an explanation of the summary measures of inflation expectations shown in Chart A on page 31 of the February 2017 Inflation

Transcription

1 BANK OF ENGLAND Mark Carney Governor Rt Hon Andrew Tyrie MP Chair Treasury Committee House of Commons London SW1A0AA 24 April 2017 At the February 2017 Inflation Report hearing on 21 February, the Treasury Committee asked for additional information on two issues that were discussed: the change that the MPC had made to its assumption about the 'equilibrium' rate of unemployment in the February 2017 Inflation Report projections, in particular the history of such changes; and an explanation of the summary measures of inflation expectations shown in Chart A on page 31 of the February 2017 Inflation Report. This letter provides the information requested. The equilibrium rate of unemployment At the Treasury Committee hearing, you asked for more information on the history of the changes that the MPC has made to its assumption about the 'equilibrium' rate of unemployment. The assumption in question concerned what economists sometimes call the 'natural rate', or the 'long-term equilibrium rate', of unemployment. This concept describes the rate at which unemployment is thought likely to settle in the long run, after all of the shocks and disturbances affecting the economy at any moment in time have dissipated. It is the rate of unemployment that the economy is capable of achieving sustainably over a long period of time. This rate is influenced by a number of structural features of the economy. Some examples include: skill levels amongst the workforce and how well they are matched with the skills required by different employers; relevant aspects of the tax and benefit system and employment legislation; how efficiently people can search for appropriate job opportunities, for example in job centres or online; and the nature of industrial relations. Over the years, successive governments have introduced policies aimed at keeping this long-run equilibrium level of unemployment as low as possible.

2 2 The long-run equilibrium rate of unemployment is a different concept from the non-accelerating inflation rate of unemployment, or NAIRU. The NAIRU is a shorter-term measure of the equilibrium unemployment rate that takes into account other temporary factors that can affect the pressure that a given unemployment rate exerts on wages, and hence inflation. For example, if there is a prolonged downturn that results in people being unemployed for an extended period, it may become more difficult for them to find a job as their skills deteriorate. That is likely to mean that they will place less downward pressure on wages. Reflecting that, the NAIRU is likely to rise above the long-run equilibrium unemployment rate and converge back only gradually as demand recovers, and people re-train or move in order to fill available vacancies. Due to the additional effects from these temporary factors, the NAIRU is likely to be more volatile than the long-run equilibrium unemployment rate. In practice, of course, the economy is never perfectly at rest and unemployment never settles stably at the hypothetical long-run equilibrium rate. This makes it difficult to identify its level with precision. In reality, as time moves on and domestic and global economic conditions evolve, new shocks and disturbances arrive to replace old ones, causing the actual unemployment rate to move around from month to month and over the business cycle. Nevertheless, the concept of the natural (or long-run equilibrium) rate of unemployment remains a very important one for economists because it is a key determinant of the economy's long-run supply capacity, future incomes and living standards. Of particular relevance to the MPC, it is an important concept also because it partly determines the level to which unemployment is able to fall without generating increased pressure on pay growth and inflation. Because the natural rate of unemployment is related to the deep-rooted structural features of the economy, it is typically thought to move relatively slowly over time. Commensurately, changes to economists' estimates of the level of the natural rate tend to be reasonably infrequent. This has also been true of the MPC. There is a relatively widely held consensus amongst UK labour market economists that the major changes in the long-run equilibrium rate of unemployment of the past few decades occurred in the second part of the 1980s through to the early 2000s. During this period the actual unemployment rate trended downwards from around 12% to around 5%, and an active debate took place (including within the Bank, government and academia) about how much of this decline reflected a fall in the long-term equilibrium rate, and how much reflected cyclical and other short-term factors. These were amongst the major modelling and forecasting decisions facing the early MPC following its creation in Along with most other economists, the MPC would have revised downwards its assumptions about the natural rate of unemployment a number of times in its early years. Overall, typical estimates suggest that perhaps 3-5 percentage points of the decline in unemployment from the mid-80s might be attributable to a reduction in the natural rate - a reduction considerably larger than the change in judgement the MPC made in its February Report. However, consistent with the standards of transparency prevailing at that time, these revisions were not publicly disclosed. Following that period but prior to the financial crisis, estimates of the long-term equilibrium rate of unemployment remained comparatively stable, in turn reflecting the relative stability of the UK and global economies (the so-called 'great moderation'). There were minor adjustments to the forecast assumptions made by staff from time to time, for example reflecting the updated calibration and estimation of macroeconomic models following the release of new economic datasets. But, by and large, the relatively benign economic conditions - and in particular the lack of need for the MPC to balance any significant trade-off between returning inflation to the target and supporting activity and employment - meant that the estimate of the natural rate, while important, was simply not a primary policy concern or particular focus of

