Reassessing productive capacity in the United States.

Holland, Dawn

This note considers the productive capacity of the United States economy, and highlights the key developments that have caused a reassessment of potential output since 2007. This issue is central to the debate on the size of the output gap in the US, which is a key input into macroeconomic policy decision-making. The Congressional Budget Office (CBO) made a downward revision to their estimate of potential output in the US last September, due to persistent effects of the recent recession (CBO, 2011). A permanent or persistent loss of output, often called a 'scar', affects the estimates of the structural and cyclical components of the government budget deficit, and hence the sustainable level of government spending and the need for fiscal adjustment.

A reassessment of potential output will also affect our judgement on the level of spare capacity in the economy, a key indicator of inflationary pressures used to guide monetary policy decisions. President of the St Louis Federal Reserve, James Bullard, recently expressed the view that the US output gap may be smaller than typical estimates, and has sparked a heated debate around the roles of demand and supply in determining the output gap.

The estimates reported here support the views expressed by the CBO and Mr Bullard that the output gap in the US is indeed smaller than typical estimates suggest. Our estimates of the scar on output relative to expectations in 2007 exceed those reported in most other studies--primarily because we consider not only the direct impact of the financial crisis and recession, but also the steep rise in the price of oil over the past four years. Some caution should be used when interpreting the impact of the oil price rise, as it is difficult to extract permanent from transitory shifts in volatile commodity prices or to identify, a satisfactory explanation for the price rise over this period. In addition, research into alternative energy sources and new gas discoveries in the US may reduce oil dependency over the medium term.

Methodology

The methodological approach of the analysis presented in this note is based on the framework developed by Barrell (2009) and extended by Barrell and Kirby (2011) and Barrell, Delannoy and Holland (2011). It is based on a production function approach to measuring capacity output, and is comparable to the methodology used by the CBO (2011) to measure capacity in the US, as well as studies by the IMF (2010) for the UK and OECD (2010) for a broader spectrum of OECD economies. This Review has discussed the magnitude of the scar on output in the UK in this context on numerous occasions, but this is the first in-depth study in this Review that is focused on the United States.

We model the potential output of an economy through an underlying production function that relates the aggregate productive capacity of the economy to the available units of production and the current state of technology. We assume an underlying functional form that can be expressed as:

Q = [gamma] [[delta] [K-.sup.[rho]] + (1 - [delta])[[([Le.sup.[lambda]t]).sup.-[rho]].sup.-(1-[alpha])/[rho]] [M.sup.[alpha]] (1)

where Q is output, K is the capital stock, L is labour input, t is a labour augmenting technology measure and M is energy input. The parameter [gamma] captures neutral technical progress, while [delta] is a distribution parameter, [lambda] is the average rate of labour augmenting technical progress, p is related to the elasticity of substitution between labour and capital ([sigma] = 1/(1+ [rho])) and [alpha] is the oil share of output in a base year. This framework imposes a unit elasticity of substitution between oil input and the labour-capital bundle, so that the share of oil costs in production is constant in the long run, although adjustment towards this long run in response to a price rise may be very protracted. The inner function imposes a CES relationship between capital and labour, with an elasticity of substitution between these two factors of production of about 0.5. (1)

A loss of potential output within this framework must reflect a decline in one or more of the available factors of production (L, K or M), or a decline in the average level or rate of technical progress (or productivity), which would be captured through a shift in [gamma], [lambda] or t. A shift in relative prices can affect the relative demand for factor inputs, while a shift in input costs relative to output costs will affect demand for all factor inputs and the potential level of production.

The associated cost minimising factor demands can be derived by setting the marginal product of each factor to its real cost. In the next section we look at the shocks that the US economy has sustained in terms of input prices and available factors of production since 2007, distinguishing between developments that affect labour input, capital input and energy input. The impact on potential output in the US is assessed through simulations using the National Institute's Global Econometric Model, NiGEM. (2)

Shocks to labour input

The total labour input available to the production process can be decomposed into a demographic component reflecting growth in the working age population (this includes net immigration), the rate of participation in the labour force, the average hours worked per employee and the inverse of the unemployment rate (the share of the labour force that is employed). Figure 1 puts developments in the labour input in the US since 2007 into context with other advanced economies that were affected by the financial crisis.

