The basis of formalized forecasting methods is mathematical theory, which increases the reliability and accuracy of forecasts, facilitates the processing of information and forecast results, and significantly reduces the time it takes to produce them.
Formalized forecasting methods can be divided into two groups: extrapolation methods and mathematical modeling methods. Extrapolation consists of studying stable trends that have developed in the past and present. economic development and transferring them to the future. With simple extrapolation, all previously operating factors that determine the trend under study in the past and present will remain unchanged in the future. However, maintaining the trends of the past and present unchanged for the future is most often unlikely. And therefore, although extrapolation is the basis of any forecast, it is capable of producing an effect only in a very narrow time range of a relatively not particularly complex process.
It is necessary to distinguish between formal and predictive extrapolation. The formal one is based on the assumption that past and present trends in the development of the object will be preserved in the future. In forecasting, the actual is linked to hypotheses about the dynamics of the object under study, and alternative changes in the object itself and its essence are taken into account in the future.
Extrapolation forecasting methods are based on the study of time series, which are sets of measurements ordered in time various characteristics forecasting object under study. Extrapolation in forecasting assumes that the process of change in the variable under consideration is a combination of two components xr - regular (deterministic non-random)

naya) and ex - random. The time series y can be represented as
(1)

The regular component is called a trend. These terms contain an intuitive idea of ​​the essence of the analyzed process, cleared of interference (intuitive because for most processes it is impossible to unambiguously separate the trend from the random component). The regular component (trend) xg characterizes the dynamics of the development of the process as a whole, the random component e% reflects random fluctuations or noise of the process. Both components of the process are determined by a functional mechanism that characterizes their behavior over time.
The task of the forecast is to determine the type of extrapolating functions x% and еr based on the initial empirical data and parameters of the selected function. The first step is to select the optimal type of function that gives the best description of the trend. The next stage is calculating the parameters of the selected extrapolation function.
When estimating dependency parameters, the most common method is least squares and its modifications, exponential smoothing method, adaptive smoothing method, moving average method, etc. The least squares method (LSM) requires finding the parameters of a trend model that minimize its deviation from the points of the original time series, i.e. minimize the sum of square deviations between observed and calculated values.
P (A
5=1 У1~У1
A
where y1 are the calculated values ​​of the original series;
u. - actual value of the original series; n is the number of observations.
P

The trend model can have different forms; its selection in each specific case is carried out according to a number of statistical criteria. In practical research the following are most often used:
y = ax + b (linear);

y = ax2 + b + c (quadratic);
y - xn (power);
y = ax (indicative);
y = aex (exponential);
A
U - (logistics).
Widely applied linear function, or linearizable, i.e. reducible to linear, as the simplest and most consistent with the original data.
The classical least squares method assumes the equivalence of the initial information in the model. In actual practice, the future behavior of a process is determined by later observations in to a greater extent than earlier. Decreasing value more early information(discounting) can be taken into account, for example, by introducing into model (2) some weights B. lt; 1. Then

(9)
The coefficient can be presented in various forms: numerical form, functional dependence, but in such a way that as we move into the past, the weights decrease.