3 3 the Committee. Again, these deliberations were not publicly disclosed and no public estimate of the natural rate was released. Following the financial crisis, aggregate demand, having contracted sharply, recovered only sluggishly, pushing the unemployment rate to around 8 1 /2%. At a similar time, movements in the exchange rate, commodity prices and the rate of VAT pushed inflation well above the target. This presented a marked trade-off for the MPC to manage between returning inflation to the target promptly, on the one hand, and supporting activity and jobs on the other. Naturally, therefore, an estimate of what unemployment rate was actually achievable in the long run became a much more critical input to monetary policy decisions. During this period, there was a question over whether the sustainable long-run equilibrium rate of unemployment might have risen as a consequence of the financial crisis and resulting recession, or whether the persistent weakness of pay growth indicated that it might instead have fallen. The MPC has debated these questions and reviewed its assumptions regarding the natural rate of unemployment frequently since the financial crisis, with its evolving thinking on the underlying drivers laid out in the Bank's Inflation Report and the minutes of its meetings and an estimate of the equilibrium unemployment rate first published in Indeed, as I mentioned at the hearing, in 2014 the MPC instituted annual stock-takes of its overall supply-side assumptions (including the long-term equilibrium unemployment rate) of which the most recent was prepared to coincide with the February Inflation Report. These annual stocktakes serve two purposes. First, they provide the MPC with a regular opportunity to step back from the month-to-month flow of economic data and consider supply-side developments as a whole and over an appropriately lengthy time period. Second, by undertaking routine systematic annual assessments of the supply-side, rather than addressing the issues ad hoc from quarter to quarter, we hoped to provide clarity on our analytical processes to observers outside the Bank. On each of the first three annual stocktakes, the Committee decided to retain its central forecasting assumption of a long-run equilibrium unemployment rate of around 5%. Notwithstanding the usual range of views amongst MPC members, the available evidence did not, on balance, present a clear enough case for revising the central assumption in one direction or the other. During the period, however, a number of MPC members referenced in speeches the possibility that the natural rate might have fallen. 1 No one publicly argued it had risen. Overtime the evidence regarding the natural rate continued to accumulate more decisively in one direction. The actual unemployment rate has continued to decline to a level a little below that prevailing before the financial crisis, while pay growth has remained subdued. This pattern had been apparent for some time. In the Minutes of its February 2016 meeting, for example, the Committee noted that recent data "raised the possibility that there was more slack than assumed and that the economy could therefore function with a rate of unemployment permanently lower than previously estimated." At that time the Committee felt these trends were not marked enough to justify a reduction in its central estimate of the natural rate of unemployment. The pattern has since continued and, coupled with further analysis by Bank staff regarding the impact of the changing composition of the workforce, the Committee concluded in the February 2017 annual supply stocktake that the long-run equilibrium rate of unemployment was likely to be somewhat lower than previously assumed. These factors, set out in more detail in the latest Inflation Report, are what motivated the adjustment to the MPC's forecasting assumption that we discussed on 21 February. 1 See, for example, 'Drag and drop', speech by Andy Haldane (2015), 'The turn of the year', speech by Mark Carney (2016), and 'The economic outlook', speech by Michael Saunders (2016).

