Home : Projects : Kenya : KTMM
Start
Micromacro Consultants
Photos
Projects
Downloads
Contact
Nederlands
Micromacro Consultants
Fotos
ProjectenNL
CuracaoAruba
DownloadsNL
Route

Treasury Macro Model Kenya (KTMM)

This part is based on the Paper to Seminar 7, November 2002, in The Hague.
This paper is  very short,  full documentation is in the model file itself: See for the theory behind the behavioural equations the sheet THEORY in the model file: KTMM.xls. That model file also contains a sheet MANUAL.

 
Table of contents:

Click here to download the word-file of this document.

 

Visit 3-10 November 2002 of two KIPPRA economists to MMC in The Hague

Background:

In January 2000 did start the EU Project nr. 7 ACP KE 080, Institutional Support to the Kenya Institute for Public Policy Research and Analysis (KIPPRA) by ISS/MMC. In this project Micromacro Consultants (MMC) has supported the construction of the macro model, and -once the operational version was running- the use of the model for Budget forecasting, Policy Analyses and Research. In this project MMC used its Macroabc methodology (that has been applied also successfully in Curaçao, Suriname, Poland, EU15, The Netherlands and Indonesia, see www.micromacroconsultants.com) and telework approach: MMC bring short visits to Kenya, Kenyan economists follow short training visits at MMC in the Hague, and in between co-operation by telework.

Target:

The target of this additional training visit of KIPPRA economists to The Hague is:
  • Increase the experience of KIPPRA economists further in using the macro model for policy and historical variants analysis. 
  • Provide KIPPRA economists with ideas for further improvement of the KTMM. 
  • Model improvements
Participants: Two economists of KIPPRA Kenya:
  1. Dr. Stephen Karing
  2. Ms. Maureen Were MA

MMC consultants who have been  involved in this project:
 (* involved in this visit, ** did also visit KIPPRA in Nairobi):
  1. *Dr. Marein van Schaaijk**           (director MMC, Macro model manager) 
  2. *Dr. Free  Huizinga**                     (specialist macro model theory) 
  3. Drs. Laurens Harteveld                (economist, assistant to project director, MMC) 
  4. *Drs. Bas van Tuijl                         (economist, assistant to project director, MMC) 
  5. Drs. Runy Calmera                       (Curalyse, macro model Curaçao) 
  6. *Dr. Jan Donders                           (head  short term division CPB) 
  7. Drs. Bjorn van Hamel                    (retired CPB, MMC consultant Indonesia) 
  8. Drs. Cees Jansen                          (expert in fiscal and macro monitoring) 
  9. Drs. Ate Nieuwenhuis                    (specialist in prices analysis) 
  10. Dr. Herman Stolwijk                       (model builder, agricultural expert) 
  11. Dr. Andrzej Tabeau                         ( MMC consultant Polish Macroabc) 
  12. *Dr. Johan Verbruggen                   (model & short term, head Short term Div. CPB) 
  13. Drs. Michiel Vergeer                        (Fiscal expert, MMC consultant Indonesia) 
  14. Drs. Eugene Verkade                      (macro models Indonesia and India) 
  15. Dr. Dirk van der Werf                        (macro model builder, revenues government) 
  16. Drs. Gerard van Welzenis               (international monitoring expert CPB) 

Organization of the training visit at MMC’s office:

The training is organized by workshops in the morning and model training and improving sessions in the afternoon. These afternoon sessions most of the time will be done with the whole group with a PC with big screen, and some training will be done individually and each participant will have a PC at his disposal.

Time table Workshops 4-8 November at MMC's office in The Hague:

9.30-13.30 discussion meeting (workshop) 9.00-9.30 participants prepare workshop,

13.30-17.45: participants prepare workshops and improve their model

Day

 

Subject

Expert

S 3 Nov.

Arrival of the participants

 

M 4 Nov.

Policy & Historical simulations

Dr. Marein van Schaaijk

Afternoon:

CPB analyses: even more than macro: micro sectoral, etc.

Dr. Free Huizinga

T 5 Nov.

Short term forecasting at CPB

Dr. Johan Verbruggen

Afternoon

Poverty Module to Curacao Macroabc

Dr. Marein van Schaaijk and Drs. Bas van Tuijl

W 6 Nov.