For this purpose, modifications of the least squares method are used.
The least squares method is widely used in forecasting due to its simplicity and the possibility of implementation on a computer. The disadvantage of the method is that the trend model is rigidly fixed, and this makes it possible to use it only for short lead periods, i.e. for short-term forecasting.
The exponential smoothing method makes it possible to obtain an estimate of trend parameters that characterize not average level process, but the trend that had developed at the time of the last observation, that is, it allows you to estimate the parameters of the model that describes the trend that formed at the end of the base period, and thus does not simply extrapolate existing dependencies into the future, but adapts to changing conditions over time. The advantages of the method are that it does not require an extensive information base, but involves its intensive analysis from the point of view of the information value of various members of the time sequence. Models describing the dynamics of the indicator have a simple mathematical formulation, and the adaptive evolution of parameters makes it possible to reflect the heterogeneity and fluidity of the properties of the time series. The method is used for short- and medium-term forecasting.
The moving average method makes it possible to align a dynamic series by dividing it into equal parts with the obligatory coincidence in each of them of the sums of model and empirical values.
The extrapolation method also includes the method called “Markov chain”. The basis of the forecast, built on the basis of simple Markov chains, is the calculation of the transition matrix, the elements of which are the probabilities of the transition of the predicted parameters from one state to another, from
one meaning to another. If we have A = (i.e.
matrix of predicted dimension indicators (t x T), where Ai is the value of the i-th indicator at time t, and if
the transition matrix P is known, then the forecast is calculated as follows:
A+1 = pL;A+2 = ^2A--A+* = -p*A, lt;10)
where is the vector of values ​​of the predicted indicators at time t.
Procedure for calculating transition matrix elements
p = (p1) )„ g, y = 1Gy (P)
involves determining the total changes in indicators Ai for each point in time t, i.e.
,?A=t=g? <12gt;
(if we forecast needs, then this will be the total resource requirement by year).
Then we determine the values ​​of the chain indices for the quantities
A
G = -4- = 1,T
Based on chain indices, we determine the possible values ​​of the predicted indicators with the structure unchanged at moments (?+1):
5 = b?i_ = bl = ]7t,
I t
those. the index is multiplied by the value of this indicator at the appropriate moment (?+1).

Elements Bi form a matrix = ) of dimension (пхТ).
The discrepancy between the real change in indicators Ay and the hypothetical one is found as their difference:
A?*, *+1 = A^ #+1 - 8i.
These mismatch values ​​determine the change in the structure of the process under study (if this is consumption, then the structure of resource consumption) and represent a generating vector
A?m+1 = (A?t)= * A?P,*+1) .
Then a normalized vector is formed that determines the change in the value of the i-th indicator in (? + 1) year compared to t-th year. It is determined by the formula
1
h‘m - - (13)
? + 1 *=1
The obtained values ​​make it possible to form the ?-th row of the matrix of the corresponding transition Рт.
Using a similar scheme, transition matrices are calculated sequentially for different points in time. The forecast is directly implemented according to formula (10).
Implementation of forecasts using Markov chains allows, as they arrive, new information regularly correct errors, take into account information inaccuracy of the forecast, which increases the reliability of the results obtained. This method can be used to predict many indicators that change from year to year simultaneously, but functional connections between them are not directly established due to the lack of information or the extreme complexity of these connections. An example is the forecast of the resource needs of sectors of the national economy. When implementing this forecast, not only the volumes, but also the very structure of resource consumption by various industries are established for the future.

