Transcript SVAR Approach to Monetary Policy Evaluation in India

A SVAR Approach to
Evaluation of Monetary Policy
in India
Soumya Bhadury 1, Taniya Ghosh 2
*1 [email protected]; University of Kansas
*2 [email protected]; IGIDR
July 23, 2015
A SVAR Approach to Evaluation of Monetary Policy in India
1
A SVAR Approach to Evaluation of Monetary
Policy in India
• Introduction and Motivation
• The Structural VAR model
• Empirical Result
• Conclusion
A SVAR Approach to Evaluation of Monetary Policy in India
2
A SVAR Approach to Evaluation of Monetary
Policy in India
• Introduction and Motivation
• The Structural VAR model
• Empirical Result
• Conclusion
A SVAR Approach to Evaluation of Monetary Policy in India
3
The Basic Tenets of the Model
• In majority of the exchange rate literature, interest rate alone plays the role
of the monetary policy instrument
• We try to introduce separately ‘money’ into the system, other than interest
rate
• Our novelty will lie in introducing the theoretically grounded, superlative
measure of money i.e. Divisia monetary aggregate
• We introduce the K&R contemporaneous restrictions customized for the
Indian economy
A SVAR Approach to Evaluation of Monetary Policy in India
4
Why Interest rate the only monetary policy instrument?
• Chrystal and McDonald (1995) claim that the velocity of the monetary
aggregate in some major countries like U.S.A and U.K., took a sharp
downward trend after 1980 ** (See Stone and Thronton, 1987)
• Leeper and Roush (2003) agree with C&M that traditionally stable
money demand functions were widely perceived to have broken down
A SVAR Approach to Evaluation of Monetary Policy in India
5
Why include ‘Money’?
• Ireland finds money plays an informational-role rather than causal-role by
helping to forecast future nominal rate of interest**(Ireland 2001a,2001b)
• Practical consideration suggests including money in the Central Bank’s
policy rule
• Goodfriend (1999) argues that money plays a critical role even under an
interest rate policy because “ credibility for a price-path objective stems
from a central bank’s power to manage a stock of money”
A SVAR Approach to Evaluation of Monetary Policy in India
6
Why include Divisia Money?
• Divisia money measures the aggregate flow of the monetary services derived from a
collection of assets that are imperfect substitutes**(Barnett, 1980)
• C&M (1995) believe that in a period of rapid financial liberalization data dynamics will
be unable to track the exchange rate movements when simple-sum money is the preferred
monetary aggregate
• Relative performance of Divisia over simple-sum has been evaluated in forecasting
inflation**( Schunk, 2001; Drake and Mills, 2005; Stock and Watson,1999)
• We believe Divisia correctly tracks the flow of money services; plays an informational
role critical to capturing exchange rate movements and determination.
A SVAR Approach to Evaluation of Monetary Policy in India
7
Indian economy at a glance
• Indian economy experienced high inflation
in last 24 years
• Monetary policy stance loosened overtime
• Simple-sum M1 & M3 growth shows
significant liquidity injection in the Indian
economy
• Divisia measure shows that simple-sum
over-estimates liquidity injection
• Industrial growth stagnated post 2007
A SVAR Approach to Evaluation of Monetary Policy in India
8
• Introduction and Motivation
• The Structural VAR model
• Empirical Result
• Conclusion
A SVAR Approach to Evaluation of Monetary Policy in India
9
Model
• 𝐵0 𝑦𝑡 = 𝑘 + 𝐵1 𝑦𝑡−1 + 𝐵2 𝑦𝑡−2 + ⋯ + 𝐵𝑝 𝑦𝑡−𝑝 + 𝑢𝑡 is the structural
model
• 𝑦𝑡 is an 𝑛 × 1 data vector,
• 𝑘 is an 𝑛 × 1 data vector of constants
• 𝑢𝑡 is an 𝑛 × 1 structural disturbances vector and is serially and mutually uncorrelated.
• 𝑝 denotes the number of lags.
• 𝑌𝑡 = [ (𝑜𝑖𝑙/𝑤𝑐𝑜𝑚)𝑡 𝑟𝑓𝑒𝑑𝑡 𝑖𝑖𝑝𝑡 𝜋𝑡 𝑀𝐷𝑡 𝑀𝑃𝑡 𝐸𝑅𝑡 ]
A SVAR Approach to Evaluation of Monetary Policy in India
10
Model
• 𝑦𝑡 = 𝑐 + ∅1 𝑦𝑡−1 + ∅2 𝑦𝑡−2 + ⋯ + ∅𝑝 𝑦𝑡−𝑝 + 𝜖𝑡 is the reduced form
model
• 𝜖𝑡 is the reduced form residuals
• 𝑢𝑡 = 𝐵0 𝜖𝑡
𝑜𝑖𝑙/𝑤𝑐𝑜𝑚
• 𝜖𝑡 = [𝜖𝑡
𝑟𝑓𝑒𝑑
𝜖𝑡
𝑖𝑖𝑝
𝜖𝑡
𝜖𝑡𝜋 𝜖𝑡𝑀𝐷 𝜖𝑡𝑀𝑃 𝜖𝑡𝐸𝑅 ]
• It is possible to recover the structural parameters from the reduced form
model
• It requires the model be either exactly identified or over-identified
A SVAR Approach to Evaluation of Monetary Policy in India
11
Structural Shocks
• VAR can be written in terms of the structural shocks
• World oil price shock/ World commodity price shock
• Foreign interest rate shock
• Domestic output shock