Monitoring monthly CPI, exchange rate, interest rate

Drs. Bas van Tuijl

Afternoon:

cost price (including indicator self-employed): re-estimation, variants, model improvement

Dr. Marein van Schaaijk

T  7 Nov.

Macro Theory

Dr. Free Huizinga

Afternoon:

16.00-17.30 Seminar at ISS

 

Evening:

Fare well dinner

 

F  8 Nov.

Macro & Fiscal models: how to cooperate?

Dr. Jan Donders

Afternoon

Historical simulations; research agenda

Dr. Marein van Schaaijk

S  9 Nov

End of this visit

 


KTMM seminar at ISS Thursday 7 November

SPECIAL ECONOMIC RESEARCH SEMINAR
7 November 2002 16.00-17.30

Kenyan Model (KIPPRA) model seminar:

Title: ''Presentation of the KIPPRA-Treasury Macro Model Kenya: 
discussion of theory behind, demonstration about how it runs and base
line and policy simulations".

Chairman: Rob Vos (ISS)
Discussant: Jan van Heemst (ISS)

The model, labeled KTMM, has been developed as part of the
KIPPRA/ISS/MMC project in 2000 and already has been used
intensively for the preparation of Budget Papers, the Medium Term
Economic Framework, a Poverty Reduction Strategy Paper and the
Development Plan.
  1. Presentation of KTMM theory by Dr. Stephen Karingi (KIPPRA) 
  2. Presentation of KTMM consistency framework by Dr. Marein van Schaaijk (MMC) 
  3. Presentations of Variants Paper by Ms. Maureen Were  MA  (KIPPRA)
Papers to be discussed:
  1. ''A Better Understanding of the Kenyan Economy: Simulations from the  KIPPRA-Treasury Macro Model" by Maureen Were and Stephen Karingi,   Discussion Paper KIPPRA, August 2002 
  2. "Introduction to KTMM". Paper can be found on www.micromacroconsultants.com

 

Documentation to KTMM

  1. A review of old models: Macro Models of the Kenyan Economy: A Review, by Karingi S. Njugana and Njugana S. Ndung'u, January 2000, KIPPRA D Paper no.2.
  2. A general theoretical underpinning of the behavioural equations, based on the theory behind the Polish Macroabc and discussions at KIPPRA. Improved version below. 
  3. Report on Macro data needs for the KIPPRA Macro Model and a work plan how to realize an operational version of the KIPPRA Macro Model in August 2000: Macro Model Work plan (Annex B and C to Inception Report) 
  4. Draft Consistency Framework for the KIPPRA Kenya Macro Model and the provisional macro database, KTMM note 5, February, improved version note 6 March 2000. With attached disk with model file. 
  5. The paper: ''Draft Estimation Procedure and Estimated Results of the KIPPRA-Treasury Macro Model'', April 2000. The paper describes each of the 12 behavioural equations: specifications and coefficients based on the initial theory paper and estimation/calibration results for 1972-1999. 
  6. The paper KIPPRA-Treasury Macro Model Kenya, Database and Consistency Framework, based on MMC's Macroabc methodology, April 2000, including documentation in sheet Manual and CDROM with data and model files and multi-media self explaining slideshow to the model.
  7. The Paper KTMM and Proceedings Workshops June 2000
  8. The Paper KTMM Workshop, Nairobi August 16-17, 2000 with model overview. 
  9. Paper to seminar at ISS  6 July 2001: ‘’The Kenya model KTMM’’ 
  10. The Paper ‘’Kenya’s exchange rate movement in a Liberalized Environment, An empirical Analysis’’  KIPPRA Discussion Paper DP/10/2001 
  11. The Paper ‘’Theoretical Base for the Kenya Macro Model: The KIPPRA-Treasury Macro Model, KIPPRA Discussion Paper 11, October 2001.
  12. The Paper KTMM Study Module +CDROM, March 2002.
  13. The Paper Kenya Variants Analysis Training MMC, June 2002.
  14. The Paper: ‘’A Better Understanding of the Kenyan Economy: Simulations from the KIPPRA-Treasury Macro Model” by Maureen Were and Stephen Karingi, Discussion Paper KIPPRA, No.. , August 2002.
  15. This Paper  ‘’Introduction to KTMM’’ seminar at ISS, 7 November 2002