Extrapolation methods based on extending past and present trends into the future can be used in forecasting only with a lead period of 3-5 years. With longer forecast periods they do not give accurate results. Using extrapolation methods, quantitative parameters are studied large systems, quantitative characteristics of economic, scientific and production potentials, data on the effectiveness of scientific and technological progress, characteristics of the relationship between individual subsystems, blocks, etc.
A large group of formalized forecasting methods consists of modeling methods. With their help, models are constructed based on a preliminary study of the object and the identification of its essential characteristics, experimental and theoretical analysis of the model is carried out, the results are compared with the object data, and the model is adjusted. Modeling is widespread not only in forecasting, but also in planning. The impetus for the development of formalized methods, including modeling methods, was the use of electronic computers (computers). Their development showed new stage- the stage of economic-mathematical methods (EMM), which combined mathematical theory and computer capabilities.
EMMs and computers based on the methods of applied mathematics and mathematical statistics have made it possible to significantly expand the possibilities of application and areas of use of formalized methods. Thus, it became possible to more deeply reveal the relationships in the national economy, comprehensively substantiate changes in economic indicators, speed up the receipt and processing of information, carry out multivariate calculations of plans, forecasts, programs and select best option according to a given criterion.
In planning and forecasting, various types (types) of models are distinguished: optimization, factor, structural, inter-industry balance models, etc. Depending on the level of aggregation, the same type can be applied to various economic objects, therefore the following models are distinguished: macroeconomic, inter-industry, inter-district , from
sectoral, regional and microeconomic (at the enterprise, association level).
An economic-mathematical model of any kind is a formalized description of the process or object under study in the form of mathematical dependencies and relationships.
Optimization models are based on the choice of an optimality criterion, based on which, by comparison various options the best (optimal) option is selected. The optimization economic-mathematical model consists of a target function and a system of constraints. The objective function describes the optimization goal and reflects the dependence of the indicator used for optimization on independent variables (constraints). The system of restrictions reflects objective economic connections and dependencies and is a system of equalities and inequalities, for example, between resource consumption or the values ​​of technical and economic indicators and established limits, as well as product output limits. The influence of each variable on the value of the objective function is expressed by a coefficient-indicator, the extremum of which acts as an optimality criterion. Examples of optimization models in planning and forecasting: models for optimizing the development and location of production, models for optimizing the structure of production of industrial products, models of the agro-industrial complex, models of transport problems, with the help of which suppliers are rationally attached to consumers and minimum transport costs are determined, and others.
Examples of macroeconomic models include static and dynamic models of inter-industry balance.
The static model looks like:

X. - gross production y-th industry-consumer (y = 1, n); X. - gross production of products of the 1st supplier industry (1=1, p);
u. - final volume Products Mr. industry.
In this case, УЦ1ач] represents an intermediate product (the amount of products of the 1st industry used in the y"th industry in the production process).
The statistical model of the input-output balance can also be expressed as follows:

where b.. is the coefficient of total material costs, reflecting the amount of production of the 1st industry, necessary at all stages of production to obtain a final unit products industry.
The coefficients of direct and full costs differ in that the former are determined per unit of gross output of the industry and are the industry average, while the latter are calculated per unit of final product and are national economic. The coefficients of total costs exceed the coefficients of direct costs by the amount of indirect costs.
The dynamic model of the intersectoral balance characterizes the production relations of the national economy over a number of years (i.e., reflects the process of reproduction in dynamics) and ensures the linking of the production forecast plan with the forecast plan of capital investments. The simplified model has the form

where t is the year index; AFu - products of the *th industry, sent as industrial capital investments to expand production in the yth industry; Z. - the sum of the final products of the i-th industry, with the exception of products aimed at expanding production.
The correlation-regression method makes it possible to quantitatively study the influence of various factors on the level of a parameter characterizing the planned (predicted) phenomenon or process, allows you to separate imaginary connections from real ones and express this connection in mathematical form (through a regression equation) and reveal the effect of factors on this parameter. The correlation-regression method is widespread and solves two main problems:

• establishes the degree of closeness of connection between the planned (predicted) parameter and the factors influencing it;
• using regression equations, determines the form of the relationship between the planned (predicted) parameter and the factors influencing it.

Page 3

One of the first tasks in this area is to provide engineering methods for pre-design analysis of existing control systems, formalized methods for the analysis itself and the presentation of its results, allowing analysis to be carried out using a computer. Such methods and models can be obtained based on the apparatus of graph theory and matrix algebra.

The construction and analysis of tests can be performed using the methods outlined in works devoted to the study of formalized methods for algorithmizing diagnostic processes. Methods for minimizing defect search programs based on assessing the amount of information should be considered somewhat separate. It is shown that evaluation of the defect search is possible for both equally probable and non-equiprobable events. Despite the apparent simplicity of the problem, determining the optimal defect search program for systems with unequally probable element defects is difficult.

The concept of a discrete converter arose on the way to the application of automata theory in the study of certain problems in programming theory and the construction of formalized methods for designing computer structures.

