Sensitivity analysis
The Finance Storyteller
https://www.youtube.com/watch?v=o6-HCOG1Cp4

Transkript (automatisch erstellt)

0:00 The clearest real-world example of sensitivity analysis that I have ever seen is one by oil and gas company Shell about the impact of changes in the oil price.
0:15 The price sensitivity at Shell group level is $6 billion of cash flow from operations per annum per $10 per barrel Brent oil price movement.
0:26 If the oil price goes up by $10 per barrel, they expect $6 billion of incremental cash flow from operations.
0:33 If the oil price goes down by $10 per barrel, they expect a decrease of $6 billion of cash flow from operations.
0:41 This sensitivity statement comes with a disclaimer: this price sensitivity is appropriate for smaller price changes, and is best used for full-year numbers.
0:52 The “format” of this sensitivity analysis is: what is the effect of a change in absolute terms of an input variable (oil price) on the absolute amount of a target variable (cash
1:04 flow from operations). Commodity trading and mining company Glencore provides a sensitivity analysis in the part
1:12 of the annual report that discusses the review of assets for impairment. For each cash generating unit with limited headroom relative to their estimated recoverable
1:23 value, a sensitivity impact of potential 10% movements in the most sensitive assumptions is provided.
1:32 For Coal South Africa, a 10% fall in coal price assumptions would lead to a possible impairment of $703 million.
1:41 For Mopani, a fall of 10% in the copper price assumption would lead to a possible impairment of $181 million, while a 10% reduction in the estimated annual production over the life
1:54 of the mine could result in an impairment of $116 million. Similar sensitivity analyses are done for cash generating units involved in the extraction
2:05 and production of nickel, oil, and zinc. The “format” of this sensitivity analysis is: what is the effect of a change in percentage
2:15 terms of an input variable (10%) on the absolute amount of a target variable (millions of $ of impairment charges).
2:26 These examples lead us to a definition of sensitivity analysis: the process of estimating how target variables change in relation to changes in input variables.
2:38 What is the effect of a change in input variable x on target variable f(x)? What is the effect of a change in the oil price on cash flow from operations?
2:51 What is the effect of a change in revenue on profitability? What is the effect of a change in estimated project benefits on net present value?
3:03 The key to sensitivity analysis is to identify the most significant assumptions that affect an output: which input variables have the strongest impact on the target variables?
3:16 Prior to starting a sensitivity analysis, you first need to decide what the key performance indicators (target variables) are!
3:25 Here’s an example of sensitivity analysis. Let’s start off with the base case on the left.
3:31 The projected income statement of a company has $1 million of revenue, minus $600K of variable costs, which leads to $400K of contribution margin.
3:43 After deducting fixed costs of $200K, operating margin is $200K. What is the effect of a change in revenue (more specifically a drop of 10%) on operating margin?
3:57 Wrong question! Not every type of change in revenue has the same effect on operating margin.
4:03 Currently, the company expects to sell 100 thousand units of one single type of product at $10 each.
4:12 If we want to analyze the impact of a change in revenue (input variable) on operating margin (target variable), then we should either take a 10% decrease in volume, or a 10% decrease
4:24 in price, or even better: analyze the effect of each of these versus the base case side-by-side. A 10% decrease in volume leads to revenue of $900K: 90 thousand units sold at $10 each.
4:41 A 10% decrease in price also leads to revenue of $900K, but with a different composition: 100 thousand units sold at $9 each.
4:52 So far, so good, so what? That question is answered in the next line of the projected income statement.
5:02 Variable costs in the case of a 10% drop in volume are $540K: 90K units that cost $6 per unit to make.
5:13 In other words, 90K units in revenue, and 90K units in variable costs. If there is a drop in volume, then both revenue and variable costs drop by 10%.
5:26 Variable costs in the case of a 10% drop in price stay at $600K: 100K units at that cost $6 per unit to make.
5:38 Revenue drops, but variable costs stay the same. As a result, the contribution margin in the volume drop scenario is $360K (90K units times
5:50 $4 per unit), while in the price drop scenario it is only $300K (100K units times $3 per unit). Deduct the fixed costs, which stay at $200K, the same as in the base case scenario.
6:06 Then take a look at operating margin: $160K in the 10% volume drop scenario, but only $100K in the 10% price drop scenario.
6:20 From this, we learn that we can expect a 10% drop in volume to lead to a 20% drop in Operating Margin, while a 10% drop in price would lead to a 50% drop in Operating Margin!
6:32 Be much more afraid of a drop in price than of a drop in volume! What we have done so far is called a “one assumption at a time” scenario analysis:
6:45 analyzing the effect of varying one model input factor at a time while keeping all other fixed. In real life, many assumptions may be linked, and move at the same time,
6:56 in the same direction, or in opposite directions. What if both volume and price drop by 10% at the same time?
7:05 Our new revenue would be $810K: 90K units at $9 per unit. Variable costs $540K: 90K units at $6 per unit.
7:21 Contribution margin $270K: 90K units at $3 per unit. Fixed costs $200K. Operating margin $70K. What we learn from this, is that a 10% drop in volume AND a 10% drop in price at the same time
7:42 lead to an expected 65% drop in Operating Margin versus the base case! In a lot of situations, people performing a sensitivity analysis mistakenly assume that
7:58 input variables and output variables are linked in a linear way: a 20% change in the inputs is expected to have double the effect on the outputs versus a 10% change in the inputs.
8:11 A positive linear scenario on the left, a negative linear scenario on the right. The impacts of input variables on target variables could however be exponential, not linear!
8:26 In the case of a “positive” scenario, this would be called convexity: exponential growth in the target variable.
8:34 This is something you try to look for, and benefit from! In the case of a “negative” scenario, this would be called concavity:
8:43 exponential decline in the target variable. This is something you try to avoid, build defenses against, in order not to get hurt,
8:53 or “blow up”! When doing a sensitivity analysis, try to detect the sensitivity of an outcome to different
9:00 sized shocks. What does sensitivity analysis give us?
9:06 In the words of Nassim Taleb: performing sensitivity analysis on assumptions does not eliminate the risk, but identifies which assumptions are key to conclusions, and thus merit close scrutiny.