The theory and estimation papers are brought into sheet Manual and Theory in the Excel version of the model, with hyperlinks between the equations in the model sheet and manual sheet. The KTMM model file has not been published by KIPPRA so far, but another Macroabc model, Curalyse, can be downloaded as shareware from www.micromacroconsultants.com

Overview KTMM

The KIPPRA-Treasury Macroeconomic Model (KTMM) model is an aggregate supply/aggregate demand type of macro model of a market economy, and it describes the behaviour of market actors. It consists of equations not only from demand side but also from supply side. The model has been made for the Kenyan Institute for Public Policy Research and Analysis (KIPPRA) to analyze the effects of the economy on the Budget of the Government as well as the other way around: the effects of the Budget on the economy. It is a simultaneous model. The theoretical base of the KTMM is described fully in Theoretical Base for the Kenya Macro Model, KIPPRA Paper 11, October 2001). The model is demand driven in the short run, with multiplier effects through consumption and investment. An important assumption of the model is that any demand is actually met, that is, it is assumed that the price system ensures that there is always some excess capacity in the economy. High demand leads to high capacity utilisation rates of capital and low unemployment rates, however, which lead to wage and price increases. With the important further assumption that the resulting inflation will decrease the competitiveness of exports, causing a reduction in exports and then lowering investments. In this way the model has a tendency to return to equilibrium with ‘normal’ capacity utilisation and unemployment rates in the medium and long run.

These main feedback mechanisms in the real economy work through the wage-price spiral and also the real exchange rate. For instance, an increase in aggregate demand raises labour demand, reduces the unemployment rate, raises wages and starts a wage – price spiral. The resulting inflation causes a real appreciation, a reduction in competitiveness and a reduction in exports. The drop in exports and investment reduces demand again, until equilibrium is restored. An important point to note is that even though total demand may be stabilised in this way, the feedback mechanism may well change the composition of demand. For instance, if the original increase in demand came from an increase in government spending, the net result will be a shift from exports and investment to government spending, resulting in a government deficit and a current account deficit.

The model is well organized in blocks and sheets. In this overview we will introduce the main relationships.

We discuss the main relations in MODFI following the numbers in the diagram and box:
  • The Flow Diagram of the Structure of KTMM 
  • Box with list main behavioural equations in KTMM
The components of GDP: see the arrows in the diagram going to GDP, coming from private consumption, investments, government consumption and investments, exports minus imports:

The real consumption growth follows real disposable income growth (with a higher saving rate for the profit disposable income then that from wages disposable income).

The private investment level is explained by a combination of flexible accelerator and net profitability (profits minus calculated interests) plus the change in public investments.

Real export growth: 80% of world trade growth + 2 * (world market price minus export price). So one percent better competitiveness gives two percent more real exports growth.

The real growth of imports equals: 1,3* the final demand growth, re-weighted with import intensities, minus 60% of difference of growth of import prices and domestic prices. So a difference of one percent in imports and domestic prices gives a half percent  lower real imports.

The Prices are mainly explained from cost price developments: the wage costs (wage rate minus trend in labour productivity), import price and change in indirect taxes pressure. The coefficients for wage costs and import price differ in the price equations for consumption, investments and exports. The export price follows for 70 %  cost price (wage costs plus import price) and 30% competitors price , plus some effect of real interest rate. Also the investments price is affected by the real interest rate, and by the utilisation rate.

The wage rate follows inflation, but with a lag and not for 100%. Furthermore the wage rate is influenced by the trend in labour productivity, and the pressure of direct taxes, and the growth of employment.

Employment of businesses follows real GDP growth and is negatively affected by real wage growth.

In the monetary side: the interest rate changes with 105 of the difference between actual and targeted inflation. Because of this low coefficient it is almost exogenous for the time being. The exchange rate equation is based on a combination of purchasing power parity, interest rate parity and exogenous shocks in the Balance of Payments.  The money supply equals real GDP growth plus the inflation target. This means that Central Bank will sterilise all money created from abroad or by government as far as it is more than the growth of real GDP plus the targeted inflation. Please note that the interest equation shows an increase if inflation is higher than the targeted level. 

