a global macroeconomic forecasting model for the philippines ruperto majuca, ph.d (illinois), j.d....
TRANSCRIPT
A Global MacroeconomicForecasting Model for the Philippines
Ruperto Majuca, Ph.D (Illinois), J.D.
De La Salle University, Manila
51st Philippine Economic Society Annual MeetingNovember 2013 (Makati City, Philippines)
Outline
Introduction The Model’s Stochastic Equations Estimation Methods Estimation Results Summary of Findings and Conclusions
Table of Contents
Introduction The Model’s Stochastic Equations Estimation Methods Estimation Results Summary of Findings and Conclusions
Motivating Questions How does a slowdown in U.S. or a U.S. debt
default affect PH economy, directly & indirectly via effects on EU, China, Japan, ASEAN?
■ How does a debt crisis in EU, or China slowdown, affect U.S., China, Japan, ASEAN, and PH directly & indirectly?
■ What has greater impact on PH, shocks from the U.S., EU, China, Japan, ASEAN, or its own shocks?
■ What are the ripple effects of the shocks to Philippine GDP, unemployment, inflation, interest rates, exchange rates, etc.?
Research Interests
Economic & financial linkages PH with ASEAN, U.S., EU, China, Japan, & those economies’ linkages with each other
■ Transmission of shocks from U.S., E.U., Japan, and China to ASEAN, & PH
■ Quantifying the ripple effects to ASEAN & AMSs’ GDP growth, inflation, interest rates, exchange rate, & unemployment
■ Implications for policy and macroeconomic management
Research Methodology, 1 Traditional PH models (equation-by-equation OLS,
ECM)NEDA QMMPIDSAteneo (AMFM), others
Simultaneity bias, exogeneity issueEstimates are biased and inconsistentIncreasing sample cannot cure bias in estimates
Lucas (1976) critiqueCoefficient estimates are not policy invariantLucas: conclusions and policy advice based on
these models are invalid and misleading
Research Methodology, 2 Post Lucas critique. Now standard: modern,
dynamic quantitative economics Dynamic stochastic general equilibrium (DSGE
models)Global projection models
Utilizes state of the art: Bayesian methods This work: global projection model to analyze
interplay of key macroeconomic variables across countries/regionsU.S., E.U., Japan, China, ASEAN, PHGDP growth, inflation, interest rates, exchange rate,
unemployment
Designed to capture cross-regions and cross-country macroeconomic linkages (e.g., US, EU, Japan, China, ASEAN, AMS)
Traces cross-border ripple effects of key macroeconomic variables (GDP growth, inflation, interest rates, exchange rates, unemployment)
Bayesian estimation techniquesPriors plus Bayesian updating via Kalman filter;
Markov Chain Monte Carlo
The Global Projection Model
Table of Contents
Introduction
The Model’s Stochastic Equations Estimation Methods Estimation Results Summary of Findings and Conclusions
Potential Output
NAIRU
Equilibrium Real Interest Rate
GPM Stochastic Equations, 1
𝑟𝑖,𝑡 = 𝑅𝑖,𝑡 − 𝑅𝑖,𝑡
Output Gap (Aggregate Demand / IS Curve)
Inflation (New Keynesian Phillips Curve)
GPM Stochastic Equations, 3
Policy Interest Rate (Taylor Type Rule)
Uncovered Interest Parity (Bilateral Real Exchange Rate)
Unemployment Rate
GPM Stochastic Equations, 4
Table of Contents
Introduction The Model’s Stochastic Equations
Estimation Methods Estimation Results Summary of Findings and Conclusions
Bayesian Estimation
Mixture between classical estimation and calibration of macro models
Puts some weight on the priors and some weight on the data
Combine prior and MLE estimation via Kalman filter
Recover posterior distribution via MCMC (Metropolis Hastings)
Estimation Strategy
Start with GPM4 (US, EU, Japan, China); estimate coefficients
■ Proceed with GPM5 (US, EU, Japan, China + ASEAN), fixing coefficient for GPM4. Assumes ASEAN doesn’t change GPM4 coefficients
■ Then proceed with GPM6 (US, EU, Japan, China, ASEAN + Philippines), mutatis mutandis
■ 250,000 MH draws each stage; first 30% used as burn-in
Data Requirements
Consumer price index Real gross domestic product Nominal interest rate Nominal exchange rate Unemployment rate Bank lending variable for US CPI, GDP and ER are in logs
Table of Contents
Introduction The Model’s Stochastic Equations Estimation Methods
Estimation Results Summary of Findings and Conclusions
Estimation Results: GPM5 Parameters
Prior distribution
Prior mean
Prior s.d. Posterior mode
s.d.
