Historical inflation - how different inflation models can affect safe withdrawal rates

OldScientist

Modelling historical outcomes of retirement spending requires knowledge of the historical returns on various assets (stocks, bonds, etc.) and also inflation rates. Of the current official measures of inflation adopted in the UK, RPI was started in 1948 and CPI in 1988 (see https://www.ons.gov.uk/economy/inflationandpriceindices). There are a number of sources of inflation values before 1948 including

1. Office of National Statistics (ONS), Retail Prices Index (long run series) (see https://www.ons.gov.uk/economy/inflationandpriceindices/timeseries/cdko/mm23)
2. Measuring Worth (MW), Samuel H. Williamson "Annual Inflation Rates in the United States, 1775 - 2020, and the United Kingdom, 1265 - 2020," MeasuringWorth, 2021. (see https://www.measuringworth.com/calculators/inflation/index.php#)
3. Barclays Equity Gilts Study (BEGS), Figure 1 in Timpson, Barclays Equity Gilt Study, 2016, Courtiers Research Note (see https://www.courtiers.co.uk/wp-content/uploads/2020/09/barclays-equity-gilt-study.pdf)
4. Hills, S, Thomas, R and Dimsdale, N.(MMD) "A millennium of macroeconomic data – version 3.1", Bank of England. Series: Headline consumer price index (cpi). (see https://www.bankofengland.co.uk/statistics/research-datasets)

Given that the inflation rates are, at least prior to 1948, derived using different sources or making different assumptions they can end up being quite different. Example values are given in the following table where the inflation rate (in percent) from 1937-1946 is shown

            ONS      MW     BEGS     MMD
1937      3.8        5.3        5.8         3.4
1938      1.2        1.1       -2.4        1.6
1939      3.0        3.0       10.6        2.8
1940    16.8      13.5       12.3      16.8
1941    10.9      10.3         3.4      10.8
1942      7.1        6.3        -0.6        7.1
1943      3.3        3.6        -0.6        3.4
1944      2.8        2.1         0.8        2.7
1945      2.7        2.1         0.8        2.8
1946      3.1        4.0         0.7        3.1

Two of the series (ONS and MMD) are broadly similar with differences of no more than 40 basis points, a third series (MW) is similar in form to the first two but the differences are larger (up to 3 percentage points). However, the fourth series (BEGS) is significantly different with relatively low inflation (and some deflation) during the latter half of the period.

So how much effect does this have on safe withdrawal rates? The Safemax (i.e. the highest inflation adjusted withdrawal that does not result in portfolio exhaustion within 30 years), P1, and P10 (the highest inflation-adjusted withdrawal that results in no more than 1% and 10% of historical retirements running out of money in 30 years, respectively) for a portfolio consisting of 60% UK stocks and 40% UK bonds with annual withdrawals and rebalancing are given in the following table (the returns for stocks and bonds from 1872-2015 are drawn from the Jordà-Schularick-Taylor Macrohistory Database, Revision 5, at https://www.macrohistory.net/database/)

            Safemax    Year       P1       P10
ONS       2.93        1937     2.95      3.29
MW        3.07        1937     3.07      3.37
BEGS     3.02        1906     3.12      3.53
MMD      2.88        1937     2.93      3.24

The outcomes exhibit a 20 to 30 basis point variation in the withdrawal rates depending on the inflation series used, pointing to a first approximation of the precision in safe withdrawal rates (and demonstrating that quoting safe withdrawal rates to two decimal places, as in the above table, is unlikely to be justified!). It is also interesting that use of the BEGS inflation values changes the worst retirement case from 1937 to 1906.

I have a couple of questions that I hope the knowledgeable members of the MSE boards can help me with:

1) Does anyone know where the Barclays (BEGS) inflation series originates from (I have been unable to find a reference to the source of these values which are significantly different to the other series)?

2) Does anyone know the source of the UK inflation values incorporated in the Dimson, Marsh and Staunton dataset (https://www.london.edu/-/media/files/faculty and research/subject areas/global investments yearbook.pdf)? I note that there is a reference to the Measuring Worth series in the 2012 version of the Credit Suisse Investment Returns Yearbook, but I’m not sure whether these are then used in the dataset.

michaels

OldScientist said:

Modelling historical outcomes of retirement spending requires knowledge of the historical returns on various assets (stocks, bonds, etc.) and also inflation rates. Of the current official measures of inflation adopted in the UK, RPI was started in 1948 and CPI in 1988 (see https://www.ons.gov.uk/economy/inflationandpriceindices). There are a number of sources of inflation values before 1948 including