In this case, in a number of cases, correlations between individual indicators can be taken into account, but the above-mentioned main drawback of formalized methods remains valid - the impossibility of taking into account the inequality of individual components of the initial information.

Of great interest is the study of the processes occurring in reservoirs when various chemical reagents are added to the injected water and, even more so, the creation of fairly formalized methods for finding the best technology for influencing the deposit.

In order to reduce the influence of subjective factors when making calculations and to increase the degree of reliability and reliability of the results obtained, an economic and mathematical model has been developed that allows, on the basis of formalized methods, to determine the optimal banknote structure of cash in circulation. This model assumes ranking at the macro level the volume of demand for banknotes and coins depending on the level of income of business entities.

When planning the development of a system, it is necessary to take into account such factors as the relationship between technology and resource tasks, the need to use several types of resources, etc. The use of formalized planning methods allows you to reasonably approach the choice of the optimal priority for developing information support for automated control systems.

Systems approach in chemical technology [4, 45, 47, 49] is a methodological direction, the main task of which is to develop a general methodology, as well as non-formalized or heuristic and formalized methods for the integrated study and creation of complex CTP and CTS different types and classes. The systems approach is based on one of the most important laws of dialectical materialism - the law of universal connection, interaction and interdependence of phenomena in the world and society, based on which any studied phenomena are considered not only as independent systems, but also as subsystems of some larger system.

In the future, of course, more in-depth formulations and substantiated mathematical models of the processes under consideration will be found, but even these first results make it possible to show the potential of formalized methods for solving problems to find optimal design solutions, which give the technologist the opportunity to consider all possible options implementation of technological processes and the most efficient use of available geological and technical information.

The systems approach in chemical technology is a methodological direction, the main goal of which is to develop overall strategy, as well as informal, or heuristic, and formalized methods for complex research and creation of complex chemical-technological processes (CTP) and CTS of various types and classes. The systemic approach assumes that the interconnection and interaction of CTPs included in a certain CTS ensure the appearance of fundamentally new properties in this CTS that are not inherent in its individual non-interconnected CTS.

It is necessary to use some of the most typical representative products, for example, of all varieties of apples - one. There are no formalized methods for such selection. As a result, the problem of constructing a good index shifts to the sphere of largely intuitive assessments: how many and what products to leave in the set so that, on the one hand, not to distort the result, and on the other, to ensure the practical feasibility of the task of obtaining initial information on prices and volumes.

It turned out that solving this problem using methods known from the compulsory school mathematics course is difficult. Therefore, a formalized solution method based on the use of matrix calculus will be proposed here. Moreover, the order in which this exception occurs is chosen by the solver himself. For solving large systems (for example, with 2000 equations and the same number of variables), the use of the above methods is inappropriate. Here we will consider one of the modifications of the method sequential elimination of variables, developed by Carl Friedrich Gauss in 1938. It allows you to write down the decision process in a convenient, understandable form and facilitates control of calculations.

At the same time, practice shows that, based on intuitive considerations, the experimenter, as a rule, cannot competently choose enough complete system competing hypotheses, especially if we're talking about about multistage reactions. There is an urgent need to develop formalized methods for solving this stage general scheme based on the stoichiometric analysis of the reaction system. The use of stoichiometric analysis techniques allows the researcher to determine all possible reactions among all molecular types of the reaction system, on their basis to build systems of hypotheses about the possible mechanisms of the complex chemical reaction and for each mechanism correctly derive the correct kinetic model, presented in the most convenient and compact form.

Such a systematic study of the formation and circulation of scientific information has not only independent significance. It is necessary when developing any formalized methods for analyzing information arrays and optimizing them. This is due to the fact that automation of management of organizational and technical systems cannot be effective without taking into account the specifics of the object of its application.

The version of the plan developed using formalized methods is the basic one, and new options appear under the influence of new ideas.

Formal forecasting methods

Forecasting using formalized methods is carried out according to a strictly defined algorithm and form.

