[PPT] - emissions control. An analysis of challenges to reach 2050 . CRREM PowerPoint Presentation

SLIDE 1

Real estate growth and carbon emissions control. An analysis of challenges to reach 2050.

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

1

CRREM PROJECT

Paloma Taltavull de La Paz Francisco Juárez Raul Pérez Sánchez Universidad deAlicante

Pacific Rim Real Estate Conference PRRES 2019 Melbourne,Australia

SLIDE 2

Agenda

2

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Introduction to CRREM.
Tool to estimate how much energy should be

saved to fulfill energy efficiency goals 2050

What this paper does and aim
Model estrategy and steps
Results
Conclusions

SLIDE 3

What is CRREM

3

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Carbon Risk Real Estate Monitor

– Project 785058 of H2020 EU program, Energy topic

Main goal: to estimate the required investment in the

existing comercial building in order to improve their energy efficiency and reduce carbon emissions

– Speed of energy renovation the building stock should follow – Identify the stranded assets ….Stranding risk

Build a tool to allow estimating to particular carbon

efforts:

– Real estate assets – Portfolios – Aggregate

SLIDE 4

The idea

4

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

SLIDE 5

The idea

Not so simple:

– At national level in EU (28!!) – At portfolio level – At building/asset type level

And..

–Climate becomes + hot

5

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

SLIDE 6

What is CRREM

6

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Consortium: 5 partners

– Coordinator: IIÖ (research centre), Austria – GRESB, The Netherlands – University of Tilburg – University of Ulster – University of Alicante

Strong links with companies (investor and

energy oriented firms)– EIC organization

http://crrem.eu

SLIDE 7

What is CRREM

7

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Calculation methodology follows several steps:

– Data base construction – Estimate the carbon impact of retrofitting in emissions and monetary investment – Fit how emissions evolve with the carbon target – Calculate the future increase on emissions

Forecast the future building trend

– All affect the emissions stream: horizon 2050 – All follow process of VERIFICATION of data and results

SLIDE 8

Step covered by this paper: Forecast building activity

Public forecasting are incompleted for our needs

Source: EUREF16, in wp2 report,

8

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

SLIDE 9

Aim of this presentation

9

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Show the forecasting strategy: 2018-2050!

– Time series arena: VAR environment methodology – Long term series are needed – Yearly data – Forecasting construction (m2) is needed…

Supply side model

– Estochastic modelling

We cannot advance any innovation nor structural change

– We can estimate the growth trend in the future done the past knowledge.

SLIDE 10

Conceptual model

10

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Di Pasquale & Wheaton (1996) shows that new supply

space reacts to changes on market prices and construction costs. Qrets = f(Pre,t, Cct ,St-1 , Gtk ) = e1 Pre,t

2 Ct 3 It 4 St-15

[Gtk ] 6t

– where: – Pre,t corresponds to real estate prices in real terms (market prices not developer prices) – Cct corresponds to the costs associated with construction materials and labor – it reflects the real interest rates paid by developers for building credits – St-1 is the existing stock at the previous moment – Gtk is a matrix of the regional market characteristics, including physical features as well as other aspects like land and market size ฀ t is a random term

฀ 1..6 are the estimatedparameters.

SLIDE 11

Problems?

11

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Long term data is not available enough.

– Time series comes from 1990 (quarterly) but prices from 2005!

Yearly base forecasting is better (less estimated

points than quarterly and with no seasonal effects) but requires long term evidence.

Proxies?
Forecasting method.

– Ideal: stochastic – Deterministic

SLIDE 12

Econometric Strategy

12

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

The analytical process follows the conventional sequential steps:

1.- stationary analysis,
2.- VAR definition and lag structure analysis,
3.- Cointegration tests identification,
4.- VECM definition, diagnosis and final model

estimation,

5.- Forecasting.
Separate country to country

SLIDE 13

Econometric Strategy

13

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Different methods for forecasting
1st. Looking for a proxies: housing prices?