1. Office of National Statistics (ONS), Retail Prices Index (long run series) (see https://www.ons.gov.uk/economy/inflationandpriceindices/timeseries/cdko/mm23)
2. Measuring Worth (MW), Samuel H. Williamson "Annual Inflation Rates in the United States, 1775 - 2020, and the United Kingdom, 1265 - 2020," MeasuringWorth, 2021. (see https://www.measuringworth.com/calculators/inflation/index.php#)
3. Barclays Equity Gilts Study (BEGS), Figure 1 in Timpson, Barclays Equity Gilt Study, 2016, Courtiers Research Note (see https://www.courtiers.co.uk/wp-content/uploads/2020/09/barclays-equity-gilt-study.pdf)
4. Hills, S, Thomas, R and Dimsdale, N.(MMD) "A millennium of macroeconomic data – version 3.1", Bank of England. Series: Headline consumer price index (cpi). (see https://www.bankofengland.co.uk/statistics/research-datasets)

Given that the inflation rates are, at least prior to 1948, derived using different sources or making different assumptions they can end up being quite different. Example values are given in the following table where the inflation rate (in percent) from 1937-1946 is shown

            ONS      MW     BEGS     MMD
1937      3.8        5.3        5.8         3.4
1938      1.2        1.1       -2.4        1.6
1939      3.0        3.0       10.6        2.8
1940    16.8      13.5       12.3      16.8
1941    10.9      10.3         3.4      10.8
1942      7.1        6.3        -0.6        7.1
1943      3.3        3.6        -0.6        3.4
1944      2.8        2.1         0.8        2.7
1945      2.7        2.1         0.8        2.8
1946      3.1        4.0         0.7        3.1

Two of the series (ONS and MMD) are broadly similar with differences of no more than 40 basis points, a third series (MW) is similar in form to the first two but the differences are larger (up to 3 percentage points). However, the fourth series (BEGS) is significantly different with relatively low inflation (and some deflation) during the latter half of the period.

So how much effect does this have on safe withdrawal rates? The Safemax (i.e. the highest inflation adjusted withdrawal that does not result in portfolio exhaustion within 30 years), P1, and P10 (the highest inflation-adjusted withdrawal that results in no more than 1% and 10% of historical retirements running out of money in 30 years, respectively) for a portfolio consisting of 60% UK stocks and 40% UK bonds with annual withdrawals and rebalancing are given in the following table (the returns for stocks and bonds from 1872-2015 are drawn from the Jordà-Schularick-Taylor Macrohistory Database, Revision 5, at https://www.macrohistory.net/database/)

            Safemax    Year       P1       P10
ONS       2.93        1937     2.95      3.29
MW        3.07        1937     3.07      3.37
BEGS     3.02        1906     3.12      3.53
MMD      2.88        1937     2.93      3.24

The outcomes exhibit a 20 to 30 basis point variation in the withdrawal rates depending on the inflation series used, pointing to a first approximation of the precision in safe withdrawal rates (and demonstrating that quoting safe withdrawal rates to two decimal places, as in the above table, is unlikely to be justified!). It is also interesting that use of the BEGS inflation values changes the worst retirement case from 1937 to 1906.

I have a couple of questions that I hope the knowledgeable members of the MSE boards can help me with:

1) Does anyone know where the Barclays (BEGS) inflation series originates from (I have been unable to find a reference to the source of these values which are significantly different to the other series)?

2) Does anyone know the source of the UK inflation values incorporated in the Dimson, Marsh and Staunton dataset (https://www.london.edu/-/media/files/faculty and research/subject areas/global investments yearbook.pdf)? I note that there is a reference to the Measuring Worth series in the 2012 version of the Credit Suisse Investment Returns Yearbook, but I’m not sure whether these are then used in the dataset.

Useful.

Of course we would ideally want a price index that reflects costs for the typical retiree who no doubt buys a different bundle of goods from the index bundle - just one example - 'technology prices' have driven lower cpi inflation over the last few decades but perhaps make up a smaller proportion of the typical retiree basket than the country as a whole.

SMcGill

THIS
|
|
|
|
|
is how far above my head that all is 

Dead_keen

I had a quick flick through the Barclays report and can see that in Figure 6 it gives the index and annual inflation (plus five year average).  The numbers I saw seems a bit different to the ones you set out above but I didn't spend any time on that.  Their index is based on 1899=100 and I'm not aware of any tables that start from then.  Their's no narrative in the one I looked at that says how it was derived.  Maybe they created it based on something else in the earlier years - like the price of bread :-) 

You could always ask Barclays and see what they say.

billy2shots

Sorry, you lost me at 60% UK stocks and 40% UK bonds. 