According to NIESR estimates, the average level of labour input in the US in 2012 will be about 2 1/2 per cent below its level before the financial crisis in 2007. A closer look at the contributions to the change in labour input since 2007 illustrates some worrying developments. While average hours of work appear to have more or less reverted to pre-crisis trends (contrary to developments in the UK and Japan), the unemployment rate remains high and the participation rate is significantly below pre-crisis levels. This is largely masked by the continued growth of the labour force driven by demographic developments. The total shift in labour input is broadly in line with that observed in the UK, and significantly less than the labour losses incurred in Ireland and Spain. Labour input in the US has also declined by less than that in Japan, although this is a reflection of differences in the demographic developments in the two countries. Labour input in Germany and France, on the other hand, has increased since 2007. Of the countries illustrated in figure 1, only in the US and Ireland do we see strong evidence that the rate of labour force participation has deteriorated significantly since 2007. This may have long-term implications for the productive capacity in these economies, as we discuss in the context of the US below.

[FIGURE 1 OMITTED]

Labour force participation

Figure 2 illustrates the steady deterioration of labour force participation in the US since the onset of the financial crisis. This decline is in contrast to the general upward trend over a longer time horizon since 1970, and in contrast to developments in most other advanced economies since 2007, as discussed above. Labour force participation in the US increased steadily between 1970 and 1990, (3) and was more or less stable over the following decade. The downward drift between 2001-2003 was never recovered, and it is worrying that we have yet to see any sign of a correction to the steep decline in participation since 2009, whereas the unemployment rate has started to recede and average hours of work have reverted towards pre-crisis levels. The CBO suggests that some of this drift reflects early retirement and other withdrawals from the labour force, highlighting the rise in applications for disability benefits. Extended periods of unemployment are also associated with disconnection from the labour force of discouraged workers. While not necessarily permanent, these types of withdrawals may prove protracted.

We assess the potential impact of the recent 2 percentage point shift in labour force participation, under the assumption that the full shift in labour force participation proves long-lasting. This is clearly a strong assumption, and somewhat more pessimistic than the assumptions of the CBO, who expect labour force participation to regain 2008 levels by 2021. However, a recent study by Aaronson et al. (2012) concludes that labour force participation may continue to decline through 2020, reflecting shifts in the demographic composition of the labour force. In this case we may well be underestimating the ongoing shock to labour input in the US. We have not observed anything like the decline in labour force participation since 2009 during previous recessions or periods of high unemployment, so it would be difficult to ascribe the development to 'normal' cyclical behaviour that will necessarily be reversed as economic conditions improve. Estimates based on NiGEM simulations indicate that a 2 percentage point decline in labour force participation would reduce the level of potential output in the US economy by about 2 per cent compared to where we expected it to be in 2007. These estimates are somewhat bigger than those suggested by the CBO, which attributes 1/2 per cent loss of potential output to labour force withdrawal, and should be viewed with an appropriate degree of uncertainty.

[FIGURE 2 OMITTED]

Equilibrium rate of unemployment

There is also some evidence that the equilibrium rate of unemployment, often referred to as the NAIRU, has increased since the onset of the financial crisis. Over the period 2000-2007, an unemployment rate of not more than about 5 per cent was generally thought to be sustainable in the US. The Federal Reserve now estimates that a medium-term target of about 6 per cent is more likely. This is in line with the assessment of the NAIRU published by the CBO, which is illustrated in figure 3. The 'short-term' NAIRU, which incorporates structural factors that are temporarily boosting the equilibrium rate of unemployment, is estimated to be roughly 1 percentage point higher than it was in 2006, while the 'long-term' NAIRU has also increased according to these estimates by about 1/2 percentage point.

