Math calculations to better utilize CPI data

Report prepared by Gerald Perrins, branch chief of Consumer Prices in the Mid-Atlantic region,
and Diane Nilsen, former regional clearance officer in the National Office of Field Operations,
Bureau of Labor Statistics.

The Consumer Price Index (CPI) is published points, because index point changes are
as an index number that shows the change in affected by the level of the index in relation to
the price of a defined market basket of goods its base period, while percent changes are not.
and services over time from a base period
which is defined as 100.0. An increase of 7 The following illustration shows a
percent from that base period, for example, is hypothetical CPI one-month change between
shown as 107.0. Alternately, that relationship April 2016 and May 2016 using the 1982-
can also be expressed as the price of a base 84=100 reference base.
period "market basket" of goods and services
rising from $100 to $107. Currently, the Reference Base
reference base for most CPI indexes is 1982- 1982-84=100
84=100 but some indexes have other May 2016 ........................................ 240.236
references bases. The reference base years April 2016 ....................................... 239.261
refer to the period in which the index is set to Index point change .............................. 0.975
100.0. In addition, expenditure weights are Divided by the earlier index ..... 0.975/239.261
updated every two years to keep the CPI Equals ............................................... 0.004075
current with changing consumer preferences. Multiplied by 100 ............................... 0.4075
Equals percent change ........................ 0.4
Index numbers are not dollar values, but
measures of the change over time relative to Over-the-year percent change
their base period value of 100.0 (for example, To arrive at a percent change over an entire
280.0 or 30.3). Index numbers also are year, the beginning and ending periods of a
commonly used to measure the size and CPI series must always be the same month,
direction of price movements between various such as May 2015 and May 2016. Note: A
time periods such as monthly, quarterly, semi- calculation using January and December data
annual, and annual percent changes. would result in an 11-month change, not a 12-
month/over-the-year change.
Effective with the January 2007 CPI, the
Bureau of Labor Statistics began to publish its The calculation below shows the over-the-
consumer price indexes rounded to three year change from May 2015 to May 2016 for
decimal places rather than one. As a result, both the 1982-84=100 and 1967=100
all percent changes in this document have reference bases. The percent change is
been calculated from three decimal place rounded:
indexes regardless of date. However, the
resulting percent changes will continue to be Reference Base
published to one decimal place. Note: using 1982-84=100 1967=100
three decimal place values to compute percent
changes eliminates nearly all rounding errors May 2015 237.805 712.357
in the resulting percent change that May 2016 240.236 719.641
occasionally occurred when using indexes Index point change 2.431 7.284
rounded to one decimal place. Additional Divided by the
information on this conversion can be found earlier index 2.431/237.805 7.284/712.357
at www.bls.gov/cpi/cpithreedec.htm. Equals 0.01022 0.01022
Multiplied by 100 1.022 1.022
What follows are mathematical concepts and Equals percent change 1.0 1.0
formulae that are useful in a variety of index
applications. There are two critical points to remember:
Percent change 1. Always use the same reference base period
Movements of an index from one month to for all calculations. If the first point uses the
another are usually expressed as percent 1982-84=100 base, the end point must also
changes rather than as changes in index use that base.
2. Calculating an over-the-year percent could buy, on average, in 2013. An automatic
change, such as May 2015 to May 2016, is not “CPI Inflation Calculator” is available online
equal to the sum of the over-the-month for annual comparisons of purchasing power
changes between those two time periods. at www.bls.gov/data/inflation_calculator.htm.