Next diagram shows these relationships and the following box presents the overview of formulas in behavioural equations (lags not mentioned).  So far the equations as they are in the actual KTMM. How did we reach those results?

Which behavioural equations, what level of coefficients?

The behavioural equations are identified in the initial theory paper and estimated in the paper: ''Draft Estimation Procedure and Estimated Results of the KIPPRA-Treasury Macro Model'', April 2000.

For each of the equations (see the list of 12 main behavioural equations) we needed a specification and levels of the coefficients. This is done by an iterative process based on:
  1. The economic theory 
  2. The estimation results 
  3. Other studies (about Kenya and other countries and other Macroabc models) 
  4. The simulation properties

The initial theory paper (based on other Macroabc models and discussions in Nairobi) provided the base for the estimations of the behavioural equations. However before starting the estimations, we of course needed to construct a database and consistency framework. A preliminary database and consistency framework were made in the beginning of the KTMM project, and then the estimation process started.

Estimations of the behavioural equations

The twelve main behavioural equations have been estimated for the period 1972-1998. See the paper: ''Draft Estimation Procedure and Estimated Results of the KIPPRA-Treasury Macro Model'', April 2000. The (improved) text of this paper is in the sheet THEORY.

In these estimates only explanatory variables from sheet MODEL are used (otherwise we would not have values for the explanatory variables for future years to make forecasts). During the estimation process we needed some additional primary variables. They were added to the data system. We did also add additional secondary variables to sheet MODEL, because they can by definition be derived from the primary variables in sheet MODEL. Only in a very special case we brought additional primary variables into the system, to prevent that the model becomes too big.

For every behavioural equation a research note has been made (with description of theoretical background, estimation results and proposed specification and level of coefficients in the model).

The results of the estimations brought into the sheet MODEL

When the behavioural equations were estimated, as expected, always some estimates are inconclusive. For example because the time series after the regime switches in the beginning of the nineties are to short, or because we don't have time series for some variables, or because the confrontation with actual figures shows that the postulated theory does not fit Kenyan reality. In that case we had to discuss the specifications and levels of coefficients for example on base of Macroabc models for other countries and simulation properties of some specifications and levels of coefficients. Some estimates provide coefficients, but with low significance, so high inaccuracy margins. Also in that case we have to analyse the simulation properties: does the model converge, does it give plausible results? These simulation properties were analysed when the behavioural equations were brought into the model and some forecasts with the model as a whole were made.

Actually, also for the semi-behavioural equations we had also the check if they give plausible results and a model that converges as a whole.  But these checks for each (semi)behavioural equation only can be made if the model runs as a whole, with already the other (semi)behavioural equations in the model. That is why we had constructed a test version of the model.  This test version had very provisional specifications and levels of coefficients. Then we brought step by step the coefficients in the behavioural equations, analysed the simulation properties and discussed the whole. That resulted in the improved pilot version of the model. Then we improved some parts of the model, and then we had to look at the plausibility of the whole again. Etc. etc. At the end this process converged into the KTMM version of August 2000, the first operational version. And afterward every year the model is updated and improved and tested again.  

We followed five rules in bringing the behavioural equations in the test version:

  1. Each coefficient has only one figure behind the comma (to demonstrate that these parameters are not that accurate). Some insignificant coefficients are left out. 
  2. We avoided to bring constant terms into the equations (because of the bad simulation properties in  longer term projections) 
  3. Those coefficients that actually represent shares, are chosen on such levels that their sum adds to 1,0 
  4. The capacity rate is only in the investment price used as explanatory variable. 
  5. Some additional variables were used in the estimation process, but they are not in the model. In those cases we used a familiar variable

This resulted in the formulas and coefficients in the behavioural equations, see table on next page.

We can look at the plausibility of the forecast results, not only concerning the variable in the equation under construction, but also the effects on all other variables. Furthermore we can also look at the results compared to a reference path (base line). The model has a block in which the results of an actual run are automatically compared with the results of a reference path (frozen values) for all variables in the sheet Model.

Next diagram shows these relationships and the following box presents the overview of formulas in behavioural equations (lags not mentioned). Then we will discuss the main sheets of the model.