• Inflation or cost push shock
• Money demand shock
• Monetary policy shock
• Exchange rate shock
A SVAR Approach to Evaluation of Monetary Policy in India
12
Identification Assumptions
A SVAR Approach to Evaluation of Monetary Policy in India
13
Identification Assumptions
• “Contemporaneously” exogenous world shock variable (alternatively captured using the world
commodity price index and world price index)
• The foreign interest rate is the Fed fund rate, the short term interest rate of the U.S. , in the small
open economy framework is only affected by the world event shocks
• Output and prices do not respond contemporaneously to changes in domestic monetary policy
variables and exchange rates.
• Real activity like the industrial production responds to domestic price and financial signals (interest
rate and exchange rate) with a lag
• The industrial production and inflation of the small, open, economy is deeply impacted by the
world or outside shocks
A SVAR Approach to Evaluation of Monetary Policy in India
14
Identification Assumptions
• Inflation is affected by the world shock and the current state of industrial
production.
• In addition to the real income and the domestic interest rate, the money demand
function also depends on the foreign (US) interest rate and the prevailing
exchange rates
• Monetary policy i.e. interest rate is set after observing the current value of money
supply, the interest rate and the exchange rate
• Exchange rate is one of the most volatile variables in the model and is quick to
react to almost all shocks be it from inside or outside, be it nominal or real
variable shock.
A SVAR Approach to Evaluation of Monetary Policy in India
15
Monetary Aggregates
 M2 = currency with the public + demand deposits with banks + other deposits
with the RBI + time liability proportion of the savings deposits with banks + term
deposits with the contractual maturity of up to and including one year with banks
+ certificate of deposits issued by banks
 M3 = M2 + term deposits with the contractual maturity of over one year with
banks + call borrowings from non-depository financial corporations by banks
 L1 = M3 + all deposits with the Post Office Savings Banks (excluding National
Savings Certificates)
A SVAR Approach to Evaluation of Monetary Policy in India
16
Model Structure
Table 1
Model
Model
Model
Model
Model
Model
Model
Model
Model
Model
Model
Model
Model
Model
Model
1
2
3
4
5
6
7
8
9
10
SVAR Model [Non-Recursive (NR) Structure]
{oilp, rfed, iip, pi, dm3, rdom, er}
(NR, OIL, DM3)
{oilp, rfed, iip, pi, m3, rdom, er}
(NR, OIL, M3)
{oilp, rfed, iip, pi, m1, rdom, er}
(NR, OIL, M1)
{oilp, rfed, iip, pi, dl1, rdom, er}
(NR, OIL, DL1)
{oilp, rfed, iip, pi, dm2, rdom, er}
(NR, OIL, DM2)
{wcom, rfed, iip, pi, dm3, rdom, er}
(NR, COM, DM3)
{wcom, rfed, iip, pi, m3, rdom, er}
(NR, COM, M3)
{wcom, rfed, iip, pi, m1, rdom, er}
(NR, COM, M1)
{wcom, rfed, iip, pi, dl1, rdom, er}
(NR, COM, DL1)
{wcom, rfed, iip, pi, dm2, rdom, er}
(NR,COM,DM2)
11
12
13
14
15
VAR Models with Cholesky Decomposition [Recursive (R) Structure]
{oilp, rfed, iip, pi, dm3, rdom, er}
(R, OIL, DM3)
{oilp, rfed, iip, pi, m3, rdom, er}
(R, OIL, M3)
{oilp, rfed, iip, pi, m1, rdom, er}
(R, OIL, M1)
{oilp, rfed, iip, pi, dl1, rdom, er}
(R, OIL, DL1)
{oilp, rfed, iip, pi, dl1, rdom, er}
(R, OIL, DM2)
Model 16
{oilp, rfed, iip, pi, rdom, er}
(R,OIL, X)
A SVAR Approach to Evaluation of Monetary Policy in India
17
A SVAR Approach to Evaluation of Monetary
Policy in India
• Introduction and Motivation
• The Structural VAR model
• Empirical Result
• Conclusion
A SVAR Approach to Evaluation of Monetary Policy in India
18
Model and Puzzles
Table 2
Model & Code
Liquidity Puzzle
Price Puzzle
Exchange Rate
Puzzle
Forward Discount
Bias Puzzle
1 (NR,OIL,DM3)
none
None
None
None
2 (NR,OIL,M3)
Insignificant
None
None
None
3 (NR,OIL,M1)
Yes
Yes
None
None
4 (NR,OIL,DL1)
none
None
None
None
5 (NR,OIL,DM2)
none
None
None
None
6(NR,COM,DM3)
none
none
None
None
7 (NR,COM,M3)
Insignificant
Insignificant
None
None
8 (NR,COM,M1)
Insignificant
None
None
None
9 (NR,COM,DL1)
Insignificant
Insignificant
None
None
10(NR,COM,DM2)
Insignificant
None
None
None
11 (R,OIL,DM3)
Yes
Yes
None
Yes
12 (R,OIL,M3)
Insignificant
Yes
Yes
Yes
13 (R,OIL,M1)
None
Yes
Yes
Yes
14 (R,OIL,DL1)
Yes
Yes
None
Yes
15 (R,OIL,DM2)
Yes
Yes
None
Yes
16 (R,OIL,X)
Yes
Yes
Yes
Yes
A SVAR Approach to Evaluation of Monetary Policy in India