– Prices prediction other prices?. Evidence – Exogenous prove, using GDP

2nd. Forecasting with proxies

– Supply model for commercial building permits

Offices
Commercial real estate (no-offices sector)

SLIDE 14

Econometric Strategy

14

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Two steps for forecast: 1st.- In-sample period [T-q+1,T], with q<T, in order to choose the model which minimize theprediction errors to fit available data.

Estimated with fixed estimated regressors (following Pesaran and Timmermann,2005).
These models have been using for prediction in real estate and construction by Jiang and Liu

(2011), Kouwenberg and Zwinkels (2014) or Bork and Moller(2015). 2nd Out-of-sample data until2050

‘the expanding window strategy’ (Pesaran and Timmermann, 2007) through which a [T+m] future

periods are estimated (with m>T) using a dynamic-stochastic simulation (Broyden 1969 solver) to calculate the future values in m-Tphases.

– Every model is repeated 1000 times allowing to a 5000 maximum iterations.

When the idiosyncratic features in the estimation process requires using more than one method of forecasting, utilizing the so-called as ‘the predictors technique’ (Clements andHendry, 1995)

i.e. by including a variable as a predictor which can be demonstrated to have strong propertiesto

approach the variable of interest. A combination forecast would be also applied to obtain consistent out of-sample predictions (Timmermann, 2006, Aiolfi et al, 2010 among othercontributions).

SLIDE 15

1st stage: housing prices model

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058. 15

SLIDE 16

1st. Housing price model
Dynamics seems to follow similar (lagged) cycle
Supporting the idea of co-movements and the

availability to be used as proxy (with the correct lag)

Figure 5.1.5-Prices: housing and offices Figure 5.1.6. Starts of housing and offices(m2)

400 800 1,200 1,600 2,000

P _ H pof_m2

Linearscaling Sources: M F O M andA N C E R T 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020

1

1 2 2,400 3

2

1975 1980 1985 1990 1995 2000 2005 2010 2015

OFIC_M2 S T A R T S

16

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058. Normaliseddata

SLIDE 17

1st. Housing price model

17

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Meen(2001):

Phdt = f(X, Z)t =  2 (Pop)t + 3 (y)t + 4 (K)t – 5 (Dh)t – 6 (Cu)t + t

Table 1. Variables in Model 1 of Housing Prices

Available

Variable

Concept period Source MFOM, Dallas Fed and Taltav Phdt Housing prices by m2 - Ph 1971-2018 and Juárez 2015 population older than 20 years. Taken in differences Pop Pob>20 1971-2018 INE Y GDP real terms -RGDP 1971-2018 INE Finance flow to buy houses (number of mortgages to K buy a house)- FF 1971-2018 INE, mortgage statistics h Changes in the stock measured by the flow of starts STARTS 1971-2018 MFOM User costs, measured by interest rates (real) and Cu inflation, RIRM, INF 1971-2018 Bank of Spain, INE

