L 2 Supply and Productivity Tools and Growth Diagnostics

Transcription

1 L 2 Supply and Productivity Tools and Growth Diagnostics IMF Singapore Regional Training Institute OT Macroeconomic Diagnostics February 26 March 2, 2018 Presenter Reza Siregar This training material is the property of the International Monetary Fund (IMF) and is intended for use in IMF courses. Any reuse requires the permission of the IMF.

2 Lecture objectives The main objective of this lecture is to acquire tools that will enable one to diagnose where an economy is in the business cycle. To do this we will: Understand the importance of GDP measurement Examine the concepts of potential output and potential growth - important inputs for macroeconomic and development policies Explore tools for the estimation of potential output and output gaps Conduct growth diagnostics and discuss potential impacts of structural reforms MDS 2

3 I. Basic Measurement of GDP 3

4 How is GDP measured? Three different approaches give different perspective on the economy but all are equivalent. Production approach + Sum value added in all sectors of econ. + taxes subsidies Expenditure approach + Cons. (C) + Investment (I) + Exports (X) Imports (M) Income approach + Wages + Interest/Rents + Profits + taxes subsidies MDS 4

5 Nominal and real GDP Nominal GDP production approach is defined as the current market value of all final goods and services of newly produced in the economy during a fixed period How to compute real GDP (i.e. net of price changes), when economy produces countless goods? For example, how do you add a ton of steel and a gallon of milk? The standard approach to measure real GDP is to use prices of a base year as aggregation weights: Y ( prices ) N = å P i *Q i i=1 MDS 5

6 Problems of measuring GDP It is easy to take macroeconomic data for granted, and it is important to have timely well-measured data to help us diagnose the economic situation. Unfortunately, computing GDP accurately is not easy. Some types of production are not accurately measured Non-market goods and services Black market economy underground economy Changes in the quality of the goods produced The choice of base year matters MDS 6

7 What happens when relative prices change? Accounting Identity: Real GDP = Nominal GDP/Price Level Suppose, relative prices of some goods are declining, and demand for those goods is rising as a result. Using old price weights will tend to produce higher estimates of GDP growth than if more recent weights are used. Classic example Demand for IT intensive goods has been rising, while the relative prices of IT goods have been falling. GDP growth rates tends to be biased upward, when measured with fixed weights. MDS 7

8 II. Potential Output: Concepts and Measurements 8

9 What is Potential Output? Output Gap? What is potential output? Potential output is the level of output that is consistent with a natural rate of unemployment and stable inflation, i.e. with full capacity utilization. The Output gap is the difference between actual GDP (Y t ) and potential GDP (Y t* ), usually expressed as a percentage of potential output. Yt Y Output gap 100% Y * t * t MDS 9

10 What does it mean to have an output gap? In the short term, output can deviate from potential and move with a business cycle. A positive output gap would signify an unsustainable inflationary expansion that is placing excessive pressure on resources. A negative output gap would signify idle resources low capacity utilization of capital stock and higher unemployment that is likely to lead to declining rates of inflation or to deflation. In the long-term, output is expected to be determined by the potential. MDS 10

11 Illustration of the output gap Output Gap: Illustration Output Positive output gap Negative output gap Actual output Potential output Time MDS 11

12 Illustration of Business Cycles Business Cycles: Illustration Business cycle stands for fluctuations of aggregate demand and supply over time resulting in changes in economic growth Output Boom Recession Trough Recovery Potential output Time MDS 12

13 Overview of Estimation Methods How can potential output be estimated? Purely statistical methods or time series filtering techniques Deterministic time trend Hodrick Prescott (HP) filter Band Pass filters More advanced filters exist (e.g. Butterworth) Quasi-theoretical approaches: Production function approach Multivariate models linking the unemployment and output gaps to inflation (e.g. Benes et al 2010) MDS 13

14 Time series filtering techniques The deterministic time trend as a fitted regression line Deterministic time trend approach o A very simple approach is to estimate the following regression, using OLS: y t t t, with yt ln( Yt ). o With the deterministic time trend approach, potential output is the predictable part of the regression: y * t ˆ ˆ t * and Yt* exp( yt ). MDS 14

15 Time series filtering techniques Interpreting a deterministic time trend United States, Quarterly data, Deterministic time trend based on 1947:3-2015:4 data o Interpret the graphs. o What can you say about the growth rates before and after 2008? MDS 15

16 Time series filtering techniques What is the Hodrick-Prescott (HP) filter? The HP filter finds the series y* that solves: T 2 T 1 * * * * * t t t 1 t t t 1 min y y λ y y y y t 0 t 2 2 Output gap Measure of trend variation Where: y* = log(trend output); y = log(actual output); λ determines the degree of smoothness of the trend. MDS 16