“Price puzzle” is a phenomenon where
a contractionary monetary policy shocks
identified with an increase in interest
rates, leads to a persistent rise in price
level instead of a reduction of it.

The liquidity puzzle is an empirical
finding when a money market shock is
associated with increases in the interest
rate instead of a decrease.

The exchange rate puzzle occurs when
a restrictive domestic monetary policy
leads to on impact depreciation of
domestic currency.

Or, if it appreciates, it does so for a
prolonged period of time violating the
uncovered interest parity condition
which is known as the forward discount
bias puzzle or delayed overshooting.
19
Impulse Response Functions in a Recursive model
A SVAR Approach to Evaluation of Monetary Policy in India

Depreciation of the currency and
persistent depreciation thereafter
(exchange rate puzzle and forward
discount bias puzzle) following
monetary policy shock

Persistent rise in inflation (price
puzzle) from a contractionary
monetary policy shocks.
20
Impulse Response Function in a Non-Recursive model
A SVAR Approach to Evaluation of Monetary Policy in India

We observe that SVAR models generally doing
way better than the recursive models

Models with Divisia monetary aggregates
(DM3) are doing better than simple-sum
monetary aggregate (M3)

Divisia aggregates (Divisa L1, Divisia M2 etc.)
did better than their simple-sum counterparts
not shown here
21
Impulse Response Function in a Non-Recursive model
A SVAR Approach to Evaluation of Monetary Policy in India
22
Variance Decomposition
Table 3

Forecast Error Decomposition: Contribution of Monetary Policy Shocks to Exchange Rate Variation (in
percentages)
Month
Model 1(DM3)
Model 2(M3)
Model 10(DM2)
Model 10 with world price of commodities
and Divisia M2 works the best

Between all the Divisia aggregates, DM2
consistently works better than DL1 and
DM3

Narrow simple-sum aggregate (M1)
consistently worked better compared to
M3

Divisia M3 consistently helps monetary
policy better explain fluctuation in the
exchange rate compared to its simple sum
counterpart
1
15.968
5.706
28.417
2
17.104
5.453
29.890
3
19.67
7.51
33.094
10
14.954
6.786
25.158
11
14.354
6.317
24.007
12
13.945
5.935
22.667
22
10.993
4.635
17.974
23
10.378
4.602
17.468
24
9.773
4.583
17.053
A SVAR Approach to Evaluation of Monetary Policy in India
23
Flip-flop analysis

Top panel represents the fluctuation in the
fundamental variables (exchange rate, inflation and
economic activity) that are being explained by the
policy variable.

