chapter 8 Flashcards
Financial service companies have access to oppertunities and exposure to risks that involve unknown values. They often represent future developments. Some values may be unknown for a reason other than timing. Such as:
- they dont have access to nonpublic information about their competitiors, suppliers, and customers.
- company into a new market has not collected its own data observation about that new market
- company loses data in an IT system conversation and needs to estimate some data values from the lost records
- comapny may need information about a new product, such as longterm care insurance, which has unknown performance charcteristics over a long span of time
- company may be interrested in understanding a new product recently developed by competitors.
- company has not consistently sored certain data elements over time
- some data are impossible to track to a fine level of detail.
- some data reflect black-martket activities that another party is purposely hidding.
- company has no data available.
define forcasting
a process for estimating unknown future values.
what is a forecast
is an estimate or pojection of unknown future conditions or events.
what do you call the mathermatical constructs used to project unknown or future outcomes using known results or data.
Forecasting models
define the term estimation
the process for approximating unknown values.
What is an estimate?
a value used to approximate unknown values.
- developed through projections, forecasting, predictions, or prognostication.
what is a projection?
an extension of a mathematical pattern observed in known data created for the purpose of estimating unknown data.
what are quantititive forcasting methods?
use mathematical models that incorporate data about past events and values in order to project future events.
- short term forcasts.
What are two main categoeis of quantitiative forecasting models?
time-series models
casual models
what is qualitative forecasting methods
use nonmathermatical models and rely on customer opinion, sales producer opinion, espert opinion, and other subjective factors to project unknown values.
What are time-series models?
mathematical models designed for estimating unknown future values solely on the basis of known, historical data about the uknown values or values.
What is time-series data?
defined as information about a variable over successive periods of time. The specific formula used on the time-series cata to make the projection depends on the model used.
What is the name of the most simplest time-series model?
naive time-series model.
it requires neither a significant amount of data nor any computational steps. A forecaster simply use the data value(s) for the most recent period as the forecast value(s) for the next future period.
-easy to use
what is the arithmetic average model?
a simple form of time-series model.
uses the arithmatic average (the mean) of the time-series data from previous periods as the forecast value for the next future period.
When is it appropriate to use the arithmetic average model?
if the values for experience data
1) show consistent variation around the mean
2) are trend-free.
here it would produce an accurate result.
What is the simple moving average model?
closely related to the arithmatic average model. Requires findings the arithmetic average (mean) of data values but it considers only a specified number of the most recent data values.
What is a weighted moving average time-series model
assigns relative weights to the data values being use in the forecast.
What is a trend
a movement in a specific direction, such as upwards or downward, that occurs over many years.
they are series of data points with values that are consistently rising of falling over time.
What is a variation in a data series?
a is a change or fluctuation in a trend. There are 3 kinds: random, seasonal, and cyclical.
What is a random variation?
changes in a trend that are either unexpected or are onetime occurances.
- difficult to preduct and include the economic inpact on the business cycle from major wars, hurricanes, floods and volcanoes.
What are seasonal variations?
result from routine patterns that typically occur in the course of one year.
- exists when a consistent pattern of change occurs with every successive 12 month period.
What is a cycle?
a rhythmically fluctuating pattern that repeats over time with reasonable regularity.
- have predictive or explanatory value.
- more it continues the more predicatable it becomes.
What is a causal forecasting model
type of quantitative model that uses historical data and other relevant variables as a basis for describing unknown future data points.
What is a causal forecasting model
type of quantitative model that uses historical data and other relevant variables as a basis for describing unknown future data points
when are variables said to be correlated?
when a change in one variable is associated with a consistent and equivalent change.
What is correlation analysis
is a statistical technique used to measure the strength of any relationship between variables.
what do you call the degree of correlatio between variables wen expressed inin a correlation statistic known.
correlation coefficient.
What is a correlation coefficient?
a statistic that indicates how closely movement in two variables track one another
Varies between +1 and -1.
What is a +1 valye of coefficient?
this means that one variable is chamging in value by a given percentage increment, the other variable will also change in the same direction by the same percentage increment
- PERFECT POSITIVE CORRELATION
What is a 0 value of coefficient?
a correlation coefficient of 0 for two variables indicates no correlation, and the two vairables have no explanatory value for eachother.
What is a -1 value of coefficient
between two variables mean that if one variable chnages in value by a given percentage increment the other variable will change in the opposite direction by the same percentage increment.
This indicates perfect negative correlation.
a scatter diagram can be used to visually check for a correlation between two variables. How do you create a scatter diagram?
plot one variable against another in two dimensions, and view the resulting data distribution.`
Regression models and econometric models are types of casual models. What are regression models?
mathematical statement of a relationship between a dependent varibale and one or more idenpendent variables.
What are linear models?
types of regression models.
- describe a relationship that forms a straight line.
-
what are nonlinear regression models?
describe relationship that follow a curve.
Regression models apply a statisticall technique called regression analysis. What is involved in this?
projecting the value of one dependent variable based on its change over time in relation to changes in one or more independent variables.
When a regression model considers only one independent variable, the technique is known as what?
whats the algorithm?
simple regression analysis
y= Mx+b
When a regression model incorporates more than one independent variable, the technique is known as what?
whats the algorithm?
multiple regression analysis
y= a + a1(X1) + a2(x2)+…
Why would a life insurance company use regression models?
to develop a variety of projections, including projection of policy loan activity, disability claims activity and sales.
What are economic models?
they use a system of interdependent regression euqations to describe a particular segment of economic activity such as the flow of funds into and out of the financial service industry or the compoenents of a country’s gross domestic product.
True or False
Forecast should be monitored for accuracy?
True,
When the data comes up it should be comapred to predictive modelling.
What are the apporaches to measuring the sucess of a quantitative model in forecasting data points when matching them to real-world data outcomes?
- error
- percentage forecast error
- mean forecast error
- mean absolute deviation
How do you calculate error?
Actual value- forecast value = error
define a “cycle”
AKA series of cyclical variations, is a rhythmically fluctuating pattern that repeats over time with reasonable regularity.
- predictive or explanatory value
- the longer it is the more predictable it is.
When would cyclical variations occur?
as an economy changes from depression to recovery or from expansion to recession
What is a causal forecasting model
type of quantitative model that uses historical data and other relevant variables as a basis for describing unknown future data points
How do you find the forecast error as a percentage?
calculate the absolute value of the difference in one observation and its forecast. divide the absolute value by the actual result for that observation, multuply by 100
what is the mean forecast error (MFE_
measure that indicates the directional tendency and magnitide of erros produces using a given quantitative forcasting model.
True or False
A value for MFE that is greater than 0 indicates that the forecasts tend to understate the actual value.
False
>0 indicate forecast tend to overstate actual value
< 0 the forecaset tend to understate the actual valye,
what is the name for the number of observations the absolute values of the forecast errros?
absolute deviations
What is the mean absolute deviation (MAD)?
is a common measure that indicates the tendency int eh absolute amount of forecast errors but does not indicate the direction of the errors.
- used to measure error in time-series analysis