Annual averages Similarly, one can calculate equivalent dollar

amounts for any two months in different years
Annual averages are the sum of the 12 using a ratio of those monthly indexes. For
monthly data points (i.e. indexes), divided by example, to determine how much money one
12. They represent an average index for a would need in May 2016 to have the same
given year, not a particular month. An annual spending power as $500 in May 2015,
average change should not be confused with multiply the dollar amount by the ratio of the
the over the year percent change, such as the indexes for May 2016 and May 2015:
calculation of the May to May changes
mentioned above.
May 2016=240.236 x $500
A percent change from December 2014 to May 2015=237.805
December 2015 is unlikely to be the same as
the change in the annual average from 2014 to or, 1.01022 x $500 = $505.11
2015. Users should take care to examine the
data with which the CPI is being compared to This means that a basket of goods costing
determine whether the annual average or 12- $500 in May 2015 would cost $505.11 in May
month change is more appropriate for their 2016.
purposes.
Constant dollars
In addition, users should note that, for an All-
items CPI that is published every other For analysis involving long time periods, it is
month, the annual average is based on 12 frequently necessary to convert current or
months of data. Many food and energy prices nominal dollars into constant or real dollars.
are collected for the "off" months, and the This is done by multiplying each dollar
unpublished “off-cycle” indexes are amount by a ratio of price indexes, as shown
interpolated and used in the annual average. below.
Most All-item CPIs for metropolitan areas are
published every other month. Suppose one’s salary was $35,000 in 2005,
$40,000 in 2010, and $45,000 in 2015. The
Purchasing power All items CPI was 195.3 in 2005, 218.056 in
2010, and 237.017 in 2015. The conversion
The CPI can be used to show how the to constant 2005 dollars would be as follows:
purchasing power of a dollar changes over
time. 2005: (195.3/195.3) x $35,000 = $35,000
2010: (195.3/218.056) x $40,000 = $35,826
The purchasing power of a dollar in 2014 was 2015: (195.3/237.017) x $45,000 = $37,080
about 98.4 percent of the purchasing power of
a dollar in 2013. This can be calculated as To convert the same data to constant 2015
follows: dollars, use 2015 as the base:
232.957/236.736 X 100 = 98.4%
2005: (237.017/195.3) x $35,000 = $42,476
2010: (237.017/218.056) x $40,000 = $43,478
with 232.957 being the CPI annual average 2015: (237.017/237.017) x $45,000 = $45,000
index for 2013, and 236.736 being the 2014
annual average index. This means that the In this example, while one’s nominal salary rose
purchasing power of the dollar declined about from $35,000 to $45,000 from 2005 to 2015, an
1.6 percent between 2013 and 2014 because increase of 29 percent, the growth in real salary,
of inflation. Or stated another way, a dollar in after adjusting for inflation, was more modest,
2014 could only buy 98 percent of what it slightly less than 6 percent.

2
Conversion to other base periods either the National CPI Information Office or
the eight Regional Economic Analysis and
It is sometimes necessary to use index values Information (EA&I) Offices. Examples
on a reference base that is no longer include All items less food and energy, All
published. In these instances, the Bureau items less energy, and All items less food.
provides rebasing factors.
A situation that requires such action could be Occasionally, a user needs to estimate a price
a long-term contract requiring the use of change that is not published by BLS. For
1957-59=100 as the base period where the instance, suppose a user needs a CPI measure
parties cannot agree to a successor index. that includes everything except gasoline. This
This often occurs when index points rather can be done by estimating a special index, in
than percent change are the basis for this case, All items less gasoline. Just
escalation. In such a case, contact the BLS remember, this estimate will not be an official
for the specific rebasing factor needed for the CPI, though it can be used to estimate
computation. changes in prices for item categories not
published by the Bureau.
Rebasing factors are unique to their index
series and can not be substituted. Once When creating a special index, the relative
calculated, the rebasing factor to move a importance of published indexes is a key
specific index from a specific base year to component. To reflect the differences in the
another specific base year will not change. average amount of the various items
purchased, weights are assigned to each item.
To convert the December 2015 (1982- As the prices of the individual items change,
84=100) CPI-U All items index to the 1957- the relative weight of each item is reported in
59=100 base, use the rebasing factor of values known as relative importance. The
0.2870447. Divide the current index by the relative importance for each published index
rebasing factor to calculate the index on a component is available for December of each
1957-59 basis. year.
The first step in making a special index is to
Example: calculate an adjusted relative importance for
All items, CPI-U, December 2015, the component being removed from All items,
1982-84=100 236.525 in this case Gasoline. Multiply the published
Rebasing factor 0.2870447 December relative importance of the
Dec. 2015 / Rebasing factor………..819.4 component index by the index value for the
All items, CPI-U, December 2015, month of interest. Then divide that amount
Rebased 1957-59=100 819.4 by the component’s index for the same month
When new base years are introduced, BLS and year as the relative importance.
recalculates each index back to the beginning For example, using the December 2015
of that series to provide a consistent stream of relative importance data and the December
data. Using the official series will minimize 2015 and February 2016 CPI-U Gasoline
rounding differences occasionally caused by component indexes for the U.S. city average
the rebasing factors. To move from a (1982-84=100):
discontinued index to a current index, the
easiest method is to request a historical table Relative importance, Gasoline,
from any BLS office. December 2015 ....................................... 3.000
Index for Gasoline
December 2015 ...................... 179.496
February 2016 ........................ 154.564
Constructing special CPI indexes and their (Rel. imp.) 3.000 x (Feb. 2016) = 154.564
percent change (Dec. 2015) = 179.496
The Bureau calculates thousands of special or, 3.000 x 0.86110 = 2.5833
indexes that are available online or from

