/istatapi

Python API for ISTAT (The Italian National Institute of Statistics)

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

istatapi is a Python interface to discover and retrieve data from ISTAT API (The Italian National Institute of Statistics). The library is designed to explore ISTAT metadata and to retreive data in different formats. istatapi is built on top of ISTAT SDMX RESTful API.

Whether you are an existing organization, a curious individual or an academic researcher, istatapi aims to allow you to easily access ISTAT databases with just a few lines of code. The library implements functions to:

  • Explore all available ISTAT datasets (dataflows in SDMX terminology)
  • Search available datasets by keywords
  • Retrieve information on a specific dataset like: the ID of the dataflow, the names and available values of the dimensions of the dataset, available filters.
  • Get data of an available dataset in a pandas DataFrame, csv or json format.

Install

You can easily install the library by using the pip command:

pip install istatapi

Tutorial

First, let’s simply import the modules we need:

from istatapi import discovery, retrieval
import matplotlib.pyplot as plt

With istatapi we can search through all the available datasets by simply using the following function:

discovery.all_available()
df_id version df_description df_structure_id
0 101_1015 1.3 Crops DCSP_COLTIVAZIONI
1 101_1030 1.0 PDO, PGI and TSG quality products DCSP_DOPIGP
2 101_1033 1.0 slaughtering DCSP_MACELLAZIONI
3 101_1039 1.2 Agritourism - municipalities DCSP_AGRITURISMO_COM
4 101_1077 1.0 PDO, PGI and TSG products: operators - municipalities data DCSP_DOPIGP_COM

You can also search for a specific dataset (in this example, a dataset on imports), by doing:

discovery.search_dataset("import")
df_id version df_description df_structure_id
10 101_962 1.0 Livestock import export DCSP_LIVESTIMPEXP
47 139_176 1.0 Import and export by country and commodity Nace 2007 DCSP_COEIMPEX1
49 143_222 1.0 Import price index - monthly data DCSC_PREIMPIND

To retrieve data from a specific dataset, we first need to create an instance of the DataSet class. We can use df_id, df_description or df_structure_id from the above DataFrame to tell to the DataSet class what dataset we want to retrieve. Here, we are going to use the df_id value. This may take a few seconds to load.

# initialize the dataset and get its dimensions
ds = discovery.DataSet(dataflow_identifier="139_176")

We now want to see what variables are included in the dataset that we are analysing. With istatapi we can easily print its variables (“dimensions” in ISTAT terminology) and their description.

ds.dimensions_info()
dimension dimension_ID description
0 FREQ CL_FREQ Frequency
1 MERCE_ATECO_2007 CL_ATECO_2007_MERCE Commodity Nace 2007
2 PAESE_PARTNER CL_ISO Geopolitics
3 ITTER107 CL_ITTER107 Territory
4 TIPO_DATO CL_TIPO_DATO12 Data type 12

Now, each dimension can have a few possible values. istatapi provides a quick method to analyze these values and print their English descriptions.

dimension = "TIPO_DATO" #use "dimension" column from above
ds.get_dimension_values(dimension)
values_ids values_description
0 EV export - value (euro)
1 TBV trade balance - value (euro)
2 ISAV import - seasonally adjusted value - world based model (millions of euro)
3 ESAV export - seasonally adjusted value - world based model (millions of euro)
4 TBSAV trade balance - seasonally adjusted value -world based model (millions of euro)
5 IV import - value (euro)

If we do not filter any of our variables, the data will just include all the possible values in the dataset. This could result in too much data that would slow our code and make it difficult to analyze. Thus, we need to filter our dataset. To do so, we can simply use the values_ids that we found using the function get_dimension_values in the cell above.