SLIDE 18

Model for housing prices

VECM modelling housing prices

– Cointegration test and error correction forequilibrium

Stochastic prediction of components for 2018-

2050

4 3 2 1

1
2
3

7 5 8 0 8 5 9 0 9 5 0 0 0 5 1 0 1 5 2 0

p o b >2 0 V I V I N RPIB RIRM P H FF_ N I N F L

18

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Norma l i s e d d a t a

SLIDE 19

Dynamic VECM

19

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

(8) J=1, Changes in housing prices

∆𝑄ℎ𝑢 = 𝛽1 +

5

𝜍1,𝑛 ∗ 𝑀𝑈𝑠𝑓𝑡𝑗𝑒𝑛 ,𝑢−1 +

3

𝜀1 ∆𝑄ℎ𝑢−𝑗 + 3

𝑛 =1 𝑗 =1 1,𝑗 𝑗 =1 2,𝑗

𝜀1 ∆2 𝑄𝑝𝑐 > 20𝑢−𝑗 +

3

𝜀1 ∆𝑆𝐻𝐸𝑄𝑢−𝑗 +

3 𝑗 =1 3,𝑗 𝑗 =1 4,𝑗

𝜀1 ∆𝐺𝐺𝑢−𝑗 +

3 𝑗 =1 5,𝑗 𝑗 =1 6,𝑗

𝜀1 ∆𝐽𝑂𝐺𝑢−𝑗 +

3

𝜀1 ∆𝑆𝐽𝑆𝑁𝑢−𝑗 +

3 𝑗 =1 7,𝑗

𝜀1 ∆𝑇𝑈𝐵𝑆𝑈𝑢−𝑗 + 𝜈1,𝑢

(9) J=2, Changes in population

𝑛 =1 𝑗 =1 1,𝑗 𝑗 =1 2,𝑗

∆2 𝑄 𝑃 𝐶 > 20𝑢 = 𝛽2 +

5

𝜍2,𝑛 ∗ 𝑀𝑈𝑠𝑓𝑡𝑗𝑒𝑛 ,𝑢−1 +

3

𝜀2 ∆𝑄ℎ𝑢 −𝑗 +

3

𝜀2 ∆2 𝑄𝑝𝑐 > 20𝑢−𝑗 +

3 𝑗 =1 3,𝑗 𝑗 =1 4,𝑗 𝑗 =1 5,𝑗 𝑗 =1 6,𝑗

𝜀2 ∆𝑆𝐻𝐸𝑄 𝑢 −𝑗 +

3

𝜀2 ∆𝐺𝐺𝑢−𝑗 +

3

𝜀2 ∆𝐽𝑂𝐺𝑢 −𝑗 +

3

𝜀2 ∆𝑆𝐽𝑆𝑁 𝑢 −𝑗 +

3 𝑗 =1 7,𝑗

𝜀2 ∆𝑇𝑈𝐵𝑆𝑈𝑢 −𝑗+ 𝜈2,𝑢

(10) J=3, changes in economic growth

𝑛 =1 𝑗 =1 1,𝑗 𝑗 =1 2,𝑗

∆𝑆𝐻𝐸𝑄 𝑢 = 𝛽3 +

5

𝜍3,𝑛 ∗ 𝑀𝑈𝑠𝑓𝑡𝑗𝑒𝑛 ,𝑢−1 +

3

𝜀3 ∆𝑄ℎ𝑢 −𝑗 +

3

𝜀3 ∆2 𝑄𝑝𝑐 > 20𝑢−𝑗 +

3 𝑗 =1 3,𝑗 𝑗 =1 4,𝑗 𝑗 =1 5,𝑗 𝑗 =1 6,𝑗

𝜀3 ∆𝑆𝐻𝐸𝑄 𝑢 −𝑗 +

3

𝜀3 ∆𝐺𝐺𝑢−𝑗 +

3

𝜀3 ∆𝐽𝑂𝐺𝑢 −𝑗 +

3

𝜀3 ∆𝑆𝐽𝑆𝑁 𝑢 −𝑗 +

3 𝑗 =1 7,𝑗

𝜀3 ∆𝑇𝑈𝐵𝑆𝑈𝑢 −𝑗+ 𝜈3,𝑢

