In the middle of last year (2023), two Geomarketing platforms vied for the title of “first to update” their database with information from the 2022 Census.
Even when, at that time, the information was still limited to the scale of municipalities and subject to change, as the IBGE itself warned.
.png)
It is worth noting that the text of the first one is more exaggerated, stating that Census 22 is already on the platform, while the other, more careful, correctly describes the data that was updated, but in the general plan it suggests that the data from the 2022 Census are on their platform.
Aside from this dispute, here at Mapfry, we continued with the projection from the 2010 Census, surprised by the creeping approach of our colleagues, who distorted complex concepts in a marketing race.
Such malice forced us to take a stand in order to avoid technically shallow movements.
We took the opportunity to stress the importance of accuracy with information by publishing the article:
Risks of using Census 22 information in advance
In August 2024, the IBGE released a revision of population estimates, something that was already foreseen in the methodology, but that surprised those in a hurry.
New census, but not much
At that time, the only information provided by the 2022 Census was that of Population and Households by Municipalities.
This metric is important to adjust the bases, but insufficient to be reflected in other dimensions such as Population by Age, Residents by Household, and Income.
The challenge is how to distribute the new numbers on a lower scale?
That is, disaggregating the information of a municipality into its neighborhoods and blocks without being aware of what the lower structures are like.
A simple way out is to make a linear inference, the famous rule of three, if the population increased by 2% in the municipality, it increases by 2% in all groups.
Considering that the greatest value of Geomarketing lies in its granularity, linear inference says nothing new, it only uses new information to adjust the 2010 Census numbers.
Which is valid and again in this text we will explain our methodology, but then calling this “Update of the new Census” is an exaggeration.
Setting aside that low mood, now, eight months later, the scenario is different.
High expectations
O 2022 Census information release schedule is closer to material deliveries for Geomarketing.
The first of these was the release of Census of the Geographic Coordinates of the Addresses.
This research, unprecedented in previous censuses, is the result of capturing georeferenced information from various moments of the research, from the routes taken to each visit to the addresses.
A monumental base of information about where things actually happened.
It will show us the places where there are new addresses, places where there are no addresses, the first point-by-point view of the new Census.
The most important release of a Census is data from the researched universe opened by Census Sectors.
When talking about the “Census in Sectors universe”, we are referring to detailed information about all the people and households surveyed.
Only at that time is it possible to have the full perspective of the quantitative and the qualitative in each location.
Census, user manual
Once we have the information about the universe, a process of comparing the information will begin.
Be very calm at this time, as we will be comparing a projection, which are super valid quantitative methods, with the research itself.
In comparative terms, a Census is worth many times more than the projection in terms of truth.
As we saw in the article A New Day, a New Census
.gif)
A new Census changes the game
It changes much more than that, the structures of the questionnaires, the design of the Census Sectors, the meaning of the answers change.
Comparing Censuses requires care.
It's like saying that the population of a municipality shrank when, in fact, a part of it was emancipated.
The oldest census we have, from 1872, counted 643 municipalities in Brazil, compared to 5,570 today.
How would you compare data from different periods in a way that reflects actual changes in population or social conditions, while eliminating or minimizing the effects of methodological or geographical variations?
To this end, there is the concept of areas that are minimally comparable in time, a sophisticated methodology of spatial equivalences throughout the numerous transformations in the areas of analysis.
Ensuring comparability over time involves the standardization of categories for the benefit of the consistency of the variables and the adaptation as a function of changes in geographical boundaries.
Comparing data also requires care
Over time, professions are created, technical courses are converted into higher education, the currency changed several times, inflation acted in waves.
It is not merely a logical-mathematical question to compare Censuses and other population information, it is necessary to have knowledge of recent history so that each information can be adjusted to its greatest informational value over time.
That's where the danger lies, that's where everyone fell beautiful
Demography is an essentially academic area of study, a master's degree attended by professionals with a strong statistical base and a lot of humanistic vision to understand people in numbers.
Our technical team is comprised of specialists in geo-demographic bases, including a master's degree in Demography from the University of Campinas (Unicamp).
Learn about some topics studied in the master's degree that are instruments for our methodologies:
- Current Demographic Issues - presentation of the diversity of interdisciplinary approaches to Demography
- Demographic Analysis - survey, validation, and normalization of data sources
- Demographic Analysis Laboratory - topics such as fecundity and mortality, mobility and population migration
- Studies on family and nuptiality - updating the concept of family based on data
- Geoprocessing applied to Social Sciences
- Population policies - How do governments use demography to design public policies
- Advanced topics in Demography - health and gender issues, history of populations and their heterogeneity
- Quantitative Methodology Applied to Population
- Databases in Demography
- Population projection using mathematical models - mathematical methods for modern demography, fertility modeling, mortality, migration, and demographic projections
Data as a Service
Geomarketing platforms fall into the Software as a Service (SaaS) category, but we offer more than systems, you receive market, population, and economic information.
