7.2 City system dynamics model

7.2.1 Purpose of the model

The mathematical model for calculating the integrated (system-dynamic) forecast of demographic and economic indicators (hereinafter - the Model) is intended for:

  • Determining long-term trends (regression, approximating functions) of changes in basic socio-economic development (SED) indicators of the city;
  • Forming a dynamic forecast (based on a dynamic function) of basic SED indicators, limited by the corridor of values of long-term trends;
  • Calculating correlation functions describing the measure of mutual influence of indicators on each other and susceptibility to changes in external factors.

The model converts a set of actual, annual (for at least 6 years) values for 27 macroeconomic SED indicators and 4 additional macroeconomic indicators characterizing external factors into sets of the following values for each city:

  • 27 macroeconomic SED indicators for the retrospective (until 2000) and prospective (until 2050) periods;
  • Correlation matrices of dimension 31 indicators x 31 indicators x 50 annual periods, reflecting the degree of mutual influence of changes in SED indicators.

7.2.2 Basic form of the model

The basic differential equation of the Model is the damped oscillation equation, describing the change in system indicators \(x(t)\) over time: \[\begin{align*} \frac{d^{2} x(t)}{dt^{2}}+ \gamma \frac{dx(t)}{dt} + \omega^{2}x(t)=f(t), (1) \end{align*}\] where system parameters are:

\(\gamma\) - characteristic of the system’s oscillation damping due to resistance (“friction”) in the system; \(\omega\) - the natural frequency of the system’s oscillations in the absence of resistance (“friction”); \(f\) – the set of external influences.

The dependencies of indicators on time are represented by the sum of 42 basic functions with different oscillation periods:

\[\begin{align*} x(t)=\sum_{i=1}^{21}(C_{i}cos\omega_{i}t+S_{i}sin\omega_{i}t)e^{-a_{i}t}(2) \end{align*}\] To determine the coefficients \((C, S, \omega, a)\), the Laplace transform is applied to the dynamic function \(x(t)\): \[\begin{align*} L\{x(t)\}(s)=\int_{0}^{\infty}x(t) e^{-st} dt. (3) \end{align*}\] Or explicitly: \[\begin{align*} L\{x(t)\}(s)=\sum_{i=1}^{21}(C_{i}\frac{s}{(s+a_{i})^{2}+\omega_{i}^{2}}+S_{i}\frac{\omega_{i}}{(s+a_{i})^{2}+\omega_{i}^{2}}). (3.1) \end{align*}\] Based on available statistical data, the parameters \((C, S, \omega, a)\) of equation (3.1) are calculated, and then a dynamic function \(x(t)\) and the rate of change \(dx(t)/dt\) are built. For all analyzed SED indicators, oscillations with periods from 6 to 68 years make the greatest contribution.

7.2.3 Description of input variables

The Model uses retrospective values of the following indicators as input data (for at least 6 years):

No Code Indicator Unit
1 C013 Population as of Jan 1 people
2 C034 Number of births people
3 C035 Number of deaths people
4 C037 Population inflow (migration) people
5 C038 Population outflow (migration) people
6 C041 Working-age population people
7 C096 Commissioning of apartment buildings m2
8 C149 Total residential space m2
9 C221 Average number of employees people
10 C241 Payroll fund rubles
11 C698 Taxable money income of individuals rubles
12 C699 Population income: social and other payments rubles
13 C712 Money income per capita rub/year
14 C713 Mandatory payments and contributions per capita rub/year
15 C714 Real disposable income per capita rub/year
16 C757 Integral provision of social infrastructure %
17 C049 Total city budget income mln rub
18 C065 Total city budget expenses mln rub
19 C087 Fixed capital investment mln rub
20 C706 Gross City Product (GCP) mln rub
21 C708 Tax revenues of all budget levels mln rub
22 C709 Local budget revenues (tax) mln rub
23 C433 Current assets mln rub
24 C475 Revenue mln rub
25 C711 Total cost (revenue - net profit) mln rub
26 C707 Gross value added of SMEs mln rub
27 C409 Fixed assets mln rub
28 C715 Regional Consumer Price Index (CPI) units
29 C716 Exchange rate rub / $
30 C717 Balance of payments of the RF dollars
31 C718 Gross Domestic Product (GDP) of the RF mln rub

7.2.4 Complete description of the calculation algorithm

The calculation algorithm consists of the following steps:

  1. Preparation of initial data including actual values of the model input variables specified in the List of model input variables.

