segsell / elder_care Goto Github PK

See Adda et al (2017)

Estimate discrete distribution of parental age at agent's initial age when they enter the model (currently age 40).

Signed Integers:

np.int8: 8-bit signed integer, from -128 to 127.
np.int16: 16-bit signed integer, from -32,768 to 32,767.
np.int32: 32-bit signed integer, from -2,147,483,648 to 2,147,483,647.
np.int64: 64-bit signed integer, from -9,223,372,036,854,775,808 to 9,223,372,036,854,775,807.

Unsigned Integers:

np.uint8: 8-bit unsigned integer, from 0 to 255.
np.uint16: 16-bit unsigned integer, from 0 to 65,535.
np.uint32: 32-bit unsigned integer, from 0 to 4,294,967,295.
np.uint64: 64-bit unsigned integer, from 0 to 18,446,744,073,709,551,615.

• wage offer (include person and year fixed effects)

   ln(wage) = omega_0 + omega_1 * high_educ + omega_2 * exp + omega_3 * exp ** 2 + v
   
   Include interactions of exp and exp ** 2 with high_educ
  

• Job offer probability after unemployment

   job_offer = beta_0 + beta_1 * I(age >= 62) + beta_2 * high_educ + e 
  

Restrict caregivers on:

• having healthy spouse (if married/in partnership)
• being healthy themselves
• intra/extra household care? (sometimes parents co-inhabit caregivers home)

Incorporate the presence of siblings.

The care mix is the share of each caregiving mode for a given demographic, e.g. people in the age interval [70, 75).

Adjust the data from the statistical office to match the "restricted" child sample. Children providing informal care to anyone else except their parents are dropped.

The share of parental care in the informal care mix is about 50%.

Fix bug in the simulation function where for some cases jnp.mean throws an error. This affects the creation of asset, wage, and savings (rate) moments.

• Care Mix by parental health type and other parent alive (SHARE parent-child data set)

   e.g. share informal care by mother: 3 health states * 2 (father dead/alive) * 2 (siblings/sister  yes/no)
  

from OECD. See also here

(Old) Formula: 1 + 0.7 * x

where x is the number of additional persons in the household.

What about number of children in the household? Would require an additional state variable.

Due to RAM limitations of the GPUs available on Google Colab, discretize age into two-year age bins.
Consider switching to Google Cloud for more powerful GPUs.

Fix import and module paths.

outcome: parent_alive (by gender, i.e., mother and father separately)

features: age, age^2, health_state, interaction between age * health?

For now, use logit for tractability. Could use machine learning also.

For RAM reasons, reduce state space. Remove married state for now.