Therefore, we also offer databases as services, since we do all the information management, from its basic structure to its reading and interpretation in the system (Data as a Service).
There are countless ways to create a computer system, different programming languages, databases, interfaces.
On the information side, there are not so many possibilities, and there is a gold standard consisting of continuously validated and established methodologies.
Understand a little more about the projection and comparison methodology
Let's think of a municipality that had 1,000 inhabitants and 300 households counted in the 2010 Census.
To make it easier, let's imagine that the main age and income groups are divided into quartiles, which are blocks of 25% each.
Then we would have 25% of children and adolescents, that is, 250 children and adolescents, 25% of young people, 25% of adults and middle age, and 25% of the elderly.
We would also have 25% of very low income, that is, 75 very low income families, 25% of low income, 25% of middle income, 25% of high income.
Five years after the 2010 Census, we decided to update this information in the projection.
To do this, we use a series of data sources, which are validated and normalized well in line with what was taught in the Demographic Analysis course.
Let's say that the data sources indicate that the population grew by 2% and now there are 1,020 inhabitants.
How to allocate these 20 new inhabitants among the households?
At the time of the 2010 Census, there were 1000 inhabitants living in 300 households, an average of 3.3 residents per household.
Challenges:
Should we allocate the new inhabitants to the same 300 households, increasing the average number of residents per household to 3.4, or should we increase the number of households to accommodate them while maintaining the average of 3.3?
Are these 1020 inhabitants the same as 1000 contacts before with the addition of 20 new ones, or were there deaths, births, and migrations during this period?
Does population growth occur in all age groups or are there some where it is more or less representative?
Can we maintain the distribution of income classes, or has it changed, perhaps with more households migrating from middle to low income?
Now that you've seen the danger, you can understand, perhaps better than many people who call themselves experts, that you can't use the rule of three in Demography.
The correct methodology will work with:
Sample and analytical weights
Information from surveys between Censuses, samples with small sizes, but proportionately similar to the whole, capable of mirroring their characteristics and thus expanding the sample information to the universe.
This expansion involves a tremendous risk of over-representing one phenomenon while underrepresenting others. Such an error can go unnoticed by data analysis systems, leading to the false impression that the model is representative of reality.
In order to avoid this scenario, the specialist will apply a countermeasure, which are the analytical weights, which define the expansion limits.
Having established these adjustments, we can process various models, scenarios and overlays, until we find a pattern composed of all the variables that allows us to estimate the possible fluctuations, while respecting local and regional dynamics.
Only then can we establish who the 20 new inhabitants are and where they are and all the changes with the other 1000.
Going forward, let's say that information from a new Census is released
They account for 980 inhabitants in this same municipality, a decrease of 4% compared to the last projection.
Having seen the methodology, you will apply the models and arrive at a distribution.
But that information is not, nor can it be called, a new Census.
This is just an update of the 2010 Census projection with some information from the new Census, still in structures from the 2010 Census.
One of the companies claimed to have prepared a comparative report between the 2010 Census and the 2022 Census 😱
Apart from the comparison between Population and Households, which were the only ones available by the IBGE at the time, and they're here in case you want to take a look, any other comparison would be between the 2010 Census and its projections.
The use of some data from the new Census in the projection does not justify saying that it is an update for the new Census.
Census is Census, projection is projection, OK?
Comparing Censuses
Remember the areas that are minimally comparable in time?
Let's use them here.
First of all, we have an advantage, if the number of municipalities grew almost 9 times between the Censuses of 1872 and the 2010 Census, the variation between 2010 and 2022 was minimal.
So the main challenge will be on the intra-municipal scale, comparing neighborhoods with neighborhoods, blocks with blocks at times as different as 2010 and 2022.
This is the main technical move we are making now, migrating the more granular information base, the 2010 Census Sectors to the 2022 Census Sectors.
The 2010 Census organized its information into 314,000 Sectors, while the 2022 Census was divided into 450,000 Sectors, an increase of 40%.
This is the first challenge, to migrate information from a base of 314 thousand Census Sectors to another 40% more fragmented one.
Therefore, the information that we are updating on the platform is a 2010 Census Projection with information from the 2022 Census and presented in the new Census Sectors structure.
This methodology was adopted in anticipation of the arrival of complete data by Census Sectors for the new Census, which will be the most important moment for us in recent years and will remain important for the next 10 years.
We are very excited about the arrival of this invaluable resource.
We are also pleased to have kept the methodology, maintained steadfastness while competitors accused¹ our base of being “out of date”.
It is now reserved for us to recognize that our information is at the highest level.
--------------
(1)
.png)