L A N D Y S Á N C H E Z
E L C O L E G I O D E M É X I C O
CHANGES IN MEXICO'S WEALTH
AND HOUSEHOLD ENERGY
MEXICO HISTORICAL TRENDS
•
Social and demographic
•
Lower fertility rates
•
Higher urbanization
•
Higher education
•
Changing migration patterns
•
Family formation
-10 -8 -6 -4 -2 0 2 4 6 8 10 1961 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011
GDP per capita growth (annual %)
ECONOMIC DEVELOPMENT AND
INCOME DISTRIBUTION
0 10 20 30 40 50 60 70 80 90 100 0 1000 2000 3000 4000 5000 6000 7000 8000 9000955
960
965
970
975
980
985
990
995
2000
2005
2010
US
d
ol
la
rs
(
co
st
an
t 2005)
GDP per capita and GINI index, Mexico
GDP per
capita
(constant
2005 US$)
GINI index
G
INI
HOW DO CHANGES IN ECONOMIC WEALTH IMPACT
HOUSEHOLD ENERGY CONSUMPTION?
•
Approaches with household data
•
Comparative aggregated data across countries and over
time
•
Response to price changes
•
My approach
•
Age-period-cohort
•
Time dimension
•
Nested/context
AGE-PERIOD-COHORT APPROACH
Alwin and McCammon 2003
Birth Cohort: shared
formative experience
Period: historical
times, impacts all
individuals, across
cohorts
HOW DO CHANGES IN ECONOMIC WEALTH
IMPACT HOUSEHOLD ENERGY CONSUMPTION?
•
Age: life course, changing household
characteristics
•
Period: changes on country’s economic wealth and
income distribution
•
Birth cohort: shaping practices and “reading the
economic context”
DATA
•
Eight Mexican Income and Expenditure Household
Survey Data 1992-2008
•
Only urban households: 97,161
•
Birth Cohorts: 5 years periods
•
Good for building up cohorts, bur relative short period
•
Direct Energy consumption
•
Electricity, gas, oil, coal, firewood at home
•
Energy consumption: expenditure >>> fuels Price, net
caloric content
MODEL
•
Age-Period-Cohort: Cross-classified Hierarchical
APCM
(Yang and Land 2006, Yang 2008, Yang and Land 2013)
•
Solution to the Identification Problem
•
Single year of age
•
Time periods correspond to years in which the
surveys are conducted
•
Cohorts can be defined either by five intervals that
are conventional in demography (also substantive
classification e.g., War babies, Baby Boomers, Baby
Busters, etc.).
YEAR Birth cohort 1992 1996 1998 2000 2002 2004 2006 2008Total Before 1913 121 92 56 44 38 32 19 24 426 1913-1917 111 117 59 65 76 72 49 45 594 1918-1922 206 230 145 100 186 196 146 178 1,387 1923-1927 269 296 191 188 307 347 270 368 2,236 1928-1932 339 440 293 296 453 630 477 627 3,555 1933-1937 372 559 388 366 581 797 576 898 4,537 1938-1942 547 672 477 451 764 1,038 835 1,152 5,936 1943-1947 603 845 557 533 949 1,252 996 1,494 7,229 1948-1952 729 992 794 655 1,232 1,645 1,307 1,902 9,256 1953-1957 883 1,208 846 788 1,443 1,800 1,493 2,300 10,761 1958-1962 945 1,301 1,010 926 1,654 2,162 1,767 2,732 12,497 1963-1968 878 1,510 1,139 1,085 2,004 2,724 2,389 3,443 15,172 1969-1973 315 829 713 710 1,388 2,108 1,865 2,817 10,745 1974-1978 17 255 322 472 929 1,522 1,556 2,281 7,354 1979-1983 0 12 36 121 363 776 984 1,604 3,896 1984-1988 0 0 1 2 38 211 407 745 1,404 1989-1993 0 0 0 0 0 6 24 105 135 Total 6,335 9,358 7,027 6,802 12,405 17,318 15,160 22,715 97,120
Data Structure
Households are nested
Simultaneouslyin two social contexts: Time periods and
cohorts of their households, in a cross-classified structure
ADVANTAGES AND LIMITATIONS
•
Advantages of CHAPC
•
Micro-level data, instead of aggregate population
estimates
•
Different definitions of times for age, period and cohorts
•
Considers hierarchical nature of the data
•
Allows the specification of fix and random effects
•
Manage unbalance data
•
Limitations:
•
Synthetic cohort, not “true” birth cohort (only with
longitudinal data)
•
But synthetic cohort: national representation of all ages
(potential long time span)
MODEL
Variables in the model:
Dependent variable: log per capita
energy consumption
Household Explanatory variables:
•
Income
•
Per capita expenditure (log)
•
Sex of Household Head (HH)
•
Education HH
•
Household Size
•
Family structure
•
Dwelling Size (number of rooms)
•
Geographical Regions
Period Explanatory variables:
•
Inequality distribution
* Also tested, occupation HH, activities outside
household, plus GDP as a period variable
RESULST HAPC MIXED
EFFECTS MODELS
AGE-PERIOD-COHORT EFFECTS
M0: unconditional model
Significant Period and cohort effects, though small
HOUSEHOLD PREDICTORS
0 500 1000 1500 2000 2500 3000 3500 100 4000 8000 12000 16000 20000 24000 28000 32000 36000 40000 44000 50000 60000 72000Income effect
Energy consumption Age positive Expenditure levels positive Household Size positiveRooms numbers non linear - low numbers, + large numbers Education Increases, thouhg less so at higher education Female headed
Household Positive Family type mixed results
PERIOD PREDICTORS
4000 104000 204000 304000 404000 504000 604000 704000 804000 904000 1004000 1104000 1204000 10 20 30 40 50 60 70 80 90Effect of income shared by the top 20%
PERIOD-INCOME EFFECTS
0 1000 2000 3000 4000 5000 6000 100 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 22000 24000 26000 28000 30000 32000 34000 36000 38000 40000 42000 44000 46000 50000 54000 60000 66000 72000 80000 en er gy con su m pt ionhousehold income & period inequality
inc_share20 inc_exp50