4 4 Finally, I should point out that the Committee's assumption on the natural rate of unemployment - important though it is - is only one part of a full assessment of the balance between aggregate demand and supply in the economy, which is itself only a part of a full assessment of medium-term inflationary pressure and therefore the appropriate stance of monetary policy. In the short-to-medium run, a number of other factors influence the inflationary impetus resulting from a given level of unemployment. For instance, as noted earlier, the mix of short and longer-term unemployed is likely to have a bearing on the inflationary or disinflationary pressure associated with a given level of unemployment. For this reason, the Bank has tended to draw a distinction between the long-run equilibrium or natural rate of unemployment (which has been the subject of most of this letter) and the medium-term equilibrium rate which the MPC believes to be generally more relevant in assessing the degree of inflationary pressure over the Committee's three-year forecast period. This distinction is set out in more detail in a box in the August 2013 Inflation Report (pages 28-29). Looking beyond the outlook for unemployment, assumptions regarding the likely evolution of other key elements of labour supply are equally important in assessing the balance of demand and supply capacity in the economy-for instance how labour-market participation versus economic inactivity in different age groups is likely to evolve, trends in household preferences for part-time versus full-time employment, what might happen to the average number of hours individuals work per week, etc. All of these factors are relevant for the outlook for inflation and therefore the appropriate stance of monetary policy. Moreover, moving beyond the labour market, a yet larger set of factors is critical to the MPC's assessment: the intensity of competition in product markets, the impact of structural change within the retail sector, the pace of technological innovation, and the speed with which movements in imported costs are passed through into consumer prices, to name but a few. The MPC assesses the full range of factors that are relevant to the outlook for inflation, making the best assumptions that it can based on the evidence that is available at the time, and with a willingness to challenge those assumptions as further evidence becomes available. This process of continual analytical reassessment and learning has been central to the MPC since its creation in 1997, and it is a great strength of the policy framework, reinforced by the structure of the Committee itself with nine individually accountable members. It is also an important aspect of the framework that the MPC is transparent in communicating its assessment of the economic outlook, how it has changed overtime, and the reasons for those changes of view. I am grateful of the opportunity to have been able to contribute in a small way to that in this letter and I hope the Treasury Committee will find it of some use. Summary measures of inflation expectations Chart A on page 31 of the February 2017 Inflation Report, replicated below as Chart 1, showed summary measures of the levels of inflation expectations at three different horizons: one year, two years, and five to ten years. These summary measures are calculated from a range of indicators of inflation expectations in the UK, as shown in Table A: inflation swaps in financial markets, two surveys of professional forecasters' expectations, three surveys of households' expectations and one survey of companies' expectations. These are the main indicators that the MPC considers when assessing inflation expectations as they have a reasonable back-run of data available and provide quantitative information on inflation expectations. 2 2 There are other measures of inflation expectations available but they do not meet both these criteria. For example, the Deloitte CFO survey includes a question on inflation expectations but this was only introduced in 2013, while the measure of inflation expectations in the GfK/EC consumer confidence survey of households is a net percentage balance constructed from a qualitative question asking about whether inflation will increase relative to its current rate.