  2. Determination of the initial parameters of the regression model A, B, c, d by the least squares method, minimizing the error for all indicators simultaneously for each city and each group of socio-economic indicators of the city:

  • Demographic (population, number of births and deaths, arrivals and departures, working-age population)
  • Economic (average number of employees, payroll fund, revenue and profit, SME income, fixed and current assets, investment in fixed capital)
  • City (Gross City Product, budget income and expenses, population income and expenses, tax revenues of all budget levels)
  • Social (total residential space, commissioning of apartment buildings, commissioning of residential houses, social infrastructure provision)
  1. Calculation of forecast and restoration of missing retrospective socio-economic development indicators based on the regression model.

  2. Determination of calculation errors: \[\begin{align*} \varepsilon_{[20XX]} = x(20XX)-X_{[20XX]}, (6) \end{align*}\]
    where:
    \(\varepsilon_{[20XX]}\) – the error value for each indicator for the year 20XX; \(x(20XX)\) – the value of the indicator obtained based on the regression model in step 3; \(X_{[20XX]}\) – the actual value of the indicator.

  3. Refinement of parameters A, B, c, d using the Newton method of unconditional optimization of the sum of squared errors \(\sum\varepsilon^{2}\), obtained in step 4.

  4. Iterative repetition of steps 2, 3, 4, 5 until the possible minimum of the sum of squared errors is reached.

  5. Construction of the upper and lower boundaries of the regression model forecast. The approximation error \(err(t)\) includes two components: constant \(\varepsilon_{const}\) and variable \(\varepsilon_{var}(t-t_{0})\), equal to 0 at the starting point \(t_{0}\) (in the presented results \(t_{0}\) corresponds to January 1, 2019): \[\begin{align*} err(t)^{2}=\varepsilon_{const}^{2} + \varepsilon_{var}^{2}(t-t_{0}). (7) \end{align*}\]

  6. To construct the upper \(H(t)\) and lower \(L(t)\) boundaries of the regression model indicators, the error \(err(t)\) is multiplied by the quantile \(z_{[P]}\), which depends on the confidence probability \(P\), the number of parameters, and the number of actual values (for the given example at \(P=90\%\), 16 parameters, and 90 actual values, \(z_{[90\%]} = 1.673\)): \[\begin{align*} H(t)=X(t)e^{z_{[P]}err(t)}, (8) L(t)=X(t)e^{-z_{[P]}err(t)}. (9) \end{align*}\]

  7. Determination of the dynamic function parameters \((C, S, \omega, a)\) based on available statistical data, the upper and lower boundaries of the regression model indicators based on equation (2).

  8. Formation of model values for each indicator for the forecast and retrospective periods based on the dynamic function built using parameters \((C, S, \omega, a)\).

  9. Calculation of correlation matrices for each indicator by formula (4) based on the obtained model values.

7.2.5 Scope of permissible application of mathematical models

The scope (boundary) of permissible application of mathematical models is:

– formation of inertial medium-term and long-term forecasts of basic socio-economic development indicators of cities; – restoration of missing values for socio-economic development indicators of cities in the retrospective period no earlier than 2000; – creation of correlation matrices for the implementation of models for assessing the impact of investment project parameters on the socio-economic development of cities in the medium-term (1-5 years) and long-term (5-15 years) horizons.

Back to Model start

Back to chapter start