 

The Flow Diagram of the Structure of KTMM

Figures refer to equation numbers in Box of main behavioural equations

Box: List main behavioural equations in KTMM

 (Lags not mentioned)

Components of GDP:

  1. Consumption: changes with disposable income (98% wages, 96% profits) (lags) 
  2. Private investments volume growth: flexible accelerator (weight 50%) and profits minus interest (weight 50%) and crowding-in: 0,3*(1+2*GDP real growth) +0,4*(disposable profits income minus long interest rate*invested wealth)+0,3* public investment - 0,4*public investment five years lagged -0,06*change in credits.  
  3. Export volume % change: 2* (world market price minus export price) + 0,8* world trade growth (lagged) + 0,2* % change in investment/GDP rate. 
  4. Import volume % growth: 1,3*real final demand growth - 0,6* (import price minus consumption price)  
Prices:

  1. Consumption price % change:  0,7* wage costs + 0,3*import price + 100% change in indirect tax pressure, 
  2. Export price  % change: 0,4* wage costs + 0,3*import price + 0,3*competitors price + 0,2*real interest  rate +change in indirect tax pressure, 
  3. Investment price % change:  0,5* wage costs +  0,5*import price + 0,4*(capacity utilisation rate -1) + 0,1* real interest rate (t-1), 
  4. Wage rate businesses: 5 + 0,8*consumer price +0,7*labour productivity trend + 0,4* increase pressure direct (wage) tax rate + 2* increase wage employment businesses +(for the time being - 0,05*informal sector rate and -0,1 increase in informal sector rate)
Employment

  1. Employment businesses % change: -0,4* real wages + 0,8* real GDP growth; Informal sector number % change:  For the time being we take:  +0,6*population % growth -0,6* employment % growth,
                    Monetary variables:

  1. Interest rate: changes with 0,1*(consumption price change minus inflation target (exogenous), 
  2. Exchange rate changes: +1,0*(domestic minus foreign % change in prices, five years moving average) +0,3*(domestic minus foreign interest rates changes), 
  3. Money supply: increases with 0,9*Real GDP growth and 100% inflation target

 

Sheets of the Kenya Macroabc Model

The Excel file with the Kenya Model (KIPPRA-Treasury Macro Model) consists of:

  1. Sheet MANUAL: contains a detailed manual organized in blocks:
  • Design of the consistency framework 
  • Discusses every sheet of the model 
  • Discusses calibration results of all semi-behavioural equations 
  • The specification and coefficients of all behavioural equations 
  • How to run a simulation 
  • Sources and literature list 
  • Micro simulation for average marginal wage tax rate 
  • Activities performed during the construction stage of the model project.
  1. Sheet HELP: variants for fifty variables, ready to be used in all kind of combinations of variables and years. 
  2. Sheet SOURCE: contains the information about macro time series that we gathered from several sources. Column A gives the name of the time series, column B the unit and column C and D indicate the source. Column E and subsequent columns display calendar years, starting with 1971. These columns contain the values from the different sources. Sheet SOURCE does not contain any formulas!  
  3. Sheet MODEL: its Northwest quadrant contains cells with values of the so-called primary variables for the years in the past. All these values are calculated by formulas or references only using information from sheet SOURCE. The Northeast quadrant contains formulas that calculate future values for the primary variables. The cells in the Southern part of the sheet Model contain help variables (also called auxiliary variables) their value is calculated by formulas using only information from the primary variables above in this sheet. These help variables are used as explanatory variables in the (semi-) behavioural equations. So the time series in sheet MODEL contain all time series that are needed to estimate the behavioural equations and to calibrate the semi-behavioural equations. Sheet MODEL does not contain any input of values (other than calculated by formulas or references to sheet SOURCE. In column C of sheet MODEL you can click in each row (variable) to go immediately to the relevant part of sheet Manual where you will find background information concerning that variable.
Scheme of the Kippra-Treasury Kenya Macro Model (KTMM)

Table: Main worksheets of Kenya Model



Table: Overview of all sheets in KTMM




  1. Sheet HISTAN contains time series and graphs to analyse the relation between variables.  
  2. Sheet OUTPUT contains the output of the key variables (in calendar years). All output variables are calculated using values from sheet MODEL and no other sheet. The OUTPUT sheet gives the variables in different presentations:
  • Selected Annual Economic Indicators (in the format of CBK as far as possible), 
  • Uses of Resources in format SNA of CBS, 
  • Monetary Survey (consolidated accounts of the banking system) in format CBK,
  • Institutional Sector Accounts (flow of funds between institutional sectors)
  • Labour Market Survey
  • National Resources and Expenditures
  • From Disposable income via consumption and investments to financial transactions 
  • Simple Social Accounting Matrix (SAM)
  • Central Government Revenues, Expenditures and Fiscal data.  

