diff --git a/pensionista/README.md b/pensionista/README.md
new file mode 100644
index 0000000..8a3013d
--- /dev/null
+++ b/pensionista/README.md
@@ -0,0 +1,20 @@
+# Dados de pensionistas
+
+## Instalação
+
+```shell
+pip install -r requirements.txt
+```
+
+## Execução
+
+Baixe os arquivos de pensionistas disponíveis [nesse
+site](http://transparencia.gov.br/download-de-dados/servidores) e coloque-os em
+`data/download/`. Depois, execute:
+
+```shell
+python convert.py <cadastro|observacao|remuneracao>
+```
+
+Os arquivos `cadastro.csv.gz`, `observacao.csv.gz` e `remuneracao.csv.gz` serão
+gerados em `data/output/`.
diff --git a/pensionista/convert.py b/pensionista/convert.py
new file mode 100644
index 0000000..a68c575
--- /dev/null
+++ b/pensionista/convert.py
@@ -0,0 +1,153 @@
+import csv
+import datetime
+import io
+from functools import lru_cache
+from pathlib import Path
+from uuid import NAMESPACE_URL, uuid5
+from zipfile import ZipFile
+
+from rows.fields import slug
+from rows.utils import CsvLazyDictWriter, open_compressed
+from tqdm import tqdm
+
+
+strptime = datetime.datetime.strptime
+
+@lru_cache(maxsize=1024 ** 2)
+def convert_number(value):
+    return value.replace(".", "").replace(",", ".")
+
+
+@lru_cache(maxsize=32 * 1024)
+def convert_date(value):
+    value = value.strip()
+    if not value:
+        return None
+    return str(strptime(value, "%d/%m/%Y").date())
+
+
+@lru_cache(maxsize=1024 * 1024)
+def person_uuid(cpf, name):
+    """Create UUID based on URLid methodology"""
+
+    if cpf is None:
+        cpf = "***********"
+    assert len(cpf) == 11, f"Invalid CPF: {repr(cpf)}"
+    internal_id = cpf[3:9] + "-" + slug(name).upper().replace("_", "-")
+    return str(uuid5(NAMESPACE_URL, f"https://id.brasil.io/person/v1/{internal_id}/"))
+
+
+@lru_cache(maxsize=128)
+def normalize_key(text):
+
+    text = text.replace("(R$)", "_brl_").replace("(U$)", "_usd_")
+    result = (
+        slug(text)
+        .replace("_registradas_em_sistemas_de_pessoal_", "_")
+        .replace("_programa_desligamento_voluntario_mp_792_2017_", "_deslig_voluntario_")
+    )
+    return result
+
+
+def convert_row(row):
+    new = {}
+    for original_key, value in row.items():
+        key = normalize_key(original_key)
+        value = value.strip()
+        if (value and value[0] == "0" and value[-1] == "0" and set(value) == {"0"}) or value in ("-", "--"):
+            value = None
+        if not key and not value:
+            continue
+
+        if key.startswith("data_") and value is not None:
+            value = convert_date(value)
+        elif value is not None and ("R$" in original_key or "U$" in original_key):
+            value = convert_number(value)
+
+        new[key] = value
+    return new
+
+
+def read_csv(fobj, table_name, year, month, input_encoding="iso-8859-1", delimiter=";"):
+    """Read binary `fobj` as CSV, convert each row, adding `table_name` as a column"""
+
+    fobj = io.TextIOWrapper(fobj, encoding=input_encoding)
+    reader = csv.DictReader(fobj, delimiter=delimiter)
+    for row in reader:
+        new = convert_row(row)
+        if "(*)" in new.get("ano", ""):  # Invalid row
+            continue
+        if "ano" not in new:
+            new["ano"] = year
+        if "mes" not in new:
+            new["mes"] = month
+        if "PENSIONISTA MENOR DE 16 ANOS" in new["cpf"]:
+            new["menor_16"] = True
+            new["cpf"] = None
+        else:
+            new["menor_16"] = False
+            new["cpf"] = new["cpf"].replace(".", "").replace("-", "")
+        new["sistema_origem"] = table_name
+        new["pessoa_uuid"] = person_uuid(new["cpf"], new["nome"])
+        if table_name == "cadastro":
+            for key in ("representante_legal", "instituidor"):
+                new[f"cpf_{key}"] = new[f"cpf_{key}"].replace(".", "").replace("-", "")
