impact of its expansion or contraction:

“While multipliers can be a useful way of summarising and quantifying interlinkages within the economy, they are more often abused than used correctly. Multipliers are used to suggest that an industry is more valuable to Western Australia than its current size would suggest. They are used to show substantial flow-on benefits to the broader economy and to justify claims for government support for that activity. However, multipliers do not provide a measure of net economic benefit of expanding activity in a particular area. They are based on limiting assumptions and dated information.” (WA Treasury 2002)

“Claims that jobs ‘gained’ directly from the cause being promoted will lead to cascading gains in the wider economy often fail to give any consideration to the restrictive nature of the assumptions required for input-output multiplier exercises to be valid. In particular, these applications fail to consider the opportunity cost of both spending measures and alternate uses of resources, and may misinform policy-makers.” (Gretton, Productivity Commission 2013 )

In April 2013 these criticisms were noted in a landmark case in the NSW Land and Environment Court. Rio Tinto applied for an extension at its Warkworth Mine, and the extension was denied by Justice Preston, in part because the proponents overstated the positive impact of the expansion on the regional economy. Justice Preston noted that”

“… employment generated from the extension of the Warkworth mine would involve currently employed skilled workers transferring from other industries, but the vacancy thereby created in the other industries may not necessarily be filled, partly because of a shortage of skilled workers and partly because the remuneration is inferior to that offered in the mining industry.“ (judgement, para 460 )

In other words, the input-output multiplier calculations did not take account of the state of the Hunter labour market, and that all those extra jobs ‘created’ by the mine extension would have to come from somewhere, and this was not included in the analysis presented.

And even if the impact estimates were more accurate, Justice Preston noted that input-output modelling is at best only a partial input to the decision making process about whether or the costs of the project outweigh the benefits.

“The deficiencies in the data and assumptions used affect the reliability of the conclusions as to the net economic benefits of approval. More fundamentally, however, the IO analysis does not assist in weighting the economic factors relative to the various environmental and social factors, or in balancing the economic, social and environmental factors.” (judgement para 451)

“It provides, therefore, some information but only on one set of matters relevant to be considered by the approval authority in determining the project application. The IO analysis is not a substitute for the decision-making process that the approval authority must undertake in determining the project application, and the conclusions the IO analysis reaches cannot be substituted for the fact finding, weighting and balancing of all of the relevant environmental, social and economic matters required to be considered by the approval authority. The conclusions the IO analysis reaches on the economic benefits of approving the Project, evaluated for their reliability and given appropriate weight, need to be balanced against all other environmental, social and economic benefits and costs.” (judgement para 463)

These limitations reduce the effectiveness of these models in forecasting the evolution of local economies, and in estimating the impacts of the changes that are affecting them.

General equilibrium modelling approaches
Computable General Equilibrium (CGE) modelling starts with similar economic flows data to input-output models but allows capital and labour to move and allows for demand and supply to respond to changes in prices. CGE modelling is designed to model the impacts of an external shock to an economy (a shock like a tax change or resource price change) to see what the economy looks like once the shock has been absorbed and the economy is again in equilibrium.

“CGE models allow for prices to change the relative use of different factors of production in the production of a good or service. That is, while input-output models are an attempt to explain how much wheat, energy, labour and capital is used to make bread a CGE model might be used to estimate the impact of a wage rise on the amount of labour used in bread production.”(Denniss 2012)

While superior to input-output modelling in being less prone to overstating impacts, weaknesses in model design and modelling methodologies bring other constraints.

“.. . standard versions of MMRF [an Australian CGE model] do not include substitution possibilities between material and capital inputs to production, constant returns to scale are imposed on all industries, workforce participation and employment are usually assumed to be a fixed share of the working age population and workforce participants, respectively.” (Gretton 2013)

The inability of CGE modelling to take account of substitution between material and capital inputs to production is a particular weakness when the models are applied to estimate changes in firm behaviour from external shocks like a new tax or a reduced water allocation. These types of shocks may well have a big impact on the material and capital inputs used by a firm, and excluding this adaptation from the analysis greatly weakens the robustness of the results.

Elasticity assumptions are crucial in driving CGE modelling outputs. They are not computable from the model but are important inputs into it. This means that CGE models are weak when applied to regions rather than nations, as they cannot take account of regional specialisations, like higher land or labour productivity, or different elasticities of supply amongst firms in a region compared with firms elsewhere.

CGE modelling is all about equilibrium, comparing the state of the economy at one equilibrium (before a shock) and then another (after the shock).