Anyone wanting a 30 year safe withdrawal  rate based on that needs to revisit the drawing board (imho). 

OldScientist

michaels said:

OldScientist said:

Modelling historical outcomes of retirement spending requires knowledge of the historical returns on various assets (stocks, bonds, etc.) and also inflation rates. Of the current official measures of inflation adopted in the UK, RPI was started in 1948 and CPI in 1988 (see https://www.ons.gov.uk/economy/inflationandpriceindices). There are a number of sources of inflation values before 1948 including

1. Office of National Statistics (ONS), Retail Prices Index (long run series) (see https://www.ons.gov.uk/economy/inflationandpriceindices/timeseries/cdko/mm23)
2. Measuring Worth (MW), Samuel H. Williamson "Annual Inflation Rates in the United States, 1775 - 2020, and the United Kingdom, 1265 - 2020," MeasuringWorth, 2021. (see https://www.measuringworth.com/calculators/inflation/index.php#)
3. Barclays Equity Gilts Study (BEGS), Figure 1 in Timpson, Barclays Equity Gilt Study, 2016, Courtiers Research Note (see https://www.courtiers.co.uk/wp-content/uploads/2020/09/barclays-equity-gilt-study.pdf)
4. Hills, S, Thomas, R and Dimsdale, N.(MMD) "A millennium of macroeconomic data – version 3.1", Bank of England. Series: Headline consumer price index (cpi). (see https://www.bankofengland.co.uk/statistics/research-datasets)

Given that the inflation rates are, at least prior to 1948, derived using different sources or making different assumptions they can end up being quite different. Example values are given in the following table where the inflation rate (in percent) from 1937-1946 is shown

            ONS      MW     BEGS     MMD
1937      3.8        5.3        5.8         3.4
1938      1.2        1.1       -2.4        1.6
1939      3.0        3.0       10.6        2.8
1940    16.8      13.5       12.3      16.8
1941    10.9      10.3         3.4      10.8
1942      7.1        6.3        -0.6        7.1
1943      3.3        3.6        -0.6        3.4
1944      2.8        2.1         0.8        2.7
1945      2.7        2.1         0.8        2.8
1946      3.1        4.0         0.7        3.1

Two of the series (ONS and MMD) are broadly similar with differences of no more than 40 basis points, a third series (MW) is similar in form to the first two but the differences are larger (up to 3 percentage points). However, the fourth series (BEGS) is significantly different with relatively low inflation (and some deflation) during the latter half of the period.

So how much effect does this have on safe withdrawal rates? The Safemax (i.e. the highest inflation adjusted withdrawal that does not result in portfolio exhaustion within 30 years), P1, and P10 (the highest inflation-adjusted withdrawal that results in no more than 1% and 10% of historical retirements running out of money in 30 years, respectively) for a portfolio consisting of 60% UK stocks and 40% UK bonds with annual withdrawals and rebalancing are given in the following table (the returns for stocks and bonds from 1872-2015 are drawn from the Jordà-Schularick-Taylor Macrohistory Database, Revision 5, at https://www.macrohistory.net/database/)

            Safemax    Year       P1       P10
ONS       2.93        1937     2.95      3.29
MW        3.07        1937     3.07      3.37
BEGS     3.02        1906     3.12      3.53
MMD      2.88        1937     2.93      3.24

The outcomes exhibit a 20 to 30 basis point variation in the withdrawal rates depending on the inflation series used, pointing to a first approximation of the precision in safe withdrawal rates (and demonstrating that quoting safe withdrawal rates to two decimal places, as in the above table, is unlikely to be justified!). It is also interesting that use of the BEGS inflation values changes the worst retirement case from 1937 to 1906.

I have a couple of questions that I hope the knowledgeable members of the MSE boards can help me with:

1) Does anyone know where the Barclays (BEGS) inflation series originates from (I have been unable to find a reference to the source of these values which are significantly different to the other series)?

2) Does anyone know the source of the UK inflation values incorporated in the Dimson, Marsh and Staunton dataset (https://www.london.edu/-/media/files/faculty and research/subject areas/global investments yearbook.pdf)? I note that there is a reference to the Measuring Worth series in the 2012 version of the Credit Suisse Investment Returns Yearbook, but I’m not sure whether these are then used in the dataset.

Useful.