Note: Make sure to pass the names of the dimensions in lower case letters as arguments of the set_filter function. If you want to filter for multiple values, simply pass them as lists.

freq = "M" #monthly frequency
tipo_dato = ["ISAV", "ESAV"] #imports and exports seasonally adjusted data
paese_partner = "WORLD" #trade with all countries

ds.set_filters(freq = freq, tipo_dato = tipo_dato, paese_partner = paese_partner)

Having set our filters, we can now finally retrieve the data by simply passing our DataSet instance to the function get_data. It will return a pandas DataFrame with all the data that we requested. The data will be already sorted by datetime

trade_df = retrieval.get_data(ds)
trade_df.head()
DATAFLOW FREQ MERCE_ATECO_2007 PAESE_PARTNER ITTER107 TIPO_DATO TIME_PERIOD OBS_VALUE BREAK CONF_STATUS OBS_PRE_BREAK OBS_STATUS BASE_PER UNIT_MEAS UNIT_MULT METADATA_EN METADATA_IT
0 IT1:139_176(1.0) M 10 WORLD ITTOT ESAV 1993-01-01 10767 NaN NaN NaN NaN NaN NaN NaN NaN NaN
368 IT1:139_176(1.0) M 10 WORLD ITTOT ISAV 1993-01-01 9226 NaN NaN NaN NaN NaN NaN NaN NaN NaN
372 IT1:139_176(1.0) M 10 WORLD ITTOT ISAV 1993-02-01 10015 NaN NaN NaN NaN NaN NaN NaN NaN NaN
4 IT1:139_176(1.0) M 10 WORLD ITTOT ESAV 1993-02-01 10681 NaN NaN NaN NaN NaN NaN NaN NaN NaN
373 IT1:139_176(1.0) M 10 WORLD ITTOT ISAV 1993-03-01 9954 NaN NaN NaN NaN NaN NaN NaN NaN NaN

Now that we have our data, we can do whatever we want with it. For example, we can plot the data after having it cleaned up a bit. You are free to make your own analysis!

# set matplotlib themes
plt.style.use('fivethirtyeight')
plt.rcParams['figure.figsize'] = [16, 5]

#fiveThirtyEight palette
colors = ['#30a2da', '#fc4f30', '#e5ae38', '#6d904f', '#8b8b8b']

# calculate moving averages for the plot
trade_df["MA_3"] = trade_df.groupby("TIPO_DATO")["OBS_VALUE"].transform(
    lambda x: x.rolling(window=3).mean()
)

#replace the "TIPO_DATO" column values with more meaningful labels
trade_df["TIPO_DATO"] = trade_df["TIPO_DATO"].replace(
    {"ISAV": "Imports", "ESAV": "Exports"}
)

# Plot the data
after_2013 = trade_df["TIME_PERIOD"] >= "2013"
is_ESAV = trade_df["TIPO_DATO"] == "Exports"
is_ISAV = trade_df["TIPO_DATO"] == "Imports"

exports = trade_df[is_ESAV & after_2013].rename(columns={"OBS_VALUE": "Exports", "MA_3": "Exports - three months moving average"})
imports = trade_df[is_ISAV & after_2013].rename(columns={"OBS_VALUE": "Imports", "MA_3": "Imports - three months moving average"})

plt.plot(
    "TIME_PERIOD",
    "Exports",
    data=exports,
    marker="",
    linestyle="dashed",
    color = colors[0],
    linewidth=1
)
plt.plot(
    "TIME_PERIOD",
    "Imports",
    data=imports,
    marker="",
    linestyle="dashed",
    color = colors[1],
    linewidth=1
)
plt.plot(
    "TIME_PERIOD",
    "Exports - three months moving average",
    data=exports,
    color = colors[0],
    linewidth=2
)
plt.plot(
    "TIME_PERIOD",
    "Imports - three months moving average",
    data=imports,
    marker="",
    color = colors[1],
    linewidth=2
)

# add a title
plt.title("Italy's trade with the world")

# add a label to the x axis
plt.xlabel("Year")

# turn y scale from millions to billions (divide by a 1000), and add a label
plt.ylabel("Value in billions of euros")
plt.gca().yaxis.set_major_formatter(plt.FuncFormatter(lambda x, loc: "{:,}".format(int(x/1000))))
plt.legend()

With just a few lines of code, we have been able to retrieve data from ISTAT and make a simple plot. This is just a simple example of what you can do with istatapi. You can find more examples in the _examples folder. Enjoy!