Monetary policy is able to explains exchange rate
fluctuations the most, followed by prices and the
least for industrial production due to weak
transmission

Bottom panel represents how much of the same
fundamental variables are able to explain the
movement in the policy variable.

Central bank in India seems to set the monetary
policy rule with inflation-targeting in mind.
A SVAR Approach to Evaluation of Monetary Policy in India
24
Forecast statistics: Theil U and RMSE
Table 4
STEP
RMSE
(DM3)
RMSE
(M3)
Theil U
(DM3)
Theil U
(M3)
1
0.016817268
0.0168186
0.9407059
0.940740
2
0.027939798
0.0279426
0.9465474
0.946622
4
0.04509268
0.045096935
0.97101692
0.97110852
8
0.081620156
0.081625622
1.02052515
1.02059349
16
0.135008923
0.135019462
1.05461511
1.05469744
20
0.130837069
0.130854786
1.06003547
1.06017902
22
0.118076898
0.118096956
1.07450819
1.07469072
24
0.08290196
0.082923123
1.13793843
1.13822892

Model with Divisia M3 records a
lower RMSE and Theil U values, when
compared with the model with simplesum M3

The difference between the RMSE and
Theil U for the different aggregate
grows over time

This might be because Divisia M3
facilitates longer-horizon forecasting.
A SVAR Approach to Evaluation of Monetary Policy in India
25
Forecast statistics: Out-of-Sample forecast

Figure compares between the model without money
and model with the Divisia M3

The forecast with no money is represented in blue
diverges from actual value over time.

The forecast with Divisia M3 (represented in coral)
stays closer to the actual log of exchange rate value
(represented in black)

Forecast band for the Divisia M3 (represented in
pink) lies within the forecast band for the no money
(represented in green)
A SVAR Approach to Evaluation of Monetary Policy in India
26
Forecast statistics: Out-of-Sample forecast

Figure compares between the model with
simple-sum M3 and model with the
Divisia M3

Divisia M3 forecast (represented in coral)
stays closer to the actual LER value
(represented in black)

Simple-sum M3 forecast (represented in
blue) diverges from actual value over time.

Divisia M3 (represented in pink) has
narrower forecast band compared to
simple-sum M3 (represented in green)
reflecting higher forecast accuracy for
Divisia.
A SVAR Approach to Evaluation of Monetary Policy in India
27
A SVAR Approach to Evaluation of Monetary
Policy in India
• Introduction and Motivation
• The Structural VAR model
• Empirical Result
• Conclusion
A SVAR Approach to Evaluation of Monetary Policy in India
28
Conclusion
• K&R contemporaneous restriction customized for the Indian economy showed mostly puzzle-free results
• In order to claim this with confidence, we compared the contemporaneous SVAR with the recursive model
• SVAR model was able to get rid of the price puzzle and the exchange rate puzzle
• Existence of output-puzzle due to weakly developed financial market impeding monetary transmission
• The Variance Decomposition showed Divisia money rightly captures the information on flow of monetary
services
• Alternatively, money has an informational role to play in our model which gets clear when we compared
across no-money, model with simple-sum and model with Divisia aggregate
A SVAR Approach to Evaluation of Monetary Policy in India
29
Conclusion
• We did the out-of-sample forecasting, across simple-sum monetary models and Divisia money
models
• In general, the inclusion of money lowered the RMSE values and Divisia M3 money model did
fairly better than simple-sum M3 model
• Finally, the flip-flop analysis showed, monetary policy explains most of exchange rate fluctuation,
followed by inflation and least of the output movements
• Alternatively, inflation is able to explain the most of the policy –variable changes, followed by
exchange rate and output, switching over from time to time
• This leads us to believe that the Central Bank of India had inflation-targets in mind, when it set up
its policy rates.
A SVAR Approach to Evaluation of Monetary Policy in India
30