alpha1_AS beta 0.750 0.1000 0.8221 0.0356 alpha2_AS gamm 0.100 0.0500 0.0687 0.0160 alpha3_AS beta 0.500 0.2000 0.4512 0.0548 beta1_AS gamm 0.650 0.1000 0.6353 0.0247 beta2_AS beta 0.150 0.1000 0.0943 0.0320 beta3_AS gamm 0.150 0.1000 0.0690 0.0126 gamma1_AS beta 0.750 0.1000 0.9430 0.0155 gamma2_AS gamm 1.100 0.1000 1.0857 0.0346 gamma4_AS gamm 0.500 0.2000 0.4719 0.0576 lambda1_AS beta 0.500 0.1000 0.6299 0.0296 lambda2_AS gamm 0.400 0.1000 0.3774 0.0198 lambda3_AS gamm 0.050 0.0100 0.0480 0.0032 rho_AS beta 0.500 0.2000 0.0110 0.0689 phi_AS beta 0.600 0.1000 0.6625 0.0258 tau_AS beta 0.050 0.0200 0.0427 0.0049 rr_bar_AS_ss norm 1.500 0.1000 1.4835 0.0494 growth_AS_ss norm 5.000 0.2000 5.0157 0.0651 beta_reergap_AS gamm 0.050 0.0200 0.0472 0.0074
Prior
distribution Prior mean
Prior s.d.
Posterior mode
s.d.
RES_PIE_AS invg 3.000 Inf 3.6982 0.4221
RES_Y_AS invg 0.500 1.0000 0.2406 0.1084
RES_RS_AS invg 0.600 1.0000 0.2213 0.0347
RES_LGDP_BAR_AS invg 0.200 Inf 18.5265 0.9173
RES_G_AS invg 0.100 Inf 0.0460 0.0381
RES_RR_BAR_AS invg 0.200 Inf 0.1877 0.5291
RES_UNR_GAP_AS invg 0.600 1.0000 0.2488 0.0478
RES_UNR_BAR_AS invg 0.100 Inf 0.0461 0.0493
RES_UNR_G_AS invg 0.100 Inf 0.0472 0.0237
RES_LZ_BAR_AS invg 5.000 Inf 4.8714 0.8198
RES_RR_DIFF_AS invg 1.000 Inf 0.4591 0.3230
Estimation Results: GPM5 S.D. of Structural Shocks
Estimation Results: GPM6 Parameters, 1
Prior distribution
Prior mean
Prior s.d. Posterior mode
s.d.
