Empirical Stock Flow Consistent Modeling

Stephen Kinsella | University of Limerick
AB-SFC Winter School, 28/01/2015

Plan.

Yesterday with Dr. Caverzasi you saw:

  • History, development and logical structure of stock flow consistent (SFC) models;
  • Examples of a theoretical SFC model;
  • Your first matrices (by hand).

Today with me you'll see:

  • The system of national accounts, 'hands on' (ie with a computer);
  • How to actually get the data you need;
  • Construction of a real world Transaction Flow Matrix, Balance sheet;
  • The structure of an empirical SFC & Estimation techniques.

A few important ideas from yesterday

  1. Account structures (BS/TFM/FoF)
  2. Identities \( (S-I) = (G-T)+ (X-M) \)
  3. Adding up constraints (i.e. \( F=C+\Delta K-W-rL_{-1} \) from Eugenio's example)
  4. Behavioural equations \( Y = \alpha_{1}(Y-T)+\alpha_{2}(V_{1}) \).

Why do this?

  1. Because it's new.
  2. Because other modeling approaches haven't done so well post-crisis.
  3. Revolution in macro-data availability.
  4. First real exposure of non-standard theory to major macro data.

Key Papers

(links in blue take you directly to the .pdf for your reference)

Key Empirical SFC Papers

(links in blue take you directly to the .pdf of the paper for your reference)

SFC vs DSGE (1)

Some merits of using SFC models Some merits of using DSGE models
Typically use national accounting constraints to provide a framework Model equations are explicitly linked to the optimisation problems of particular agents.
Allow modelling of gross flow and balance sheet positions by sector Offer a well-established framework which can be extended – for example, financial frictions can be added into workhorse DSGE models.
Can be used to model feedback from financial asset and liability positions to the paths for production and spending The models are usually simple enough that the main economic mechanisms at work can be explained easily.
Can include an important role for money, credit and the financial system Easy to take to the data – for example, they can be linearised and mapped into VAR representations of the endogenous variables.

SFC vs DSGE (2)

Some merits of using SFC models Some merits of using DSGE models
Can offer a framework for exploring different specifications for agents’ expectations Since the model parameters are ‘structural’, they should not be affected by changes in policy regime or time series properties of the driving processes.
Arguably SFC models have more realistic behavioural assumptions than many models which are microfounded. Models are clearly linked to economic theory.

Examples & Review of state(s) of the art(s)

Structure of the data you'll need.

We need to understand the structure of the European System of Accounts in order to understand the data we will be using in a while.

ESA is a codified method to measure the structure of a total economy. Used throughout OECD & internationally compatible.

To see how it's used, see Be Duc & Le Bretton, 2012, see also Lequellier (2006) on National Account principles.

Important concepts/delineations: statistical units and their groupings/flows and stock/the system of accounts and the aggregates/the input-output framework.

ESA: Stocks and Flows

1.65: Flows refer to actions and effects of events that take place within a given period of time, while stocks refer to positions at a point of time.

1.69 Flows reflect the creation, transformation, exchange, transfer or extinction of economic value.

1.85 Stocks are the holdings of assets and liabilities at a point in time. Stocks are recorded at the beginning and end of each accounting period. The accounts that show stocks are called balance sheets.

ESA: Sequencing of accounts.

Very important to understand this. See ESA, pg. 20.

ESA, 2010, p.20

ESA, 2010, p.20

You're going to start thinking in terms of B9 and S112. Importantly, The balancing item of financial assets and liabilities is called financial net worth (BF.90)

Let's also see page 189 for a discussion on how to move from opening time t of a balance sheet to t+1.

Actually getting the data

We'll stay fairly Anglo-Saxon in this lecture, simply because this is where the data are the best.

-ECB Flow of Funds

-US Flow of Funds

-UK Flow of Funds

-ECB Bulk download facility

-ECB Bulk Download (guide)

  • See R Package eurostatr for help getting .tsv data into R.

When you get the data in the .Rdata format, you can do pretty cool things like making sectoral balance charts easily ——->

plot of chunk unnamed-chunk-1

Checking the data

These data are:

  1. Noisy
  2. Really noisy
  3. Messy
  4. Not Stock flow consistent.
  5. So we'll have to change that.

Here's another cool chart. Pretty much just because I can make them.

plot of chunk unnamed-chunk-2

Checking the data: An example from our work for the Icelandic model

We know the Current account and the financial accounts are intimately related. But in practice these data are not gathered by the same people. Get data for both B9 and BF9 and see if you can match them.

Icelandic example.

Icelandic example.

So you know you need to check the crisis-years, and the accounts therein, to make sure these two are consistent with one another.

Now: Staring at the data

Correlation matrices plotted as 'heatmaps' help you 'see' the entirety of the data set. Looking at it by sector.

First period: 2002-2007.