17 Time series filtering techniques Choosing λ for the HP filter The case of λ=0: zero weight on smoothness If λ = 0, then only the output gap is minimized, and so the filtered data exactly parallels the actual data. MDS 17

18 Time series filtering techniques Choosing λ for the HP filter: Rule of thumb for US data Setting λ (USA data). 2 o By convention, 100q where q = periodicity of the data. E.g. λ= 100 for annual, and λ= 1600 for quarterly. o This rule was calibrated to be consistent with the average length of a recession/boom in the USA as well as its variability. o The rule might not apply to more volatile economies. MDS 18

19 Time series filtering techniques: Rule of thumb for λ when HP filtering: The US example Typical weight on smoothness for quarterly data (λ=1600) MDS 19

20 Time series filtering techniques Shortcomings of the HP filter Shortcomings of the HP filter: o It is not clear what value of λ should be used o The HP filter suffers from an end-sample bias o Like other time series filters, it lacks a theoretical foundation MDS 20

21 Time series filtering techniques: The Band-Pass (BP) filter: What is it? o The BP filter isolates the component of a time series that lies between a particular band of frequencies Christiano and Fitzgerald (CF), Christiano and Fitzgerald (1999). The Band Pass Filter NBER Working Paper o The typical bounds for GDP s short term fluctuations lie between 6 and 32 quarters (1.5 and 8 years) NBER s estimates for the frequency of U.S. business cycles. o If interested in the long-run growth look at lower frequencies MDS 21

22 The Production function approach MDS 22

23 Challenges with time series filtering call for more theorybased estimation approaches Purely statistical methods applied to historical data on GDP levels have been criticised in that they can misidentify boom and bust periods End-point bias, choice of parameters, theoretical foundation.. Also many EMEs tend to be subject to structural changes that cannot be identified by filtering techniques In these cases, production function-based methodologies with somewhat stronger theoretical foundations may be preferred, But as noted below, the production function approach may also need to rely on time series filters 23

24 What are the determinants of potential growth? Potential output growth depends on long-term developments of factors of production: Physical capital buildings, machinery and equipment, natural resources Employment labor force and participation rate Human capital knowledge and skills of the labor force Total factor productivity (TFP) amount of technical know-how available and economic structure determine how efficiently capital and labor are used to produce goods and services 24

25 Factors of production Production Function Approach: Big Picture Labor (L) Produced Y=AF(K, L) Capital (K) Natural Human Output (GDP) Productivity (A) or Solow Residual Technology Institutions 25

26 Estimating potential GDP with the production function approach Special case of production function: Cobb-Douglas Y t A t K t L 1 t or lny t ln A t ln K t (1 ln L t o 0<α<1 output elasticity of capital and capital income share o Similarly, 1- α is the output elasticity of labor and the labor income share o Since α+(1- α)=1 constant return to scale: doubling capital and labor endowments doubles the output 26

27 Estimating potential GDP with the production function approach lny * t ln A * t ln K t (1 ln L * t Potential output Trend Total factor productivity (TFP*) Capital s share in national income Capital stock Labor input based on NAIRU Computation of Potential Output: 1. Assume that actual capital stock represents potential utilization of capital stock (K*) 2. Substitute actual labor input with potential labor input (L*) which corresponds to a NAIRU (non-accelerating inflation rate of unemployment) 3. Substitute actual TFP with a filtered trend TFP series (TFP*) 27

28 Estimation of potential labor input (L*) The labor force is the economically active population all persons aged 15 years or over who are classified as employed or unemployed according to the ILO methodology Employment : L t = WAP t LFPR t (1-NAIRU t ) Total Hours : L t = WAP t LFPR t (1-NAIRU t ) AHW t where WAP : Working age population LFPR: Labor force participation rate AHW: Average Hours Worked 28

29 Estimation of potential labor input The reduced-form Phillips Curve approach: Directly links inflation to the unemployment gap* t * b ( u (This requires separate estimates for π* and b) t NAIRU t ) The filtering approach: It consists of applying time series filters to the unemployment rate series. But it has no explicit link to inflation. 29

30 Estimation of the capital stock Perpetual inventory method The method utilizes an estimate for the capital stock (K) in a given initial year, and builds up a capital stock series from that date on. Where I t is Investment in year t, δ is depreciation K K (1 ) 1 I t t t 30

31 Estimation of the TFP trend TFP, also known as the Solow residual, is calculated as: L Yt K At 1 t t TFP has trend and stochastic components. The TFP trend could be estimated with time series filters. 31

32 The global picture of long-term growth: How strong has growth been? Over time: o Median long-term growth has become weaker in AEs and stronger in EMDEs. This prompts two questions: What are the structural factors inhibiting growth in AEs? Can the drivers of growth in EMDEs be expected to last? WEO Chapter: Resilience in emerging markets and developing economies: Will it last? IMF October

