Yes. Consumption data are widely used as an efficient proxies of households’ incomes. Several studies use this approximation such as transfers targeting, demand systems, among others.
I think this will depend on the stage of your study. If it is still a proposal, the comments of Adrian, Mohamed, and Achille would most likely be accepted by your panel or sponsor. But if you have already the data collected based on the approved research problems (income and not Consumption Data one of them), it will be too late to replace. However, you can still do it by revising your research problems, related literature and methodology.(if your panel or sponsor approves it).
The issue of this question is whether consumption can be used as a substitute for income.
A short answer is “no.” It is not advisable to use consumption data as a substitute for income. The reasoning for this answer follows.
PURCHASING POWER AND DISPOSABLE INCOME
Traditionally, purchasing power may be equated with disable income. Under such an approach, it would be possible to substitute consumption for income. Assume that:
Y = income
D = deductions, i.e. tax, insurance, social security contribution, etc.: (d1, …, dn)
Y(d) = disposable income
S = savings
Then, disposable income may be construed as:
(1) Yd = Y – D – S
The ability for the agent to consume or the limit of his/her purchasing, i.e. purchasing power (P) is maximized at Yd or simply:
(2) P = Yd
However, advances in financial instruments, such as consumer finance and personal credit cards had made the classical approach to disposable income untenable. Yd above must be modified to accommodate this change. Assume that C = credit on a 30 days cycle. The revised purchasing power for time period t1, t2, … T with a corresponding purchasing power P(t1), P(t2), …, P(T) also changes accordingly. Under this circumstances, the condition P = Yd no longer hold because P > Yd.
To the extent that the extension of credit requires income (Y), there still is a relationship between Yd and P. However, the value of P no long is equal to Yd because with the extension of credit the agent now can purchase more, i.e. larger consumption then the agent’s actual disposable income.
The patterns of credit consumption also shows that credit consumers do not pay off 100% of the amount purchases in one payment, installment payment is used. Under this circumstance, continued spending would reach the credit limit of the credit extension. However, to creditor, i.e. credit card issuers, generally increases the consumer’s credit line in order to stimulate additional spending and partially payment, thus, creditors making profit from interest charge on outstanding balance. If this event repeats, the condition P = Yd will become widen.
The widening of P = Yd or P>Yd would differentiate the two arrays (purchasing power) and disposable income (Yd) in the time series t1, t2, … T “significantly.” In order to test this argument, one may employ the following Paired t-Test. The definition follows:
If the primary data of income is missing, perhaps secondary data may be used. A good source is the World Economic Report 2013 produced by the IMF. The data file in excel may be downloaded from this website:
Reliability is concerned with the reproducibility of the result of the prior study. In this case, the result may be difficult to reproduce by independent subsequent studies if (i) some data is missing and (ii) the imputation of consumption for income as equivalence where in fact these two data may not be substituted.
VALIDITY
Validity test is concerned with precision. In this case, there might be some difficulty in meeting this standard depending on the level of development in the country’s financial system. In a developed and developing countries where credit spending is a common practice, the substitution of consumption data for income would not be advisable as it would fail precision test. In a less developed economy, i.e. cash economy, the substitution might work---but with great caution because even in poorly develop economies deficit spending (not credit spending) would still produce a condition discussed above.
At business cycle frequencies, consumption and income are positively correlated and almost of same degree of variability in the US and I am pretty sure that is the case with OECD countries as well. This could be evoked as sufficient basis for replacing income data with consumption. But if the horizon of analysis is longer than that then consumption might be much more smoother than income and hence the substitution may not work.
Assume following facts of a person living in Massachusetts, USA:
Salary 5,000
Les deduction
Income tax: 27% (1,350)
Social security: 2.5%(125)
Other: IRA (500)
Total deductions (1975)
Adjusted income 3025
Less savings: 20% (605)
Disposable income 2420
In this case, the maximum that the person could spend is his/her disposable income in the amount of 2420 per month.
The situation would change if the person has a visa/mastercard for credit spending with a credit line of $10,000. Say … an average American packs 3 credit cards and would tend to use all of them. At $10,000 per card, 3 cards would add an additional $100,000 x 3 = 30,000 extra purchasing power in a form of credit spending. The current spending power now looks something like this:
Disposable income 2420
Credit available to spend 30,000
Purchasing power 32420
In this case, the maximum that the person could spend is his/her disposable income in the amount of 30,000 + 2420 = 32,420 per month. In the US, this may very well be the fact---so much so that the government has made Chapters 7, 9 and 11 in its bankruptcy law waiting to accommodate those who live their lives beyond their means.
Consumption level does not depend on income. Therefore, one cannot be a substitute for the other.
CORRELATION
By definition correlation is the measure for the level of relationship between two variables. A positive relationship only shows that the two variables are moving in the same direction. It does not mean that one may be substituted for the other. “A person has more income has a propensity to spend more”---thus goes the saying. However, this is not the same as saying that they are the same. To say so would be equivalent of saying that X is the same as Y. No, Y responds to changes in X; not that X and Y are the same. One is dependent variable and the other is independent variable. We keep them in separate coordinate precisely because they are not the same nor subsrtitutable.
normally, there is relatonship between income and consumption;
Generally, if I do research on household consumption, income are often less than consumption, and then ask the respondents again; it is difficult to get income sometimes, in this situation, I often use consumption rather than income to get some results. for example, demand income elastisity often is replaced by demand expenditure elastisity,
The use of consumption in place of income has been a standard practice in applied demand analysis for decades. As has been mentioned, it is generally thought to be less subject to error than disposable income, and, even in the absence of error, can be justified as a superior measure of permanent income in the Permanent Income Hypothesis and a better measure of lifetime income in the Life Cycle Model of Consumption. It was used by H. S. Houthakker and myself in all three editions of our book Consumer Demand in The United States (detailed justification for its use can be found in Chapter 6 of the Third Edition [Springer, 2010]), and more recently by myself in The Internal Structure of Consumption (Springer, 2013).
If you intend to explore well being of households, consumption expenditure is a better indicator than income for two reasons 1) it is theoretically closer to welfare 2) It is not strongly influenced by the strategic bias of the respondent - it is reliable.
Nice discussion. - Also, if you already have some income data and consumption data on the same respondents, you might want to regress one on the other. If you look at the scatterplot, that might help identify how reliable income has been, and a linear regression can give you the regression coefficient and its standard error to compare income and consumption. Doing this by stratum may be informative. Then if you go forward with just consumption, you may have more justification. The comments above regarding the measurement error on income sound logical, but if you already have the data to make this comparison, it may be useful. In fact, I suspect that several of those following this question may have done this already. (I worked for years with establishment surveys.)