“One of the most important, and least understood, features of CGE models is that they assume that, in the long run, the economy will be in full employment and that the path that the economy follows has no impact on its long run destination. … the economy is assumed to “converge‟ towards a “well defined steady state‟ in the longer run. Put another way, the modelling starts with the answer for what the long run level of GDP will be and then tries to plot the course that the economy will follow from where it currently is to where they know it will end up.” (Denniss 2012)

This means that CGE modelling has to take an economy-wide view of the shock, mapping both ‘winners’ and losers’ at the new equilibrium point, but not mapping how either made the transition or how the transition is actually effected, or where in an economy (below the national level) the impacts will be dispersed.

Local production systems approach
The alternative to scaling down a national economic model is to build up a local economic model of the agricultural production system. While this approach is likely to be weak in its ability to track the flow-on effects and links between all the different parts of the economy (the strength of input-output and CGE modelling) it gives a much more accurate picture of the scale of activities along a local supply chain, and the responses to increased or decreased activity (a weakness of input-output and CGE modelling).

The local production systems approach set out here is based on an agricultural supply chain. The approach is based on the collection of data from local growers, handlers and value-adders on their scale of operations (including employment), the factors that determine this, and the main upstream and downstream links in the supply chain. A local production systems model can be used as the basis for estimating the impact on local economies of changes in farm production. The estimated impacts are based on the historical experiences of businesses over the last decade, a period which in Australia covers both low-production drought years and some high production years.

The first step is to map the most important supply chains in the area being researched, and in Australia a good overview can be prepared from the ABS Business Counts data series (ABS Cat 8165.0) when used alongside Census and agricultural production data. Industry networks and phone index listings and local knowledge should be used to build a more complete picture of the main businesses operating along the selected supply chains.

Once the main supply chain(s) has been mapped using readily available information, the survey/interview process for the key businesses in the supply chain(s) can be determined. For an agricultural supply chain this would look like:
• Growers/producers
• Contractors and freight
• Processor/packers
• Manufacturers
• Rural supplies/rural services businesses

The interviews need to explore how businesses in the supply chain have made (and are likely to make in future) changes to their operations under different climate, market and regulatory scenarios. The interviews should probe for how these businesses have managed their employment and contracting to ride out the peaks and troughs in their business cycles. With information on peak and trough activity levels it is possible to identify the thresholds that will make operators change the way they run their businesses.

The peaks and troughs data gathered from the businesses in the local supply chain(s) enables modelling of the impact of changes in activity levels on direct employment and flow-on employment and business activity.

I recommend that three scenarios be modelled:
• ‘Best case’ with historically high levels of activity, planting, cropping or water availability
• A ‘medium case’ with intermediate levels of planting or activity
• ‘Worst case’ with historically low levels of activity, planting, harvesting or water availability.

In running the calculations you will need to draw from the interviews done, or from other industry data applicable to your supply chains, to find the key activity ratios from historical performance data to benchmark the high, medium and low scenarios ratios that drive the modelling:
• Activity level (eg turnover or planting area) (best case, typical case and worst case)
• Local spend (best, typical, worst)
• Employment (full time and casual, best, typical, worst)
• Profitability (best, typical, worst)

The value in the modelling, and the contribution it makes to understanding community wealth and wealth being, lies in the ability to estimate the flow-on impact on local spending and supply chain employment across the three scenarios – and to do it locally.

In assessing different future scenarios for communities, the modelling helps understand the total contribution to a local economy of the supply chains that have been analysed. When the modelling is applied to more than one supply chain, it can be used to compare the net impact on local employment and spending of changes in each. So, for example, when adaptability in tourism businesses is included, expansion or contraction in visitor numbers or visitor spend can be modelled to estimate the impact on local employment and local spending, in the same way the changes in farm scale activity have been modelled in the examples above. The impacts of a ‘farming led’ or ‘tourism led’ future can then be compared, with the contributions from each supply chain better understood.

References:
The Socio-Economic Impact of Cotton on Cotton Catchment Communities in NSW and QLD Centre for Agricultural and Regional Economics, CARE 2008
Denniss, R, The use and abuse of economic modelling in Australia , Technical Brief No. 12 The Australia Institute, January 2012.
The Use and Abuse of Input-Output Multipliers, Economic Research Articles, WA Treasury March 2002
Australian National Accounts: Input-Output Tables – Electronic Publication, 2005-06 Final (Cat 5209.0.55.001), Canberra 2009
Gretton, P, On input-output tables: uses and abuses, Productivity Commission Staff Research Note, Sept 2013.
NSW Land and Environment Court Bulga Milbrodale Progress Association Inc v Minister for Planning and Infrastructure and Warkworth Mining Limited [2013] NSWLEC 48, http://www.caselaw.nsw.gov.au/action/PJUDG?jgmtid=164038

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s