chapter 2 risk and decision making Flashcards
sensitivity analysis
Ads and disads
(NPV of project/PV cash flows subject to uncertainty) x 100
+ve:
-presented to mgmt in a form which facilitates subjective judgement
- identifies areas critical to success of project (which need to be carefully monitored)
- straightforward
- ve:
- can only analyse one factor at a time
- assumes changes ti variables can be made independently
- only looks at HOW FAR a variable can change not the PROBABILITY
- provides info to make a decision, does not point to the correct decision
predictive analytics
Uses historical and current data to create predictions about the future
linear regression - quantify relationship between dependant and independent variable so forecasts can be made
decision trees- identify the impact of different decisions and variables on the outcome of an investment
simulation- allows the effect of more than one variable changing at the same time to be assessed
Linear regression +/-
+ve:
- Simple to use + easy to explain
- used to predict impact of expanding variables beyond current estimates
- Ve:
- not always a linear relationship between variables and outcomes
- complex models needed to consider multiple variables
- Fake relationships between variables and outcomes may be identified
- data collected may be inaccurate or may be a large error variable
decision trees +/-
+:
- simple trees are easy to explain and logical to use
- can analyse different outcomes based on a number of variables
- :
- Variables have to be simplified and restricted to avoid overcomplicating tree
- large trees can be difficult to interpret
simulation
monte carlo - technique based on use of random numbers and probability statistics to investigate problems
example: setting up restaurant
need profit to attract finance
establish parameters - no. customers served(covers), spend per cover on food and drink, price to be charged per cover, cost of staff, opening closing times, required capital to fund restaurant
each then fed into software package, generated multiple alternatives for each and calculated profits, alternatives then plotted on a graph to give a spread of the profits that can be earned.
simulation presents results on standard deviation (riskiness) doesn’t tell which is better, that depends on attitude to risk.
simulation +/-
+:
- gives more info on possible outcomes and their relative probabilities
- useful for problems which cannot be solved analytically
-:
- not a technique for decision making-only provides info
- can be expensive (designing and running simulation)
- Monte C. techniques require assumptions to be ,ade about profitability distributions and relationships between variables- may be inaccurate
Prescriptive analytics what is it + pos and neg
combining statistical tools used in predictive analytics and combining it with AI and algorithms. PA can be used to calculate the optimum outcome from a variety of business decisions, e.g:
- capital rationing decisions
- replacement analysis
- optimal financing balance (debt v equity)
+:
can identify optimum decisions whilst incorporating multiple variables
- :
- creating models is complex and requires specialist data science skills
- reliability of models depends on reliability of the data and relationships between the past and future
Data bias types 7 types
selection bias- when data not selected randomly, sample is not representative of the population. all items should have an equal chance of being chosen.
self selection bias- when individuals select themselves to be part of the sample. only get results of people who are interested not those that aren’t
observer bias- when observing and recording results relates to interpretation. researcher allows their assumptions to influence their observations
omitted variable bias- key variables not included in the data to be analysed.
confirmation bias-when people see data that confirms their beliefs and ignore data that disagrees with their beliefs.
cognitive bias- human interpretation based on how data is presented, or the context it is presented in and ‘anchoring’
survivorship bias-where the sample contains items that survived some previous event. e.g. 98% of people who passed the mock passed the exam, what about the ones who failed the mock?
statistical tools
=AVERAGE to calculate mean
standard deviation shows the variability in the data set- higher the SD the greater the data spread and the greater the risk and return. =STDEV
coefficient of variation - ratio of the STDEV to mean calculated as (STDEV/mean)*100. the higher the % the wider the dispersion of data around the mean.
Expected values and its +/-
an average of possible outcomes, weighted by the probability of each outcome occurring
+:
- information is reduced to a single number for each choice
- idea of an average is readily understood
- :
- probabilities of diff possible outcomes may be difficult to estimate, possible to use; objective probabilities based on past events/similar projects or subjective probabilities (eg market research where project is different)
- average may not correspond to any of the possible outcomes
- unless same decision has to be made many times, average will not be achieved. it is not a valid way to make a decision in ‘one off’ situations(unless firm has a number of independent projects and there is a portfolio effect)
- The average gives no indication of the spread of possible results (IGNORES RISK)