5 5 The summary measures of inflation expectations shown in Chart 1 are intended to extract an underlying signal from the range of different indicators. While it is, of course, helpful to consider movements in each indicator separately - the inflation expectations of different groups may be formed in different ways or have differing effects on future inflation outcomes - the individual indicators can be noisy at times (Chart 2). There is, therefore, benefit to looking at summary measures that capture the broad movements in the data while abstracting from the volatility in each series. How are the summary measures constructed? The summary measures of inflations expectations at different horizons are constructed by using statistical methods to extract a common signal from all of the various available indicators. This is done by treating each indicator as being driven by a small number of underlying factors that determine the signal, plus some random noise that reflects other influences including sampling error. The underlying factors are interpreted as representing the level, slope and shape of a curve showing underlying inflation expectations at different time horizons, and are assumed to change smoothly overtime. The signal from each indicator of inflation expectations at each horizon can be thought of as being a weighted combination of these three factors. The weights, and the underlying factors, are estimated simultaneously across the whole range of indicators and horizons, with the model allowing for the fact that different indicators of inflation expectations are available at different horizons. More details can be found in the recently published Kapetanios. Maule & Young (2016) "A new summary measure of inflation expectations", Economics Letters vol. 149, p What measure of inflation does the summary measure refer to? The indicators underlying the summary measures of inflation expectations reference a variety of measures of inflation. As shown in Table A, the only indicators that reference CPI inflation are the surveys of professional forecasters. The indicators derived from financial markets - inflation swaps - refer to RPI inflation. The indicators from surveys of households do not refer to a specific rate of inflation and instead are phrased in more general terms, such as "prices in the shops". This phrasing helps ensure that the questions can be better understood by respondents: in the 2017 Q1 Bank/TNS Inflation Attitudes Survey, 44% of respondents answered "don't know" when asked "what do you think the current rate of inflation is, as measured by the 12 month percentage change in the Consumer Prices Index (CPI)?", whereas only 17% of respondents answered "no idea" when asked how prices have changed over the last twelve months. The survey of companies asks about "the general level of selling prices in the UK markets that your firm competes in". The levels of the inflation expectations indicators used within the summary measure that don't explicitly refer to CPI inflation are adjusted up or down to try to make them more comparable to CPI inflation. Table A shows the precise adjustment made in each case. For most indicators, we start by calculating the difference between CPI inflation and inflation expectations at the shortest horizon measured by that indicator - for example, for the Bank/GfK/TNS Inflation Attitudes Survey, 3 it would be households' perceptions of inflation over the past year whereas for the YouGov/CitiGroup survey, that would be oneyear inflation expectations. We then take the average of this difference over the period from the start date of the indicator to 2013 Q2 and add this to the measures of inflation expectations at all horizons 3 The survey provider changed from GfK to TNS in 2016 Q1.

6 6 covered by that indicator. 4 The reason for applying the same adjustment at all horizons is so that the information each indicator contains about the term structure of inflation expectations is preserved. There are some exceptions to this general procedure. For example, the measure of long-term inflation expectations derived from financial markets - the five-year, five-year forward inflation swap - is reduced by 0.95 percentage points, as this is what information from financial market contacts suggests they expect will be the value of the wedge between RPI and CPI inflation in the long run. 5 Are the current levels of the summary measures consistent with inflation expectations being to the inflation target? well-anchored At the Inflation Report hearing, Chris Philp noted that the five-to-ten year summary measure is currently higher than the 2% CPI inflation target and questioned whether that was consistent with inflation expectations being well anchored to the target. Mechanically, the elevated level of the summary measure at this horizon is largely a reflection of the inclusion of the YouGov/Citigroup survey measure of household inflation expectations. In this survey, inflation expectations at the five-to-ten year horizon have been persistently higher than expectations at both the one and two-year horizons for almost the entire period since the survey began in late 2005 (Chart 3), averaging around 3.2% compared with an average of 2.4% at the one-year horizon. That wedge between short and long-term expectations gets picked up in the estimation of the summary measures and so, all else equal, leads to the five-to-ten year summary measure being higher than the shorter-term summary measures. Moreover, one-year inflation expectations in the YouGov/CitiGroup survey have generally been below CPI inflation. So the adjustment we apply to try to bring the inflation expectations measures from this survey more into line with CPI inflation results in them shifting up at all horizons (Table B). All else equal, that pushes up the longer-term summary measure further still. The adjustment to the indicators of inflation expectations that do not reference CPI inflation is, necessarily, imperfect since it is impossible to know the true mapping between the measures of inflation referenced by these indicators and CPI inflation. When gauging how well anchored inflation expectations are to the inflation target, it is therefore more appropriate to compare the summary measures of inflation expectations to their averages over a period when inflation was close to the target than it is to compare them to 2%. The past averages provide a more useful guide because the fact that CPI inflation was close to the target over that period suggests that inflation expectations were well-anchored at that time - hence deviations from those past averages could provide evidence that they have become less well anchored. For the five-to-ten year summary measure, we compare the current reading with the average over the period from 2006 to 2007, when inflation was close to the target and had been for some time. We cannot start that period any earlier as only one indicator of five-year inflation expectations was available prior to then. 6 The five-to-ten year summary measure is currently very close to its average level (Chart 1). 4 The one exception to this is the Barclays Basix survey: that started in 1986, but the ONS official data on CPI inflation begin in 1989 Q1, so that is the start period used. 5 For a discussion of the factors affecting the long-run RPI-CPI wedge see a box on pages of the February 2014 Inflation Report. 6 The YouGov/Citigroup survey began in November 2005, meaning the first quarter for which we have a full three months of data is 2006 Q1.