(11) J=4, changes in financial flows to housing

𝑛 =1 𝑗 =1 1,𝑗 𝑗 =1 2,𝑗

∆𝑔𝑔𝑢 = 𝛽4 +

5

𝜍4,𝑛 ∗ 𝑀𝑈𝑠𝑓𝑡𝑗𝑒𝑛 ,𝑢−1 +

3

𝜀4 ∆𝑄ℎ𝑢 −𝑗 +

3

𝜀4 ∆2 𝑄𝑝𝑐 > 20𝑢−𝑗 +

3 𝑗 =1 3,𝑗 𝑗 =1 4,𝑗 𝑗 =1 5,𝑗 𝑗 =1 6,𝑗

𝜀4 ∆𝑆𝐻𝐸𝑄 𝑢 −𝑗 +

3

𝜀4 ∆𝐺𝐺𝑢−𝑗 +

3

𝜀4 ∆𝐽𝑂𝐺𝑢 −𝑗 +

3

𝜀4 ∆𝑆𝐽𝑆𝑁 𝑢 −𝑗 +

3 𝑗 =1 7,𝑗

𝜀4 ∆𝑇𝑈𝐵𝑆𝑈𝑢 −𝑗+ 𝜈4,𝑢

SLIDE 20

Dynamic VECM

20

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

(12) J=5, Changes on inflation

𝑛 =1 𝑗 =1 1,𝑗 𝑗 =1 2,𝑗

∆𝑗𝑜𝑔

𝑢 = 𝛽5 + 5

𝜍5,𝑛 ∗ 𝑀𝑈𝑠𝑓𝑡𝑗𝑒𝑛,𝑢−1 + 3 𝜀5 ∆𝑄ℎ𝑢−

𝑗 + 3

𝜀5 ∆2𝑄𝑝𝑐 > 20𝑢−𝑗+

3 𝑗 =1 3,𝑗 𝑗 =1 4,𝑗 𝑗 =1 5,𝑗 𝑗 =1 6,𝑗

𝜀5 ∆𝑆𝐻𝐸𝑄𝑢 −

𝑗 + 3

𝜀5 ∆𝐺𝐺𝑢−𝑗 + 3 𝜀5 ∆𝐽𝑂𝐺𝑢−

𝑗 + 3

𝜀5 ∆𝑆𝐽𝑆𝑁𝑢 −

𝑗+ 3 𝑗 =1 7,𝑗

𝜀5 ∆𝑇𝑈𝐵𝑆𝑈𝑢−

𝑗+𝜈5,𝑢

(13) J=6, Changes on real interest rates

𝑛 =1 𝑗 =1 1,𝑗 𝑗 =1 2,𝑗

∆𝑆𝑗𝑠

𝑢 = 𝛽6 + 5

𝜍6,𝑛 ∗ 𝑀𝑈𝑠𝑓𝑡𝑗𝑒𝑛,𝑢−1 + 3 𝜀6 ∆𝑄ℎ𝑢−

𝑗 + 3

𝜀3 ∆6𝑄𝑝𝑐 > 20𝑢−𝑗+

3 𝑗 =1 3,𝑗 𝑗 =1 4,𝑗 𝑗 =1 5,𝑗 𝑗 =1 6,𝑗

𝜀6 ∆𝑆𝐻𝐸𝑄𝑢 −

𝑗 + 3

𝜀6 ∆𝐺𝐺𝑢−𝑗 + 3 𝜀6 ∆𝐽𝑂𝐺𝑢−

𝑗 + 3

𝜀6 ∆𝑆𝐽𝑆𝑁𝑢 −

𝑗+ 3 𝑗 =1 7,𝑗

𝜀6 ∆𝑇𝑈𝐵𝑆𝑈𝑢−

𝑗+𝜈6,𝑢

(14) J=7, Changes on housing starts

𝑛 =1 𝑗 =1 1,𝑗 𝑗 =1 2,𝑗

∆𝑡𝑢𝑏𝑠𝑢

𝑢 = 𝛽7 + 5

𝜍7,𝑛 ∗ 𝑀𝑈𝑠𝑓𝑡𝑗𝑒𝑛,𝑢−1 + 3 𝜀7 ∆𝑄ℎ𝑢−

𝑗 + 3

𝜀7 ∆2𝑄𝑝𝑐 > 20𝑢−𝑗+

3 𝑗 =1 3,𝑗 𝑗 =1 4,𝑗 𝑗 =1 5,𝑗 𝑗 =1 6,𝑗

𝜀7 ∆𝑆𝐻𝐸𝑄𝑢 −

𝑗 + 3

𝜀7 ∆𝐺𝐺𝑢−𝑗 + 3 𝜀7 ∆𝐽𝑂𝐺𝑢−

𝑗 + 3

𝜀7 ∆𝑆𝐽𝑆𝑁𝑢 −

𝑗+ 3 𝑗 =1 7,𝑗

𝜀7 ∆𝑇𝑈𝐵𝑆𝑈𝑢−

𝑗+𝜈7,𝑢

SLIDE 21

In-sample forecast

Quite accurate in house price and starts
predictions. Also in GDP!!