+                new[f"{key}_uuid"] = person_uuid(new[f"cpf_{key}"], new[f"nome_{key}"])
+        yield new
+
+
+def extract_year_month(filename):
+    """Extract year and month from ZIP filename"""
+
+    part = filename.name.lower().split(".zip")[0]
+    return int(part[:4]), int(part[4:6])
+
+
+def extract_origin_system(filename):
+    return filename.split(".zip")[0].split("_")[-1]
+
+
+if __name__ == "__main__":
+    import argparse
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("table_name", choices=("cadastro", "remuneracao", "observacao"))
+    args = parser.parse_args()
+
+    # Make sure all working paths exist before anything
+    DATA_PATH = Path(__file__).parent / "data"
+    DOWNLOAD_PATH = DATA_PATH / "download"
+    OUTPUT_PATH = DATA_PATH / "output"
+    for path in (DATA_PATH, DOWNLOAD_PATH, OUTPUT_PATH):
+        if not path.exists():
+            path.mkdir(parents=True)
+
+    # Create one compressed-CSV writer
+    filename = OUTPUT_PATH / f"pensionista_{args.table_name}.csv.gz"
+    fobj = open_compressed(filename, mode="w", buffering=8 * 1024 * 1024)
+    writer = CsvLazyDictWriter(fobj)
+
+    # Read each ZIP file, then each inner ZIP file, then filter desired
+    # inner-inner CSV file, convert it and write to the output CSV.
+    progress_bar = tqdm()
+    filenames = DOWNLOAD_PATH.glob("*.zip")
+    for filename in sorted(filenames, key=extract_year_month):
+        year, month = extract_year_month(filename)
+        progress_bar.desc = f"{year}-{month:02d}"
+        progress_bar.refresh()
+        zf = ZipFile(filename)
+        for fileinfo in zf.filelist:
+            origin_system = extract_origin_system(fileinfo.filename)
+            progress_bar.desc = f"{year}-{month:02d}/{origin_system}"
+            inner_zf = ZipFile(zf.open(fileinfo.filename))
+            for inner_fileinfo in inner_zf.filelist:
+                table_name = inner_fileinfo.filename.split(".")[0].split("_")[-1].lower().replace("observacoes", "observacao")
+                if table_name != args.table_name:  # We don't want this file
+                    continue
+                progress_bar.desc = f"{year}-{month:02d}/{origin_system}.{table_name}"
+                fobj = inner_zf.open(inner_fileinfo.filename)
+                reader = read_csv(fobj, origin_system, year, month)
+                for row in reader:
+                    writer.writerow(row)
+                    progress_bar.update()
+    progress_bar.close()
diff --git a/pensionista/download-old.sh b/pensionista/download-old.sh
new file mode 100755
index 0000000..4184d62
--- /dev/null
+++ b/pensionista/download-old.sh
@@ -0,0 +1,38 @@
+#!/bin/bash
+
+mkdir -p data/download
+cd data/download
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/PENSIONISTAS_112019.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/PENSIONISTAS_122019.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/PENSIONISTAS_012020.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/PENSIONISTAS_022020.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/PENSIONISTAS_032020.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/PENSIONISTAS_042020.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/1994.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/1995.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/1996.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/1997.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/1998.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/1999.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2000.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2001.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2002.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2003.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2004.