  1. Sheet FY: the same information as in sheet OUTPUT for Calendar Years for the Central Government also for Fiscal Years. In sheet FY is, unlike in sheet OUTPUT, also information used from another sheet than MODEL. In this sheet we use also information from sheet SOURCE,
  2. Sheet THEORY:  This sheet contains the text of the first theory paper,
  3. The sheet CALIBRATION contains for each (semi)behavioural equation a graph with a comparison of the actual figures of the historical time series, and the simulations based on the forecasting formula, using the realised, actual figures for the explanatory variables, 
  4. The sheet TAX contains the calculation of the average marginal wage tax rate, based on the tax rates per bracket and a division of wage earners to income brackets,
  5. The sheet MICROEXPORT contains micro data for tea and coffee, to be used for the break down of the export equation, 
  6. The sheet ADD-FACTOR contains the add-factors that are in use in the (semi) behavioural equations in sheet MODEL,
  7. Sheet sectors gives a breakdown of forecasts to production growth of sectors of industry,
  8. Sheet IMF OUTPUT gives some of the output variables in IMF format,
  9. Sheet decomposition gives a breakdown of the forecasts for main variables to the contribution of each explanatory variable. This sheet makes analyses of forecasts and variants more easy,  
  10. Sheet MONITOR gives monthly information concerning CPI etc. and a calculation scheme to judge the plausibility of the forecast of the running year, given the already available monthly figures and trends, 
  11. Sheet LOGBOOK is used to document all changes in the model that are brought into the most recent official version of the model.

 

How to run a simulation, some examples

What will be the effects on the economy and government finance if wages get an impulse of additionally 2% each year, starting in 2003?

Start the PC, go in Excel, read KTMM.XLS, then you have two ways to run a variant, in sheet MODEL or in sheet HELP.

  1. Go to sheet MODEL, Go to cell AK35, change the 3 before +HELP in 5, enter, copy cell AK 34 to AL34 .. AO 34 (otherwise the impulse is only in the year 2003). Push F9 (for recalculation if the default is not on ''automatically''). Then look at the results: in AQ244.. AU244 you read the wage bill in this simulation variant minus its value in the base line. Look in AK.. AO to see the level of the wage bill in this simulation run. See the effects to the wage rate in sheet OUTPUT, AQ38 – AU38 and AK38 AO38. Have also a look to the changes in other key variables.  
  2. You can also run this simulation by setting in sheet HELP AK34..AO34 on 2.
What will be the effects on government finance and the economy of 10% increase of the VAT of local manufactures?

Go to cell AK67 in sheet MODEL. Edit (F2) the formula in that cell: multiply with 1,10. (Do not copy this to the right of this cell, unless you want to increase the revenues of these taxes year after year with 10%). Push F9 for recalculation and then go to see the results in AQ67..AU67, and analyse the other differences with the base line in columns AQ..AU in sheet OUTPUT

What will be the effects on government finance and the economy of 10% increase of the corporate income tax?

Go to cell AK77 in sheet MODEL. Edit (F2) the formula in that cell: multiply with 1,10. (Do not copy this to the right of this cell, unless you want to increase the revenues of these taxes year after year with 10%). Push F9 for recalculation and then go to see the results in AQ77..AU77, and analyse the other differences with the base line in columns AQ..AU in sheet OUTPUT

What will be the effects of 10 % higher exchange rate level in 2003 on government finance and the economy?

Go to cell AK142 in sheet HELP and change zero in 10. Check what happens in MODEL row AW142. Then see the effects in deviation of the base line in column AQ..AU in sheet OUTPUT.