Long-term unemployment is at exceptionally high levels, as illustrated in figure 4. The median duration of unemployment exceeds 20 weeks, significantly higher than at any other point since at least 1950 for which comparable statistics are available, while the mean duration of unemployment stands at 40 weeks. The rise is partly linked to the extension of eligibility for unemployment benefits from the standard 26 weeks to up to 99 weeks in 2009. Extended periods out of work are linked to a permanent rise in the equilibrium rate of unemployment, as skills depreciate and connections to the workforce erode. Significant structural shifts in the economy, such as that observed in the construction and housing market, may also be associated with protracted periods of high unemployment, as the reallocation of labour across sectors may be obstructed by skill mismatches. A sharp rise in the share of homeowners suffering from negative equity --about 20 per cent of those with mortgaged homes--may also affect the mobility of labour in the short to medium term, delaying the reallocation of resources across locations to their most productive uses, which may raise the equilibrium level of unemployment for a protracted period.

[FIGURE 3 OMITTED]

[FIGURE 4 OMITTED]

According to NiGEM simulations, a 1 percentage point rise in the equilibrium rate of unemployment can be expected to reduce potential output by about 1 per cent compared to its expected level before the onset of the crisis. (4)

Average hours

While average weekly hours of work have reverted to pre-crisis norms, the share of the population working part-time for economic reasons (ie involuntarily) remains stubbornly high. This may reflect a shift in preferences, perhaps as a precaution against the heightened risk of periods of unemployment. Alternatively, it may point to some form of measurement error in the assessment of average weekly hours of work. If average working time remains below levels indicated, the shock to labour input suffered since the recession may prove larger than our current estimates suggest.

[FIGURE 5 OMITTED]

Risk premia and capital demand

We have argued many times in previous issues of this Review that the financial crisis has been associated with a reappraisal of risk, and that this rise in risk premia has driven the long-term expected user cost of capital upwards, leading firms to reassess their desired level of capital. Barrell (2009) illustrates that the magnitude of the output scars suffered from this channel depend both on the magnitude of the rise in the user cost of capital and on the initial capital-output ratio of the economy. More capital intensive economies such as the US can be expected to suffer greater losses as the economy adjusts to a lower capital-output ratio.

Figure 6 illustrates the spread between corporate and government bond yields in the US, which we use as a proxy for the corporate sector borrowing premium to capture both the price and availability of credit to firms. Corporate spreads in the US have receded from the peaks reached in 2008-9, but remain elevated compared to their average level in 2000-2006. We interpret this as a rise in the risk premium on corporate borrowing relative to what had been expected and used in firm development plans and bank lending decisions during the period leading up to the financial crisis.

If we assume that this reappraisal of risk has raised the effective user cost of capital by about 0.6 percentage points on a permanent basis, compared to where we expected it to be before the financial crisis, NiGEM simulations indicate that this will have reduced potential output in the US by about 1/2 per cent. The CBO estimates also include a scar from the loss of capital accumulation, although they do not explicitly offer an explanation for this loss, only that the rebound in investment "will probably not be enough by 2021 to offset all of the capital accumulation that was foregone during the recession and early recovery". Their estimates suggest that this reduces the potential level of output in the US by about 1 per cent - roughly double our estimates. We should allow for a margin of error around our estimates.

[FIGURE 6 OMITTED]

Oil price and productive capacity

While it would be difficult to attribute the rise in the oil price over the past four years to the financial crisis and global recession, the price of energy nonetheless plays an important role in determining the productive capacity of the economy. The unexpected upward shift in prices must cause us to reassess the level of potential output in the US compared to where we expected it to be in 2006. Figure 7 illustrates the oil price assumptions underlying NIESR's current forecast, compared to our average assumptions in 2007. While we observed significant volatility at the height of the financial crisis itself, the dramatic rise in the trend level of the oil price over this period is striking. The average price in 2012 is expected to be roughly $55 per barrel higher than anticipated in 2007--a rise of closes to 90 per cent in the price level. While it is difficult to say with any certainty that the observed rise is permanent, pricing in the futures markets is currently consistent with this assumption.

[FIGURE 7 OMITTED]

It is difficult to explain the sharp rise in the price of oil over the last several years using simple models. Some studies have argued that the shift partly reflects an unexpected rise in demand for energy from China and India, and that this may be associated with an unexpected rise in demand for other goods and services in these countries. The stimulus from a global demand shock may partly offset the negative impact of the price rise itself, which suggests that treating the rise as exogenous, as we do in this study, may overstate the associated decline in capacity. These estimates, therefore, should be interpreted with caution.