Of course we would ideally want a price index that reflects costs for the typical retiree who no doubt buys a different bundle of goods from the index bundle - just one example - 'technology prices' have driven lower cpi inflation over the last few decades but perhaps make up a smaller proportion of the typical retiree basket than the country as a whole.

Completely agree - even within the broad category of 'retiree', inflation is likely to be very different between individuals. I do note that the US does produce an inflation index for retirees (see https://www.bls.gov/cpi/research-series/r-cpi-e-home.htm), so maybe the ONS will oblige at some point in the future.

OldScientist

Dead_keen said:

I had a quick flick through the Barclays report and can see that in Figure 6 it gives the index and annual inflation (plus five year average).  The numbers I saw seems a bit different to the ones you set out above but I didn't spend any time on that.  Their index is based on 1899=100 and I'm not aware of any tables that start from then.  Their's no narrative in the one I looked at that says how it was derived.  Maybe they created it based on something else in the earlier years - like the price of bread :-) 

You could always ask Barclays and see what they say.

Cheers for looking for me - I may have to contact Barclays in the end (the numbers I gave above were based on the figure in the Timpson report - I assumed these were correctly plotted).

There appear to have been a number of approaches in calculating inflation rates (some including all sorts of goods - one paper I looked at had table after table of prices - I'll admit to only having a casual interest in this, so didn't read it in detail).

OldScientist

billy2shots said:

Sorry, you lost me at 60% UK stocks and 40% UK bonds. 

Anyone wanting a 30 year safe withdrawal  rate based on that needs to revisit the drawing board (imho). 

Since I was only interested in the effect of different inflation models (and not, for example, exchange rates - which at least are consistently documented), for simplicity I just used a domestic portfolio. I'd agree that holding a portfolio with stocks drawn solely from UK would lack diversification and be rather unwise (although historically, this would have been the norm for many UK retail investors).

Dead_keen

OldScientist said:

Dead_keen said:

I had a quick flick through the Barclays report and can see that in Figure 6 it gives the index and annual inflation (plus five year average).  The numbers I saw seems a bit different to the ones you set out above but I didn't spend any time on that.  Their index is based on 1899=100 and I'm not aware of any tables that start from then.  Their's no narrative in the one I looked at that says how it was derived.  Maybe they created it based on something else in the earlier years - like the price of bread :-) 

You could always ask Barclays and see what they say.

Cheers for looking for me - I may have to contact Barclays in the end (the numbers I gave above were based on the figure in the Timpson report - I assumed these were correctly plotted).

There appear to have been a number of approaches in calculating inflation rates (some including all sorts of goods - one paper I looked at had table after table of prices - I'll admit to only having a casual interest in this, so didn't read it in detail).

Here's the one I googled - page 81 - http://topforeignstocks.com/wp-content/uploads/2012/10/Barclays-Gilt-Study-2012-Full-Report.pdf

OldScientist

Dead_keen said:

OldScientist said:

Dead_keen said:

I had a quick flick through the Barclays report and can see that in Figure 6 it gives the index and annual inflation (plus five year average).  The numbers I saw seems a bit different to the ones you set out above but I didn't spend any time on that.  Their index is based on 1899=100 and I'm not aware of any tables that start from then.  Their's no narrative in the one I looked at that says how it was derived.  Maybe they created it based on something else in the earlier years - like the price of bread :-) 

You could always ask Barclays and see what they say.

Cheers for looking for me - I may have to contact Barclays in the end (the numbers I gave above were based on the figure in the Timpson report - I assumed these were correctly plotted).

There appear to have been a number of approaches in calculating inflation rates (some including all sorts of goods - one paper I looked at had table after table of prices - I'll admit to only having a casual interest in this, so didn't read it in detail).

Here's the one I googled - page 81 - http://topforeignstocks.com/wp-content/uploads/2012/10/Barclays-Gilt-Study-2012-Full-Report.pdf

Many thanks again - you are obviously more patient with google than I am! For me, that document didn't turn up on a general search until about the 10th page of search results...

Thankfully, the numbers on page 81 are similar enough to the ones I extracted from the graph in the Timpson report so as to be unlikely to change the outcomes by too much (and are still very different to those in the other sources)...

Grumpy_chap

OldScientist said:

So how much effect does this have on safe withdrawal rates? The Safemax (i.e. the highest inflation adjusted withdrawal that does not result in portfolio exhaustion within 30 years), 

Top marks for taking a detailed and data-based approach, however I can't help think you don't need any of that historic data to work out the safe withdrawal rate for a new retiree in 2022.

You need a crystal ball.