A carregar...
Publicação
Time Series Forecasting: An Application to Balance Sheet
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.96 MB | Adobe PDF |
Orientador(es)
Resumo(s)
The Internal Capital Adequacy Assessment Process (ICAAP) provides a
qualitative and quantitative assessment of capital risks to which banking
institutions are exposed to in their activity. Caixa Geral de Dep´ositos
(CGD) is a relevant player in the Portuguese banking system, and as such
it has to perform an ongoing review of ICAAP exercise to evaluate its
ability to identify, assess, mitigate and report on its risks.
In order to properly quantify all the risks the institution is exposed to,
several models need to be developed to help estimate the amount of capital
that is needed to cover potential unexpected losses arising from each type
of risk. Given the European and Portuguese guidelines these models also
have to comply with certain requirements defined by Banco de Portugal,
European Central Bank (ECB) and European Banking Authority (EBA)
regarding ICAAP exercise.
One of the risks CGD is exposed to is the risk of an unfavourable evolution
of the main credit items in its Balance Sheet and as such, it is necessary
to estimate the evolution of certain credit items (in terms of their volumes
and spread rates). These estimations are needed for relevant segments
such as housing credit, consumer and other credit, public sector credit,
real estate activities credit, non-financial corporate credit and term and
sight deposits. To estimate the evolution of these balance sheet items, a
robust and reliable methodology must be applied, so that it can truly help
strategic decision-making process over a horizon period of three years and
the appropriate amount of capital can be allocated.
At CGD, Balance Sheet credit volumes and spread rates had been being
estimated through multiple linear regressions to which macroeconomic indicators
are added as explanatory variables. The problem with this methodology,
is that these type of dependent and explanatory financial variables
are usually in the form of time series, indicating the existence of correlation
between any observation and the previous one, meaning that there is
dependence on the past historical information. Applying multiple linear
regressions to this type of data leads to poor statistical results and to the
non-compliance of all the statistical assumptions linear regressions must
respect.
Within this context, the need to turn to a more adequate and robust
methodology became more evident and time series forecasting appeared
to be the so long needed solution that would allow to reach reliable statistical
results.
Time series forecasting is commonly used in economics and finance, denoting
a robust technique to predict macroeconomic variables representing a feasible approach to apply to estimate CGD’s main credit volumes and
spread rates of the balance sheet. In this project, we investigate the estimation
of Balance Sheet credit volumes and spreads rates using time series
forecasting aiming to assess the models suitability to quantify the risk of
unfavourable balance sheet evolution of the main credit segments.
The models proposed for this purpose, are the Autoregressive Integrated
Moving Average with exogenous variables (ARIMAX) models. The results
obtained proved to have robust statistical results and high performance,
which were verified by analysing residuals statistical behaviour and key
performance indicators such as the Mean Squared Error (MSE) and the
Akaike Information Criterion (AIC) of the final models selected for each
target variable.
Descrição
Internship Report presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science
Palavras-chave
CGD ICAAP balance sheet projection risk forecasting time series ARIMAX