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2005.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2006.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2007.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2008.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2009.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2010.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2011.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2012.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2013.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2014.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2015.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2016.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2017.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2018.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/2019.zip
+wget -c -t 0 http://repositorio.dados.gov.br/segrt/pensionistas/PENSIONISTAS_052020.zip
+cd -
diff --git a/pensionista/import-pgsql.sh b/pensionista/import-pgsql.sh
new file mode 100755
index 0000000..b3fcf1a
--- /dev/null
+++ b/pensionista/import-pgsql.sh
@@ -0,0 +1,16 @@
+#!/bin/bash
+
+if [ -z "$DATABASE_URL" ]; then
+	echo "ERROR: must set $DATABASE_URL with postgres connection string"
+	exit 1
+fi
+
+for table in cadastro observacao remuneracao; do
+	rows pgimport \
+		--dialect=excel \
+		--input-encoding=utf-8 \
+		--schema=schema/pensionista_${table}.csv \
+		data/output/pensionista_${table}.csv.gz \
+		$DATABASE_URL \
+		pensionista_${table}
+done
diff --git a/pensionista/indexes.sql b/pensionista/indexes.sql
new file mode 100644
index 0000000..6664f79
--- /dev/null
+++ b/pensionista/indexes.sql
@@ -0,0 +1,15 @@
+CREATE INDEX idx_pensobs_id ON pensionista_observacao (id_servidor_portal, ano, mes, sistema_origem);
+CREATE INDEX idx_pensobs_uuid ON pensionista_observacao (pessoa_uuid);
+CREATE INDEX idx_pensobs_orig ON pensionista_observacao (sistema_origem);
+
+CREATE INDEX idx_penscad_id ON pensionista_cadastro (id_servidor_portal, ano, mes, sistema_origem);
+CREATE INDEX idx_penscad_uuid1 ON pensionista_cadastro (pessoa_uuid);
+CREATE INDEX idx_penscad_uuid2 ON pensionista_cadastro (representante_legal_uuid);
+CREATE INDEX idx_penscad_uuid3 ON pensionista_cadastro (instituidor_pensao_uuid);
+CREATE INDEX idx_penscad_orig ON pensionista_cadastro (sistema_origem);
+
+CREATE INDEX idx_pensrem_id ON pensionista_remuneracao (id_servidor_portal, ano, mes, sistema_origem);
+CREATE INDEX idx_pensrem_uuid ON pensionista_remuneracao (pessoa_uuid);
+CREATE INDEX idx_pensrem_orig ON pensionista_remuneracao (sistema_origem);
+
+ALTER TABLE pensionista_cadastro ADD PRIMARY KEY (id_servidor_portal, ano, mes, sistema_origem);
diff --git a/pensionista/list_zips.py b/pensionista/list_zips.py
new file mode 100644
index 0000000..519e571
--- /dev/null
+++ b/pensionista/list_zips.py
@@ -0,0 +1,22 @@
+import json
+from urllib.parse import urljoin
+from urllib.request import urlopen
+
+
+def ckan_package_resources(base_url, resource_id):
+    template_url = urljoin(base_url, "/api/3/action/package_show?id={resource_id}")
+    url = template_url.format(resource_id=resource_id)
+    response = urlopen(url)
+    data = json.loads(response.read())
+    return data["result"]["resources"]
+
+
+if __name__ == "__main__":
+    resources = ckan_package_resources(
+        base_url="http://www.dados.gov.br", resource_id="c76a1bc6-2330-4b05-b3dd-491124931496"
+    )
+
+    for resource in resources:
+        if not resource["url"].lower().endswith(".zip"):
+            continue
+        print(resource["url"])
diff --git a/pensionista/requirements.txt b/pensionista/requirements.txt
new file mode 100644
index 0000000..dbf8108
--- /dev/null
+++ b/pensionista/requirements.txt
@@ -0,0 +1,2 @@
+calculadora-do-cidadao
+https://github.com/turicas/rows/archive/develop.zip