Second Period, 2008-2014

So now you're sure the data are OK. Let's make a TFM & look at the behavioural eqs

R Script Example

So we made a Balance Sheet yesterday.

x HH NFC SPV Banks Financial Firms Gov CB ROW Sum
Deposits +Dh +Dn -D 0
Equities +p.E -p.E +p.eb +p.Eff +p.Er 0
Bonds +Bh +Bb +Bff -B +Bcb +Br 0
Shares +ShF -ShF 0
Loans -Lh -Lnf +Ls +L -Lff 0
Real K +K K
Advances -A +A 0
Reserves +R -R 0
Securities -SE +SE 0
Net Wealth Vh Vf Vspv Vb(+K) Vf Vg Vcb Vrow

And we made a TFM.

x HH NFC SPV Banks FF Gov CB ROW
Consumption -C +C
Wages +W -W
Taxes -Th -TF -Tff +T
Investment +\( \Delta \) K -\( \Delta \) K
Gov. Exp +G -G
Trade Balance -Tr +Tr
Int. D +i.Dh +i.Df -i.D
Int. Se -i.SE +i.SE
Int L -i.Lh -i.Lf +i.Ls +i.Lb
Int B +iBh +i.Bb -i.Bff +i.Bcb +i.Br
Dividends +Fh -F +Fb +Ff +Fr
Shares +is.sh -is.sh
Total SAVh SAVf SAVspv Savb SAVf SAVg SAVcb SAVcb

And we made a revaluation matrix

x HH NFC SPV BANK FF GOV CB ROW Total
Total SAVh SAVf FSP SAVb SAVf SAVg
\( \Delta \) Deposits -\( \Delta \) Dh + \( \Delta \) Db 0
\( \Delta \) Equities -p. \( \Delta \) E +p. \( \Delta \) Ef -p. \( \Delta \) Ef -p. \( \Delta \) Er 0
\( \Delta \) Bonds -\( \Delta \) Bh -\( \Delta \) Bb -\( \Delta \) Bf +\( \Delta \) Bg -\( \Delta \) Bcb -\( \Delta \) Br 0
\( \Delta \) Share -\( \Delta \) Sh +\( \Delta \) Sh 0
\( \Delta \) Loans +\( \Delta \) Lh -\( \Delta \) Lspv -\( \Delta \) Lb 0
\( \Delta \) Real K - \( \Delta \) K 0
\( \Delta \) Advances +\( \Delta \) A -\( \Delta \) A 0
\( \Delta \) Reserves -\( \Delta \) R +\( \Delta \) R 0
\( \Delta \) Securities +\( \Delta \) Se -\( \Delta \) Sef 0

Let's break into groups and try to get the date for the US

Irish TFM

X incomes_hh expenditures_hh incomes_fc expenditures_fc incomes_nfc expenditures_nfc incomes_gg expenditures_gg incomes_row expenditures_row
F2: Deposits +DEP_FLOW_h +DEP_FLOW_f -DEP_FLOW_f +DEP_FLOW_n 0 +DEP_FLOW_g -DEP_FLOW_g +DEP_FLOW_r -DEP_FLOW_r
F3: Bonds +B_FLOW_h +B_FLOW_f -B_FLOW_f +B_FLOW_n -B_FLOW_n +B_FLOW_g -B_FLOW_g +B_FLOW_r -B_FLOW_r
F4: Loans -L_FLOW_h +L_FLOW_f -L_FLOW_f +L_FLOW_n -L_FLOW_n +L_FLOW_g -L_FLOW_g +L_FLOW_r -L_FLOW_r
F5: Equities -EQ_FLOW_f -EQ_FLOW_n +EQ_FLOW_r
F6: ITRs +ITR_FLOW_h -ITR_FLOW_f +ITR_FLOW_r
F7: Other Accounts P/R +OTH_FLOW_h -OTH_FLOW_h +OTH_FLOW_f -OTH_FLOW_f +OTH_FLOW_n -OTH_FLOW_n +OTH_FLOW_g -OTH_FLOW_g +OTH_FLOW_r -OTH_FLOW_r
B9_F: Net financial transactions (balancing liabilities) NTR_h NTR_f NTR_n NTR_g NTR_r
Net Output +NOUT (+NOUT_n) -NOUT_n
Gov Profits -GOVP_n +GOVP
Wage Bill +WB -WB_f -WB_n -WB_g
Net Taxes (+T_h) -T_h -T_f (+T_n) -T_n +T (-T_g) (+T_r) -T_r
Property Income:Interest (+INT_h) -INT_h +INT_f (-INT_f) (+INT_n) -INT_n (+INT_g) -INT_g (+INT_r) -INT_r
Incomes: Pensions & other +PENS_h -PENS +PENS_r
Dividends -DIVD_f -DIVD_n +DIVD
Financial Consumption (+FINCONS_h) -FINCONS_h +FINCONS_f (-FINCONS_f) (+FINCONS_n) -FINCONS_n (+FINCONS_g) -FINCONS_g +FINCONS_r (-FINCONS_r)
Consumption -CONS_h +CONS -CONS_g
Capital Gains +CAPG -CAPG_f
Capital Transfers +CAPT_f (-CAPT_f) +CAPT_f (-CAPT_f) +CAPT_n (-CAPT_n) (+CAPT_g) -CAPT_g +CAPT_r (-CAPT_r)
Gross capital formation -GFCF_h -GFCF_f +GFCF (-GFCF_n) -GFCF_g
Net Exports +NX_n (-NX_n) (+NX_r) -NX_r
social contributions (+SOCCON_h) -SOCCON_h -SOCCON_f +SOCCON
social benefits +SOCBEN_h (-SOCBEN_h) -SOCBEN_f -SOCBEN +SOCBEN_r (-SOCBEN_r)
Economic Disappearance +ECONDIS_h (-ECONDIS_h) +ECONDIS_f (-ECONDIS_f) (+ECONDIS_n) -ECONDIS_n +ECONDIS_g (-ECONDIS_g) +ECONDIS_r (-ECONDIS_r)
Net Lending/Borrowing in non-financial account NL_h NL_f NL_n NL_g NL_r

Summary