7 7 The summary measures are only one of the many metrics that the MPC uses in the assessing the extent to which inflation expectations remain well-anchored to the target. The Committee also gives consideration to how the level of each indicator individually compares with its historic averages and the Inflation Report forecasts. In addition, the MPC considers a broader range of information on inflation expectations, such as the sensitivity of longer-term inflation expectations to short-term economic news and levels of uncertainty about future inflation. The box in the Inflation Report provides information on this wider set of metrics. The MPC continues to judge that inflation expectations remain well anchored but will monitor them closely as inflation rises above the 2% target.

8 Chart 1: CPI Inflation and summary measures of the levels of inflation expectations Per cent Ftve to ten year i 4 expectations Chart 2: Indicators of one-year inflation expectations used in the summary measure and the summary measure Individual indicators* Summary measure Per cent expectations W ^ 2 One year / y 1 expectations Dashed fcnes: averages i_ S * Individual indicators are shown after their adjustment to be more comparable to CPI. Chart 3: YouGov/Citigroup inflation expectations and CPI inflation* CPI irf tatior YouGov/Gtigroup ore-year inflation expectations** YouGev.'Gtigroup five to ten year inflation expectations** Percent 2C06 20OS M 2016 * Data are quarterly averages ot monthly data. ** Data shown after their adjustment to be more comparable to CPI.

9 9 Table A: Indicators of inflation expectations within the summary measure* Indicator used Horizon used Index used/asked about Adjustment made within summary measure** Financial markets Instantaneous 1 year Average difference between 1 year Instantaneous 2 years Inflation swaps RPI and CPI inflation Five-year, five-year forward Reduce level by 0.95pp*** Professional forecasters Survey of External 1 year forecasters, Bank of 2 years CPI None England 3 years Survey of 1 year independent 2 years forecasters, HM 3 years CPI None Treasury 4 years Households 1 year Average difference between perceptions of Bank/TNS Inflation 2 years Prices in the shops how prices have changed Attitudes survey 5 years over the past year and CPI inflation 1 year Average difference Rate of inflation Barclays Basix 2 years between 1 year (unspecified) 5 years and CPI inflation 1 year Consumer prices Average difference YouGov/Citigroup 5-10 years of goods and services between 1 year and CPI inflation Companies Level of selling prices in the UK Average difference CBI Distributive 1 year markets in which between 1 year Trades survey your company and CPI inflation competes * For each of the surveys the median level of inflation expectations is used. Average differences are taken between the start of the available sample period of each series and 2013 Q2. ** As explained above, an adjustment is applied to the measures of inflation expectations that do not explicitly reference CPI inflation to try to make them more comparable with CPI inflation. *** Based on information from market contacts about their expectations for the RPI-CPI inflation wedge in the long run. Table B: Adjustment to YouGov/Citigroup data Average over period* Adjustment applied to each data period** Average over period* after adjustment CPI inflation 3.0% n.a. n.a. YouGov/Citigroup one year 2.7% = 0.3pp = 3.0% inflation expectations YouGov/Citigroup five to ten year inflation expectations 3.4% = 0.3pp = 3.7% * Period used is 2006 Q1, the first quarter for which data are available for all three months, to 2013 Q2. ** As explained above, an adjustment is applied to the measures of inflation expectations that do not explicitly reference CPI inflation to try to make them more comparable with CPI inflation