alpha1_PH beta 0.750 0.0500 0.7810 0.0434
alpha2_PH gamm 0.100 0.0500 0.0882 0.0503
alpha3_PH beta 0.500 0.2000 0.4673 0.3086
beta_fact_PH gamm 0.150 0.1000 0.1171 0.1024
beta1_PH gamm 0.650 0.1000 0.5710 0.0806
beta2_PH beta 0.150 0.0500 0.1234 0.0446
beta3_PH gamm 0.150 0.0200 0.1310 0.0181
gamma1_PH beta 0.900 0.0500 0.9101 0.0207
gamma2_PH gamm 1.100 0.5000 0.8872 0.3600
gamma4_PH gamm 0.500 0.2000 0.4034 0.1745
growth_PH_ss norm 5.000 0.2000 5.0000 0.2000
lambda1_PH beta 0.500 0.0500 0.5616 0.0477
lambda2_PH gamm 0.400 0.1000 0.3522 0.0876
lambda3_PH gamm 0.050 0.0300 0.0390 0.0292
lambda1_RS_PH beta 0.500 0.1000 0.4469 0.0868
Estimation Results: GPM6 Parameters, 2phi_PH beta 0.600 0.0500 0.6303 0.0364
pietar_PH_ss gamm 4.714 0.3000 4.6951 0.2994
rho_PH beta 0.500 0.2000 0.2675 0.1123
rr_bar_PH_ss norm 1.500 0.5000 1.5000 0.5000
tau_PH beta 0.050 0.0200 0.0436 0.0188
beta_reergap_PH gamm 0.050 0.0100 0.0480 0.0098
chi_PH beta 0.050 0.0100 0.0481 0.0098
growth_PH_ss norm 5.000 0.2000 5.0401 0.1632
pietar_PH_ss gamm 4.714 0.3000 4.6951 0.2994
rr_bar_PH_ss norm 1.500 0.5000 1.4629 0.4574
beta_reergap_PH gamm 0.050 0.0100 0.0503 0.0097
Estimation Results: GPM6 S.D. of Structural Shocks
Prior distribution
Prior mean Prior s.d. Posterior mode
s.d.
RES_PIETAR_PH invg 0.250 Inf 0.1028 0.0338
RES_PIE_PH invg 3.000 Inf 3.2548 0.4240
RES_Y_PH invg 0.500 1.0000 0.4392 0.0933
RES_RS_PH invg 0.600 1.0000 0.2533 0.0486
RES_LGDP_BAR_PH invg 0.200 Inf 0.0900 0.0353
RES_G_PH invg 0.100 Inf 0.0442 0.0168
RES_RR_BAR_PH invg 2.500 Inf 1.8752 0.4157
RES_UNR_GAP_PH invg 1.000 1.0000 0.9615 0.1265
RES_UNR_BAR_PH invg 0.100 Inf 0.0464 0.0192
RES_UNR_G_PH invg 0.100 Inf 0.0477 0.0210
RES_LZ_BAR_PH invg 5.000 Inf 5.7296 1.3704
RES_RR_DIFF_PH invg 1.000 Inf 0.4587 0.1857
RES_DOT_LZ_BAR_PH invg 0.100 Inf 0.0461 0.0188
Table of Contents
Introduction The Model’s Stochastic Equations Estimation Methods Estimation Results
Summary of Findings and Conclusions
Findings and Conclusions, 1 Existing PH models (NEDA QMM, PIDS, AMFM,
etc.) using equation-by-equation OLS, ECM)Simultaneity bias/inconsistency issueLucas (1976): coefficients not policy invariant;
conclusions & policy advice are invalid and misleading
Now standard: modern, dynamic quantitative economics (utilizing Bayesian methods)Dynamic stochastic general equilibrium (DSGE
models)Global projection models
Findings and Conclusions, 2
This work: cross-region ripple effects to key macro variables (GDP growth, inflation, unemployment, etc.) traced via GPM
Greatest influence on ASEAN macroeconomic variables come from ASEAN’s own internal shocks; followed by shocks from U.S., China, Japan, then Euro area, in that orderASEAN own AD shocks’ impact on ASEAN
GDP, 0.4 ppt; US AD impact (peaks after 5 or 6 quarters), about 1/7 of ASEAN impact; China AD shock, about 1/9 ASEAN’s; Japan, 1/10; EU, 1/11
Findings and Conclusions, 3 For AMS like PH, domestic shocks also capture
much of the influence on own macroeconomic variables. In the case of PH, this is followed by shocks from the U.S., Japan and China, then ASEAN and Euro area.PH AD shock’s impact to PH GDP, about 0.5
percent; US shock’s impact (peaks after about 5 or 6 quarters), about 1/7 of PH shock’s impact; Japan and China, about 1/10; ASEAN and EU, about 1/17.
Findings and Conclusions, 4
■ Impulse responses of PH macro variables Shock to domestic AD results in 0.5% increase on
PH real GDP on impact; positive impact persists for more than 2 years
Results in decrease in umeployment (lasts for ~ 3 years before returning to steady state)
Demand pull increase in inflation Appreciation in currency BSP increase policy rates via Taylor-type reaction
function