Figure 5.1.8. In-Sample predictions of the model. Accuracy and confidencebands Panel 1- housing prices Panel 2.-GDP

5 1 ,0 1 ,5 2 ,0 2 ,5 8 9 1 2 3 4 5

Actu a l P H (BaselineM e a n )

P H

2 , 4 , 6 , 8 , 1 , , 1 , 2 , 1 , 4 , 8 9 1 2 3 4 5

Actual R P I B (BaselineM e a n )

21

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

R P IB

SLIDE 22

In-sample forecast

800,000 600,000 400,000 200,000 80 90 00 10 20 30 40 50

Actual VIVIN (Baseline Mean)

22

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

VIVIN

SLIDE 23

Prediction for 2018-2050

1 ,0 2 ,0 3 ,0 4 ,0 5 ,0 8 9 1 2 3 4 5

Actual P H (BaselineMean)

Figure 5.1.9. Out of-Sample predictions of the model with the horizon in 2050. Accuracy and confidence bands Panel 1-housing prices Panel 2.-GDP

P H

500,000 1,000,000 1,500,000 2,000,000 2,500,000 80 90 00 10 20 30 40 50

Actual R P I B (BaselineMean) 23

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

R P IB

SLIDE 24

Prediction for 2018-2050

Less accurate in the case of starts

24

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

SLIDE 25

2nd stage: commercial space new supply (permissions)

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058. 25

SLIDE 26

2nd stage: office space

Figre 5.1.4. Office building space and office transaction prices.Spain

3 2 1

1
2

1990 1995 2000 2005 2010 2015 2020

OFIC_M2 p o f _ m 2

26

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

SLIDE 27

2nd step. Office space prediction

27

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Lack on data: 2 estimations:
A)Supply-focused model

∆𝑅𝑝𝑔𝑛2𝑢 = 𝛽 + 2 𝜍𝑨 𝛾1𝑅𝑝𝑔𝑛2𝑢−1 + 𝛾2𝑞_ℎ𝑢−1 + 𝛾3𝑑𝑑𝑝𝑜𝑡𝑢𝑢−1 + 𝛾4𝑠𝑗𝑠𝑛𝑢−1+

𝑨=1

SLIDE 28

2nd step: Office space prediction

Figures 5.1.12. Semilog supply model forecast ofOffice_m2 Figures 5.1.14. Demand model. Deterministic forecast of Office_m2 using thepredictor

3,000 2,000 1,000 80 90 00 10 20 30 40 50

Actual OFIC_M2 (Scenario 1)

28

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

OFIC_M2

SLIDE 29

25,000 20,000 15,000 10,000 5,000 1980 1990 2000 2010 2020 2030 2040 2050

renoff renoff (Baseline Mean)

2nd step: commercial space

renoff

renoff renoff (Scenario 1)

29

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Figures 5.3.5. Demand model. Deterministic forecast of Commercial real estate_m2 using thepredictor renoff

24,000 20,000 16,000 12,000 8,000 4,000 1980 1990 2000 2010 2020 2030 2040 2050

SLIDE 30

Conclusions

30

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058.

Strong needs to forecast basic variables in order

to take decisions

– Climate change: reduce energy consumption by increasing efficiency but…. How much??

Not technical variables at all,
Socioeconomic relationships explain main

variables to be forecast

Needs to use econometric techniques based on

the theory and evidence in real estate markets

It works! (at the moment)

– High uncertainty but we do not have the cristal ball

SLIDE 31

Thanks for your attention

This project has received funding from the European Union’s Horizon 2020 research andinnovation programme under grant agreement no. 785058. 31