3
Do the same calculation using the same consumer expenditures in the intervals
periods for the other index, in this case All between weight revisions, because
items. The All items adjusted relative consumption patterns are influenced by
importance result is 100.2478, where the factors other than price change. These factors
original relative importance was 100 and the include income, variations in climate, family
index rose from 236.525 (December 2015) to size, and the availability of new and different
237.111 (February 2016). kinds of goods and services.
Next, calculate the base relative importance Discontinued indexes and publication
of the All items less gasoline index. Do this changes
by subtracting the original relative importance
of gasoline from the original relative There will be occasions where, especially for
importance of the All items index. longer-term contracts, a specified index may
not be precisely available for the duration of
December 2015: the contract. An example of such an
Relative importance, All items, occurrence is the index for Washington, D.C.-
December 2015 ........................ 100.0 Md.-Va., which is now part of the larger
Less relative importance, Washington-Baltimore, D.C.-Md.-Va.-W.Va.,
Gasoline, December 2015 .......... - 3.0000 metropolitan area. It is not possible to
Relative importance, All items directly calculate a fifteen-year change
less gasoline, December 2015 .... 97.000 starting in 1995 and ending in 2010 since
Then do the same with the adjusted relative publication of Washington data only stopped
importances for the second month: in 1997. In its place, data for the
Washington-Baltimore consolidated area has
February 2016: been published since 1998.
Relative importance, All items,
February 2016 .......................... 100.2478 In a case where both parties have agreed upon
Less relative importance, an index to replace the discontinued (original)
Gasoline, February 2016 ............ - 2.5833 one, the calculation can be performed in two
steps. First, for the series of interest, calculate
Relative importance, All items the percent change from the beginning period
less gasoline, February 2016 ...... 97.6645 month and year until an agreed upon month
Finally, calculate the percent change between and year where data exist for both the
the results for the two months: discontinued and the successor (new) series.
Adjust the dollar value in the contract based
(Rel. imp. Feb. 2016) = 97.6645 x 100 - 100 = upon this percent change. The dollar value
(Rel. imp. Dec. 2015) = 97.0000 should be the amount estimated to start on, or
1.00685 x 100 - 100 = 0.685 = 0.7% closest to, the month and year chosen for the
beginning index.
In short, one could estimate that prices for All
items less gasoline rose 0.7 percent from Second, choose the end-period month and
December 2015 to February 2016. As before, this year for the new successor series. The month
is an estimate and does not represent an official and year you defined as the previous step’s
BLS index. end period will now become the first period
month and year in this step. Calculate the
In short, the relative importance of a percent change from this beginning period to
component is its expenditure or value weight, the selected end period of the new series. Use
expressed as a percentage of All items within the calculated percent change to adjust the
an area. These ratios represent an estimate of contract dollar value from the first step. This
how consumers would distribute their is the contract’s new dollar amount adjusted
expenditures over time. for the overlap between the discontinued and
Relative importance ratios cannot, however, successor indexes.
be used as estimates of current spending
patterns or as indicators of changing