In the longer term, we can expect the shift in the relative price of factor inputs to induce firms to shift towards more energy-efficient production processes and partially substitute the oil input with other factors of production, such as labour and capital. A shift in the relative price of oil will lead to a decline in demand for oil as a factor input and a relative increase in demand for labour and capital. If the initial factor bundle was optimal this will lead to a decline in the productive potential of the economy and a permanent loss of output. The magnitude of the long-term impact on output is driven by the oil intensity of production, which is also given by [alpha] in our production function described in equation (1) above.

Barrell, Delannoy and Holland (2011) relate the expected loss of trend output in response to a permanent oil price rise to the oil intensity of production across countries. As production in the US is much more energy intensive than in countries such as the UK and France, losses are expected to be greater in the US than elsewhere. (5) According to our NiGEM simulations, the rise in the trajectory of oil prices since 2007 may have reduced the potential level of output in the economy by up to 4 per cent compared to where we expected it to be in 2007. This reassessment of the productive capacity of the US economy is an important factor that most production function based studies of the US output gap, such as the estimates of the CBO, are lacking.

Measuring the output gap

Output gap measurement is typically addressed either through filtering techniques that extract a trend from the actual and projected level of GDP, or through production function techniques, such as we adopt in this study. The Hodrick-Prescott (HP) filter is the most commonly used tool, although Barrell and Sefton (1995) demonstrate that HP filters are subject to severe endpoint problems, especially if the endpoint is affected by an unknown structural break. The HP filter tends to follow the data too closely, because it is a symmetric filter. This point is illustrated by the HP filter around US output produced by Bullard (2012), which suggests that the US output gap had closed in the second half of 2010, an assertion that would be difficult to reconcile with an unemployment rate of 9 1/2 per cent and a level of output that remained below its pre-crisis peak.

The production function approach that we adopt is similar to that of the OECD, the IMF and the CBO, with our measure of trend output related to inputs of labour, capital and oil, as discussed above. In order to calibrate an estimate of the level of trend output we require two assumptions--the total magnitude of the scar on potential output that the economy has suffered since 2007 and underlying trend rate of capacity growth abstracting from this loss. We estimate underlying trend growth through the assumption that labour force participation and average hours of work remain constant at the average levels of 2006-7, while relative factor prices also remain stable, so that the desired capital-output ratio and energy intensity of production are constant. Trend growth can then be described through demographic developments and through an assumption on the trend rate of productivity growth.

While endogenous growth theory tells us that productivity gains are not exogenous, and indeed may be slower at a lower level of investment, we make the simplifying assumption that the technology frontier has the capacity to expand by roughly 1.8 per cent per annum, in line with its historical average over the 15-year period leading up to the financial crisis. Actual labour productivity growth in the US since 2008 has been somewhat less than this on average, but the sample period is rather short to establish a shift in the trend. In any case, if the scar on output exceeds the scar on employment, we would expect a decline in the level of productivity that may not necessarily affect the underlying trend rate of growth. Nonetheless, this suggests that our estimates may slightly overestimate trend growth since the crisis. The CBO makes an explicit assumption on the effect of recession on productivity growth until 2014, arguing that the downturn may delay the reallocation of resources to their most productive uses, slow the rate at which workers gain new skills as technologies evolve and curtail businesses' spending on research and development. They estimate that the gross impact on the level of potential output from the decline in productivity growth is about 1/2 per cent, implying a decline of not more than 0.1 percentage points per annum for five years.

In figure 8, we illustrate the path of actual GDP in the US, and compare this to a series of estimates for capacity output, cumulating the potential scars on output discussed in this note. These can be decomposed into the components attributable to the decline in labour force participation, the rise in the NAIRU, the rise in risk premia and the rise in the oil price. Our estimates point to a cumulative loss of capacity of up to 7 1/2 per cent since 2007. Of this, about half is attributable to the rise in risk premia and decline in labour input, with the other half attributable to the unexpected 90 per cent rise in the oil price over this period. The loss of labour input probably reflects a combination of lingering effects from the financial crisis as well as effects from the rise in commodity input prices, and it is difficult to distinguish between the two. Studies that focus exclusively on the scarring from the financial crisis are clearly omitting a vital component of factors driving a reassessment of the level of potential output across oil importing nations. There is also a question around the extent to which the calibrated losses to potential output are permanent, or whether they can either self-correct or be corrected through policy intervention. The scars on labour input clearly have the potential to be reversed, although it is not yet clear that this reversal will take place without policy assistance. We would also reiterate the risk around our oil price assumptions as there is a high level of uncertainty around the future path for oil prices.