Formalized methods are based on mathematical theory, which increases the reliability and accuracy of forecasts, reduces the time required for their implementation and facilitates information processing and evaluation of results. Formalized forecasting methods include methods extrapolation and methods modeling(Fig. 4).

Rice. 4. Formalized forecasting methods.

The essence of extrapolation consists in studying the stable trends in the development of the forecast object that have developed in the past and present and transferring them to the future.

There are formal and predictive extrapolation. Formal is based on the assumption that past and present trends in the development of the forecast object will be maintained in the future; at forecast actual development is linked to hypotheses about the dynamics of the process under study, taking into account changes in the influence of various factors in the future.

Extrapolation methods are the most common and well-developed. The basis of extrapolation forecasting methods is the study of time series. Dynamic series - it is a set of observations obtained sequentially over time.

Widely used in economic forecasting a method of mathematical extrapolation, in a mathematical sense meaning the extension of the law of change of a function from the region of its observation to a region lying outside the observation segment. A trend described by some function of time is called a trend. Trend - This is a long-term trend in economic indicators. The function is the simplest mathematical-statistical (trend) model of the phenomenon being studied.

It should be noted that extrapolation methods must be applied to initial stage forecasting to identify trends in indicators.

Feature selection method- one of the common extrapolation methods. The main stage of trend extrapolation is the selection of the optimal type of function that describes the empirical series. To do this, preliminary processing and transformation of the source data are carried out in order to facilitate the selection of the type of trend by smoothing and leveling the time series. The task of selecting a function is to select based on actual data (x i , y i) the shape of the dependence (line) so that the deviations (∆ i) of the data of the original series y i from the corresponding settlement y i , those on the line were the smallest. After this, you can continue this line and get a forecast.

Where n– number of observations.

Model selection is carried out using specially developed programs. There are programs that provide the ability to model economic series using 16 functions: linear (y= a + bx), hyperbolic of various types (y = a + b/x), exponential, power, logarithmic, etc. Each of them can have its own specific area of ​​application in forecasting economic phenomena.

So, the linear function (y = a + bx) used to describe processes that develop uniformly over time. Parameter b(regression coefficient) shows the rate of change of the predicted at when it changes X.

Hyperbolas well describe processes characterized by saturation when there is a factor restraining the growth of the predicted indicator.

The model is selected, firstly, visually, based on a comparison of the type of curve, its specific properties and the qualitative characteristics of the trend of the economic phenomenon; secondly, based on the value of the criterion. The sum of squared deviations S is most often used as a criterion. From the set of functions, the one that corresponds to the minimum value of S is selected.

The forecast involves extending the past trend, expressed by the selected function, into the future, i.e. extrapolation of time series. The value of the predicted indicator is determined programmatically on a computer. To do this, the value of the period for which it is necessary to obtain a forecast is substituted into the formula describing the process.

Due to the fact that this method is based on the inertia of economic phenomena and the prerequisites that General terms, which determine development in the past, will not undergo significant changes in the future; it is advisable to use it when developing short-term forecasts, always in combination with methods expert assessments. Moreover, a time series can be built on the basis of data not by year, but by month and quarter.

Extrapolation using the function selection method takes into account all data in the original series with the same “weight”. The classical least squares method assumes the equivalence of the initial information in the model. However, as experience shows, economic indicators tend to “age.” The influence of later observations on the development of the process in the future is more significant than earlier ones. The problem of “aging” of time series data is solved exponential smoothing method with adjustable trend. It allows us to construct a description of the process (time series) in which later observations are given greater “weights” compared to earlier ones, and the “weights” of observations decrease exponentially. As a result, it is possible to obtain an estimate of trend parameters that characterize not the average level of the process, but the trend that had developed at the time of the last observation.

The rate of data aging is characterized by the smoothing parameter A. It varies within 0 < A< 1.