4
In the following example, Index A is now before and immediately after the missing
discontinued and the parties have agreed to month.
use Index B as the successor index.
For example, to approximate the April 2016
Value of contract in January 1995 ..........$9,000.00 All items (CPI-U) index for the Washington-
Index A Baltimore, D.C.-Md.-Va.-W.Va. metropolitan
January 1995 .....................................158.7 area, multiply the March index by the May
January 1998 .....................................178.9 index and take the square root of the product:
Percent change ................................... 12.7 May 2016 .................. 157.770
9,000 x (1+ 0.127)......................10,143.00 March 2016 ............... 156.493
Value of contract in January 1998 Product ................. 24,689.901
after adjustment of 1995 amount .......$10,143.00 Square root ................ 157.130
Index B
Note: data calculated in this way cannot be
January 1998 .....................................104.7 interpreted as official CPI indexes, as the
January 2010 .................................216.687 calculation is based on two data points and
Percent change ..................................107.0 not on CPI’s aggregation method.
10,143 x (1+1.07).......................20,996.01
Value of contract in January 2010 Furthermore, if bimonthly CPI data are
after adjustment of 1998 amount .......$20,996.01 volatile, then less confidence should be placed
Approximating missing data points in estimates for the missing months. Percent
changes based on approximated data should
Many indexes are published every other also be considered as unofficial estimates.
month. BLS publishes CPI information for Examples of volatile series would be apparel,
All items and for major components (Food household furnishings and operations, and
and beverages, Housing, Apparel, fuels and utilities.
Transportation, Medical care, Recreation,
Education and communication, and Other Cost of living comparisons not appropriate
goods and services) for 11 metropolitan areas
on a bimonthly basis. Some of these Each CPI index represents the change in
metropolitan areas, as defined by the U.S. inflation for a good or service in the market
Census Bureau, include suburbs or counties basket over time for various areas and the
that extend across state boundaries. These 11 nation. However, CPI indexes for one
bimonthly metropolitan areas are available on metropolitan area can not be compared with
the following schedule: Boston-Brockton- those in other areas to determine where it is
Nashua, Mass.-N.H.-Maine-Conn.; more or less expensive to live. It is still
Cleveland-Akron, Ohio; Dallas-Fort Worth, possible to calculate the percentage change,
Texas; and Washington-Baltimore, D.C.-Md.- that is, how fast or slow prices are rising in a
Va.-W.Va.; in odd months (January, March, given metropolitan area, but that can not be
etc.) and Atlanta, Ga.; Detroit-Ann Arbor- equated to higher or lower actual costs, as the
Flint, Mich.; Houston-Galveston-Brazoria, next example illustrates.
Texas; Miami-Fort Lauderdale, Fla.; Bread was priced in City A and City B.
Philadelphia-Wilmington-Atlantic City, Pa.-
N.J.-Del.-Md.; San Francisco-Oakland-San City A City B
Jose, Calif.; and Seattle-Tacoma-Bremerton, Price of bread in:
Wash.; in even months (February, April, etc.). September 2014 $1.00 $0.50
September 2015 1.50 1.00
In a few instances, this is a change from prior
years and a contract might call for a specific Index values (Sep. 2014=100):
month which is no longer available. If both September 2014 100.0 100.0
parties agree, the missing data point can be September 2015 150.0 200.0
approximated using the month immediately Index point change 50.0 100.0

5
 Divided by the since the index values for the two groups will
earlier index 50.0/100.0 100.0/100.0 not be the same.
Equals 0.50 1.0
Multiplied by 100 Contracting parties are free to determine what
indexes they wish to use, but it is important to
equals percent change 50.0 100.0 note that while the CPI is used for a variety of
purposes, it measures only the change in the
Note: the price in City B rose at a faster pace price of a representative market basket of
than in City A, but the actual dollar price of consumer goods and services. At the All
bread in City A was still 50 cents above City items and individual component levels, only
B. items that would be purchased by consumers
are eligible for pricing. Thus, an item like
Which index should you use? commercial rent would not be included in the
The CPI program publishes data for the U.S., "rent of primary residence" category.
geographic regions, and many metropolitan
areas. For each of these, data are available for Although BLS cannot make a decision as to
"All items" as well as for a variety of the which index a user should reference, it has
components that are aggregated together to published a reference guide, "Using the
become the "All items" index. Consumer Price Index for Escalation,"
available at www.bls.gov/cpi/cpi1998d.htm.
Each of these many series are published for This publication and other information such
two population groups: All Urban Consumers as frequently asked questions, historical data,
(CPI-U) and Urban Wage Earners and and rebasing factors are available through the
Clerical Workers (CPI-W). When using the national office’s and eight regional offices’
CPI for calculating price change, you must Web sites. (See table A.)
specify one population group or the other
Table A. Regional Economic Analysis and Information Offices and the National Consumer
Price Index Information Office in Washington, D.C.

Information Office Web Address Phone Email

Boston www.bls.gov/regions/new-england (617) 565-2327 [email protected]

New York www.bls.gov/regions/new-york-new-jersey (646) 264-3600 [email protected]
Philadelphia www.bls.gov/regions/mid-atlantic (215) 597-3282 [email protected]
Atlanta www.bls.gov/regions/southeast (404) 893-4222 [email protected]
Chicago www.bls.gov/regions/midwest (312) 353-1880 [email protected]
Dallas www.bls.gov/regions/southwest (972) 850-4800 [email protected]
Kansas City www.bls.gov/regions/mountain-plains (816) 285-7000 [email protected]
San Francisco www.bls.gov/regions/west (415) 625-2270 [email protected]
Washington, D.C. www.bls.gov/cpi (202) 691-7000 [email protected]