33 Growth Accounting: Key diagnostic questions Are there structural problems inhibiting potential growth? Is potential labor declining because of a rising NAIRU or a falling labor participation rate? Do increased social benefits or stronger unions lead to greater disincentives for labor supply? What is the role played by institutional quality and the business climate on TFP growth? Do strong capital growth and negative TFP growth signal misallocation of investment? 33

34 Growth Rate Representation Take logs in Cobb-Douglas production function: logy t = loga t +α logk t + (1-α) logl t Take difference with previous year: log(y t /Y t-1 )= log(a t /A t-1 )+α log(k t /K t-1 )+(1-α) log(l t /L t-1 ) Rewrite in terms of growth rates: g Y = g A + α g K +(1-α) g L Remember for any variable x: Output growth Growth of capital log(x t /X t-1 )=log(x t-1 *(1+g x )/X t-1 ) =log(1+g x ) g x Growth of total factor productivity (TFP) Solow residual Growth of labor 34

35 Description of the growth accounting exercise The growth rates of capital (gk), labor (gl) and output (gy) are known from national statistics Given that α is calibrated to actual data, the growth rate of TFP (ga) is determined as a residual As such TFP picks everything other than factor inputs: Advances in technology and productive inefficiencies Institutional changes, war, and political instability Unrecorded factor inputs (e.g. natural resources) Any mismeasurement of K or L 35

36 Diagnosing growth with growth accounting in emerging economies Svirydzenka and Petri (2014): Mauritius: The Drivers of Growth Can the Past be Extended? 36

37 China: Growth Accounting Average annual rate of change, in percent Source: OECD Economic Survey, China,

38 Determinants of the labor force input Factors affecting the labor force Dynamics of the working age population (WAP) Dynamics of the labor force participation rate (LFPR) Factors affecting the employment rate Flexibility of the labor market (e.g. employment protection; use of temporary contracts) Additional mandated costs of labor (e.g. large employer taxes or severance payments) Skills and geographical mismatches 38

39 Selected determinants of investment and TFP growth Quality of institutions Business climate Trade openness Quality of public infrastructure 39

40 III. Structural Reform and Potential Growth

41 Structural reforms: Policy measures that reduce or remove impediments to the efficient allocation of resources. Business climate: technology and innovation, industry and business regulations, legal systems and property rights reforms Financial sector: domestic financial markets, banking sector, and capital account reforms Fiscal and infrastructure reforms Real sector reforms: product/trade/labor markets reforms, agricultural and other sectoral reforms 41

42 How do structural reforms affect potential growth? Factors affecting potential growth: o saving/investment rate o public infrastructure o investment in human capital o potential labor input o TFP growth Y AK 1 L Initial conditions: o geography o natural resource endowments natural capital o culture, social capital structural reforms: o business climate reforms o financial sector reforms o fiscal and infrastructure reforms o real sector reforms Quality of institutions: o property rights o regulatory quality o corruption 42

43 Reform payoffs by income level 43

44 China: Growth has been led by investment and sectoral allocation TFP (residual factors) growth reflects the re-allocation of labor away from agriculture towards services and manufacturing. The remaining growth of efficiency (or multi-factor productivity, MFP) is estimated to have slowed down Possible explanation: productivity in the state sector has slowed down 44

45 Structural Reform Challenges in the Region I OECD, Economic Outlook for Southeast Asia, China and India 2016 Enhancing Regional Ties 45

46 Structural Reform Challenges in the Region II OECD, Economic Outlook for Southeast Asia, China and India 2016 Enhancing Regional Ties 46

47 Overview of Workshop You will: Estimate potential GDP, the output gap and potential GDP growth using the production function approach. Assess contributions to growth Diagnose supply conditions in Indonesia Identify bottlenecks to higher growth. 47

48 Main Takeaways Overtime, potential growth has become less volatile in world economies. It has become smaller in advanced economies and stronger in emerging and developing economies. As it separates the different contributions of capital, labor, and TFP to GDP, growth accounting provides a basis for diagnosing bottlenecks to growth Structural reforms can boost potential GDP growth if they are designed to address identified key bottlenecks to growth 48

49 Appendix 49

50 Time series filtering techniques: The CF s recommended approximation to the ideal BP filter The Christiano and Fitzgerald (CF) s recommended BP filter approximates the ideal BP filter s series y t with y, given only a finite number of observations for : y B x B x... B x A x t 0 t 1 t 1 T 1 t T 1 T t T x t + B x... B x A x, 1 t 1 t 2 2 t 1 1 t where the B j, and A j s are trigonometric functions of p l and p u. See Christiano and Fitzgerald (1999) for the exact formulas of the B j s and A j s. MDS 51

51 How do policies and institutions affect economic growth? lus/ /jep