Depending on the parameter value, forecast estimates take into account the influence of the initial series of observations differently: the more A, the greater the contribution of recent observations to the formation of the trend, and the influence of the initial conditions quickly decreases.

At low A forecast estimates take into account all observations, while the decrease in the influence of “older” information occurs slowly, i.e. the less A, the more stable the data is, and vice versa.

In the field of economic forecasting, the most commonly used limits are
0,05 < A< 0,3 . Meaning A in the general case, it should depend on the forecast period: the shorter the period, the larger the parameter value should be.

This method is implemented on a computer using specially developed programs in the “time series” block, which is an integral part of the economic calculation package.

Modeling involves constructing a model based on a preliminary study of an object or process, identifying its essential characteristics or features. Forecasting economic and social processes using models includes the development of a model, its experimental analysis, comparison of the results of forecast calculations based on the model with actual data on the state of an object or process, adjustment and refinement of the model.

Depending on the level of management of economic and social processes, macroeconomic, intersectoral, interdistrict, sectoral, regional models and micro-level models (firm development models) are distinguished.

According to aspects of economic development, forecasting models for the reproduction of fixed assets are distinguished, labor resources, prices, etc. There are a number of other features for classifying models: time, factor, transport, production.

IN modern conditions development of modeling and practical application models began to be given particular importance in connection with the strengthening of the role of forecasting and the transition to indicative planning.

Let us consider some of the most developed economic and mathematical models that have been widely used in the practice of economic forecasting,

TO matrix models include models of input balance (IB): static and dynamic. The first are intended for carrying out forecast macroeconomic calculations for the short-term period (year, quarter, month), the second - for calculating the development of the country's economy in the future. They reflect the reproduction process in dynamics and link the forecast of production of products (services) with investments.

The static model of the MOB in the national economic balance system has the form

Where t- year index; I t ij- industry products i, directed as productive investments into t- m year to expand production in the industry j; Y t i - final product volume i-and industries in t-th year, with the exception of products sent for production expansion.

An intersectoral balance formed on the basis of models can be used to solve many problems: forecasting macroeconomic indicators, intersectoral connections and flows (supplies), economic structure, industry costs, price dynamics, production efficiency indicators (materials, energy, metal, chemical and capital intensity).

Economic-statistical models are used to establish quantitative characteristics of the relationship between dependence and interdependence of economic indicators. The system of this kind of models includes: single-, multi-factor and econometric models.

Examples one-factor models: y = a + bx; y = a + b/x, y = a + b log x u etc.,

Where at - the value of the predicted indicator; A - free term that determines the position of the starting point of the regression line in the coordinate system; X - factor value, b - parameter characterizing the rate of change at per unit X.

Multifactorial models allow you to simultaneously take into account the impact of several factors on the level of the predicted indicator. In this case, the latter acts as a function of the factors:

y = f (x 1 , x 2 , x 3 , …, x n)

where x 1, x 2, x 3, …, x n - factors.

At linear dependence multifactor models can be represented by the following equation:

y = a 0 + a i x i + a 2 x 2 + ... + a a x a ,

Where a 0- free member; A a 1 , a 2 , …, and p - regression coefficients showing the degree of influence of the corresponding factor on the predicted indicator with a fixed value of other factors.

With a nonlinear relationship, a multifactor model can have the form

y = a x a 1 * x a 2 2 * … * x an n.

Multifactor models are used in forecasting macroeconomic indicators, indicators of demand for products, costs, prices, profits, etc.

Econometric model called a system of regression equations and identities that describe the relationships and dependencies of the main indicators of economic development. A system of economic and mathematical models of econometric type serves to describe complex socio-economic processes. Factors (variables) of the econometric model are divided into exogenous (external) and endogenous (internal). Exogenous variables are selected so that they influence the modeled system, but are not themselves influenced by it. They can be entered into the model based on expert assessments. Endogenous the variables are determined by solving stochastic and identity equations. For each endogenous variable, several variants of regression equations are estimated using the least squares method and the best one is selected for inclusion in the model. For example, industrial investments depend on the amount of profit (endogenous factor) and the price index for investment goods (exogenous factor).