[FIGURE 8 OMITTED]

Our estimates reported in figure 8 suggest that the US economy is still suffering from a shortfall in demand, with an output gap of at least 2 per cent. Estimates that take no account of the loss of output potential since 2007 would put the gap at something closer to 9 per cent, which is clearly out of line with other measures of spare capacity in the economy, such as the unemployment rate and the rate of capacity utilisation in industry (although caution should be attached to business survey estimates of capacity utilisation, since by definition they only capture the spare capacity of the existing firms and not the wider economy). An output gap of 2 per cent would be significantly smaller than most studies suggest. The CBO puts the US output gap at about 5 3/4 per cent, while the OECD estimates a gap of 3.6 per cent and the IMF a gap of 4.9 per cent. This note highlights the risk that these estimates may overstate the level of spare capacity in the US. An output gap of 2 per cent would not be out of line with other measures of spare capacity in the economy. The unemployment rate is currently about 2 percentage points above the CBO and Federal Reserve estimates of the equilibrium rate. The Federal Reserve's estimate of the rate of total industrial capacity utilisation stands at roughly 1 1/2-2 percentage points below its average in 2006-7. NIESR's forecast presented in this Review sees both the output gap and the unemployment rate gap in the US closing by 2015-16.

REFERENCES

Aaronson, D., Davis, J. and Hu, L. (2012), 'Explaining the decline in the US labor force participation rate', Chicago Fed Letter, March 2012, No. 296.

Barrell, R. (2009), 'Long-term scarring from the financial crisis', National Institute Economic Review, 210, pp. 36-38.

Barrell, R. and Kirby, S. (2011), 'Trend output and the output gap in the UK', National Institute Economic Review, 215, pp. F63-F74.

Barrell, R. and Sefton, J. (1995), 'Output gaps: some evidence for the UK, France and Germany', National Institute Economic Review, 151.

Barrell, R. and Pain, N. (1997), 'Foreign Direct Investment, technological change, and economic growth within Europe', Economic Journal, 107, pp. 1770-6.

Barrell, R., Delannoy, A. and Holland, D. (2011), 'Monetary policy, output growth and oil prices', National Institute Economic Review, 215, pp. F37-F43.

Bullard, J. (2012), 'The U.S. monetary policy outlook', presented at InvestMidwest Venture Capital Forum, St. Louis, Missouri, April 2012.

CBO (2011), 'Persistent effects of the recent recession on potential output', Box 2.2 in The Budget and Economic Outlook: An Update, August 2011, pp. 54-5.

Ellis, C. and Price, S. (2004), 'UK business investment and the user cost of capital', Manchester School, 72, pp. 72-93.

IMF (2010), United Kingdom: Selected Issues Paper, IMF country report 10/337.

OECD (2010), 'Prospects for growth and imbalances beyond the short term', Economic Outlook, 2010/1.

NOTES

(1) Our estimate of the elasticity of substitution is in line with studies such as Barrell and Pain (1997) and Ellis and Price (2004) and outperforms the standard unit elasticity imposed by a Cobb-Douglas relationship between labour and capital.

(2) A brief overview of NiGEM is given in the Key Assumptions section of the world economy forecast chapter and further detail is available from http://nimodel.niesr.ac.uk.

(3) This is largely a reflection of increased female participation in the labour force.

(4) Technically this is effected by introducing a wedge between wages and prices, sufficient to reduce demand for labour by I per cent and increase the unemployment rate by I percentage point relative to baseline.

(5) The calculation of energy intensity is based on final consumption, which nets out the effects from the oil production sector itself.

doi: 10.1177/002795011222000108