The Short-Run and Long-Run Determinants of Household Saving: Evidence from OECD Economies

Abstract

This paper examines the relationship between interest rates and household saving rates for an uneven panel of 19 OECD countries during the period 1995–2018. Unlike earlier studies, it uses the pooled mean group methodology to investigate which of the interest rate effects, income or substitution, dominates in the short run, long run, or both periods. With the baseline estimations, I find that the income effect outweighs the substitution effect in the short run, and vice versa in the long run. I also find that inflation (both expected and actual), household wealth through housing prices, unemployment rate, current taxes on income and wealth, and general government debt have significant negative impact on household saving in the long run. I find that financial development has a positive effect on household saving in the long run. Current taxes on income and wealth have a strong negative impact on household saving in the short run.

Appendix: Data Descriptions and Sources

Real Gross Saving Rate (in percent of gross disposable income)—$s_{i t}^{rg}$: Saving of households and non-profit institutions serving households (NPISH) without deducting consumption of fixed capital and deflated by the price deflator private final consumption expenditure. Source: AMECO database.

Real Net Saving Rate (in percent of gross disposable income)—$s_{i t}^{rn}$: Saving of households and non-profit institutions serving households (NPISH) after deducting consumption of fixed capital and deflated by the private final consumption expenditure. Source: AMECO database.

Short-term Nominal Interest Rate—$i_{i t}^{sn}$: Mainly policy interest rates or 3-months money market rates. Source: AMECO database.

Long-term Nominal Interest Rate—$i_{i t}^{ln}$: Mainly Central Government bonds of over 10 years. Source: AMECO database.

Long-term Real Interest Rate—$i_{i t}^{lr}$: Long-term nominal interest rate (iln) adjusted for price deflator private final consumption expenditure. Source: AMECO database.

Unemployment Rate—$u_{i t}$: Share of the total active population (labour force). Source: AMECO database.

Current Taxes on Income and Wealth—$t_{i t}$: Comprise taxes on income from employment, property, entrepreneurship, pensions, etc., including taxes deducted by employers (PAYE taxes) and other current taxes on capital, poll taxes, levied per adult or per household, independently of income or wealth, expenditure taxes, payable on the total expenditures of persons or households. Source: AMECO database.

General Government Gross Debt—$gd_{i t}$: General government net financial liabilities as percent of GDP. Source: AMECO database. Given the lack of general public debt data for some countries such as Switzerland, central government debt data is used as an alternative.

Domestic credit to private sector (as a percentage of GDP—$dcr_{i t}$): Financial resources provided to the private sector by financial corporations, such as through loans, purchases of non-equity securities, and trade credits and other accounts receivable, that establish a claim for repayment. Source: IMF, IFS, and World Bank and OECD GDP estimates.

Real Housing Price Index—$hpr_{i t}$: Nominal house price indices deflated by the consumer price index. Source: OECD Analytical database.

Expected Inflation—$\pi _{i t}^{e}$: Computed (on the assumption of adaptive expectations of price expectations) from the annual rate of change ($\Delta pcdef$) of the price deflator private final consumption expenditure (pcdef) in per cent per annum with adjustment weights ($\lambda$) varying from 0.1 to 1. Series with the best prediction of dpcd was used as expected rate of inflation:

$$\begin{aligned} \pi _{it}^{e} =\lambda \Delta pcdef_{it}+(1-\lambda ) \pi ^{e}_{_{it}-1} \end{aligned}$$

(14)

Rate of Inflation—$\pi _{i t}$: Computed as annual rate of change of the price deflator private final consumption expenditure (pcd) in percent per annum:

$$\begin{aligned} \pi _{it} =100.0 *((\text{ pcd }_{\mathrm{it}}/ {{\text {pcd}_\mathrm{it-1}}})-1) \end{aligned}$$

Similar results could be obtained with:

$$\begin{aligned} \pi _{it} =100.0 *(\log ({\text {pcd}}_{\mathrm{it}})-\log ({\text {pcd}}_{\mathrm{it}-1})) \end{aligned}$$

Source: Author’s computation.