Inter-industry balance can also be an organic part of the econometric model. Typically, the number of model equations is equal to the number of endogenous variables.

Econometric models make it possible to predict wide circle indicators (GNP, personal income, consumption of goods and services, etc.). In conditions of automation of calculations, it becomes possible to develop alternative options for economic development, taking into account changes in external and internal conditions (factors). It should be noted that the use of econometric models requires the creation of data banks and the training of highly qualified specialists in the development and implementation of these models.

Control questions

1. Name the main forecasting methods and give them brief description?

2. Describe the main methods of individual expert assessments (“interviews”, analytical method) and the script writing method?

3. Describe the main methods of collective expert assessments (idea generation, “635” method, “Delphi” method, commission method)?

4. Reveal the essence of extrapolation methods and give them a brief description?

5. The essence of modeling methods is forecasting?

6. Describe the main types of forecast models (matrix, economic-statistical, econometric)?

Information sources

1. Alekseeva M.M. Planning the company's activities: Educational and methodological manual. – M.: Finance and Statistics, 1999

2. Basovsky L.E. Forecasting and planning in market conditions. Tutorial. – M.: INFRA-M, 1999. 260 p.

3. Goremykin V.A. and others. Planning at the enterprise: Textbook / V.A. Goremykin, E.R. Bugulov, A.Yu. Bogomolov. – 2nd ed. - M.: Kolos, 2000

4. Organization of agricultural production / F.K. Shakirov, V.A. Udalov, S.I. Gryadov et al.: Ed. F.K. Shakirova. - M.

5. Forecasting and planning in market conditions. Ed. T.G. Morozova, A.V. Pikulkina. Tutorial. - M.: UNITY-DANA, 199.-318 p.

6. Chernysh E.A., Molchanova N.P., Novikova A.A., Saltanova T.A. Forecasting and planning. Tutorial. – M.: 1999. – 174 p.

7. Serkov A.F. Indicative planning in agriculture. M.: Informagrobusiness, 1996. 161 p.

8. Economic Encyclopedia/ Scientific editor. Council of the Publishing House "Economics", Institute of Economics of the Russian Academy of Sciences; Ch. ed. L.I. Abalkin. Moscow: OJSC Publishing House Ekonomika, 1999.

Formalized methods

Desk research as a method of collecting information

Quantitative Research Methods

Desk research is a set of methods for collecting and assessing marketing information contained in sources (statistical data or reports) prepared for any other purpose.

Desk methods of collecting information rely on secondary sources. Desk methods are used in preparing field research.

Desk methods are also used as independent methods of collecting information in market research.

Document analysis methods are divided into two main groups: informal (traditional) and formalized.

Informal methods do not use standardized techniques for isolating units of information from the content of a document; they require a painstaking analysis of each source; therefore, they are more often used for processing individual (unique) documents or a small array of documents, when there is no extreme importance in quantitative processing of information. Traditional analysis can serve as a prerequisite for formalized document analysis.

An alternative to informal methods of document analysis has become formalized methods, using unified (standard) methods for registering document content elements. Standardization of information collection methods freed researchers from labor-intensive registration procedures and subjectivity in data interpretation; made it possible to switch to automated registration and processing of information using special computer programs. At the same time, other problems arose: difficulties in developing unambiguous rules for recording the necessary elements and the impossibility of exhaustively disclosing the content of each individual document.

When conducting desk research, the traditional (classical) method of document analysis, information-target analysis and content analysis of documents are most often used, the main characteristics of which are presented in Table 1.

Table 1 general characteristics desk methods of collecting information

Formalized methods - concept and types. Classification and features of the category "Formalized Methods" 2017, 2018.